Optimal Operation for Economic and Exergetic Objectives of a Multiple Energy Carrier System Considering Demand Response Program

Article Optimal Operation for Economic and Exergetic Objectives of a Multiple Energy Carrier System Considering Demand Response Program

Yu Huang, Shuqin Li, Peng Ding, Yan Zhang *, Kai Yang and Weiting Zhang

Department of Automation, North China Electric Power University, Baoding 071003, China; huangyuﬁ[email protected] (Y.H.); [email protected] (S.L.); [email protected] (P.D.); [email protected] (K.Y.); [email protected] (W.Z.) * Correspondence: [email protected]

Received: 4 August 2019; Accepted: 18 October 2019; Published: 21 October 2019

Abstract: An MECS (multiple energy carrier system) could meet diverse energy needs owing to the integration of different energy carriers, while the distinction of quality of different energy resources should be taken into account during the operation stage, in addition the economic principle. Hence, in this paper, the concept of exergy is adopted to evaluate each energy carrier, and an economic–exergetic optimal scheduling model is formulated into a mixed integer linear programming (MILP) problem with the implementation of a real-time pricing (RTP)-based demand response (DR) program. Moreover, a multi-objective (MO) operation strategy is applied to this scheduling model, which is divided into two parts. First, the ε-constraint method is employed to cope with the MILP problem to obtain the Pareto front by using the state-of-the-art CPLEX solver under the General Algebraic Modeling System (GAMS) environment. Then, a preferred solution selection strategy is introduced to make a trade-off between the economic and exergetic objectives. A test system is investigated on a typical summer day, and the optimal dispatch results are compared to validate the effectiveness of the proposed model and MO operation strategy with and without DR. It is concluded that the MECS operator could more rationally allocate different energy carriers and decrease energy cost and exergy input simultaneously with the consideration of the DR scheme.

Keywords: multiple energy carrier system; ε-constraint method; exergy; optimal operation; demand response

1. Introduction At present, to realize the sustainable development of society, many scholars have been paying attention to energy savings and eﬃciency improvements due to severe fossil energy crises and environmental problems [1]. In such context, an MECS (multiple energy carrier system), which was formulated in a project “Vision of Future Energy Networks” at ETH Zurich together with other partners, has been regarded as a promising approach for next-generation energy systems [2]. An MECS refers to a system coupled with diverse kinds of energy carriers, such as electricity, gas, heat, and cooling to achieve coordinate energy production and delivery, as well as to meet various energy consumptions in coordination [3]. Modeling and optimal operation are the two key issues in achieving the expected potentials of an MECS. The basic concept of an energy hub (EH) has been introduced from a global perspective to build up the model of the MECS [4], where the core of the EH is a coupling matrix description. A standardized matrix modeling method was proposed, and the coupling matrix was automatically built based on graph theory and the Gaussian elimination technique, wboth hich could be conveniently

Energies 2019, 12, 3995; doi:10.3390/en12203995 www.mdpi.com/journal/energies Energies 2019, 12, 3995 2 of 21 implemented in a computer to facilitate the modeling effort [5]. Another standardized modeling approach for a complex EH model was introduced in [6]. The coupling matrix of complex EH model was divided into several simple EH models based on nodal arrangement and virtual node insertion methods. Though most researchers have adopted the EH concept to model an MECS [7,8], some scholars have formulated the model from local perspectives. The energy conversion and coupling relationship are separately described for each device and energy carrier, with obvious an distinction compared to the EH concept, which is utilized to establish the relationship of the whole system from a global perspective. The device models for photovoltaic (PV), electrical energy storage (ESS), combined heat and power (CHP) plant, gas boiler (GB), heat pump (HP), and thermal energy storage (TES) have been developed with input–output characteristics, capacity constraints, ramp rate limits and other conditions [9]. In addition, models for each energy carrier (electricity, gas, and heat) have been proposed based on the energy balance principle in the MECS [9,10]. Based on MECS modeling, many scholars have focused on optimal operation strategy to realize rational energy distribution and efficient energy management. For minimizing the day-ahead operation cost, a mixed integer nonlinear programming (MINLP) approach was employed to solve the optimal scheduling problem of building an energy system, which is coupled with electricity and cooling under dynamic electricity pricing [11]. In [12], operation strategies for both grid-connected and islanded modes of a multi-energy microgrid were implemented by the CPLEX solver in General Algebraic Modeling System (GAMS) with the aim of minimizing net operating cost. A multi-objective (MO) optimal electrical and thermal energy management for a residential EH was conducted to acquire the Pareto front when considering the environmental emissions and total energy cost, and the fuzzy method was employed to determine the best possible solution on the Pareto front [13]. However, most researchers have focused on the EH model and the single-objective optimal economic operation in the MECS, in which the local relation of each component might not be adequately considered; additionally the operation goal neglects the characteristics of different energy carriers. To realize the rational utilization of energy in an MECS, the energy quality should be taken into account in the operation strategy, and the measurement of energy quality for different energy carriers needs to be conducted. Exergy, derived from the second law of thermodynamics, has been employed as a promising tool to analyze energy quality. It is the maximum theoretical work obtainable from the system as it reaches the equilibrium state with its surrounding environment, the so-called “reference environment” [14,15]. Exergy could be regarded as the usable part of energy, and the other part which cannot be utilized is called anergy [16]. Some scholars have conducted considerable research about the exergy issue. An exergetic analysis and a performance evaluation were conducted on some renewable energy resources, including solar energy systems, wind energy systems, geothermal energy systems, and biomass systems [17]. Energy, exergy and environmental analyses of a hybrid combined cooling heating and power (CCHP) system driven by biomass and solar energy have been studied, and the complementarity of the two kinds of energy has been analyzed [18]. To determine the best design parameters of a polygeneration system which was composed of gas turbine cycle, a Rankine cycle, an absorption cooling cycle, and domestic hot water, a multi-objective optimization approach was applied based on a genetic algorithm considering the total cost and exergy efficiency as objective functions [19]. Most studies have concentrated on the system design and performance evaluation of different energy systems using the exergy method, but few researchers have contributed to the optimal operation strategy of the MECS. In [20], the concept of exergy hub, which is analogous to EH, was proposed, and three single-objective optimal dispatch cases of MECSs were investigated for exergetic efficiency maximization, energetic efficiency maximization, and cost minimization. A multi-objective mixed integer linear programming (MOMILP) problem for the operation optimization of a distributed energy system was solved, and the weighted sum approach was employed to convert the MOMILP problem to a single-objective MILP according to [21]. However, these studies for optimal operation lack flexibility, and the interaction between the MECS and load demand side are inadequately considered. Energies 2019, 12, 3995 3 of 21

Demand response (DR) could contribute to the rational use of electricity by realizing the peak load shifting and curtailment, benefitting both grid utility and customers. Generally, DR is divided into two categories, price-based DR (PBDR) programs and incentive-based DR (IBDR) programs [22]. In PBDR programs, dynamic electricity pricing is employed to decrease the peak load and increase the valley load, where three major cases are developed—time of use (TOU), critical peak pricing (CPP) and real-time pricing (RTP). On the other hand, direct load control (DLC), emergency demand response program (EDRP), capacity market program (CAP) and other terms are included in IBDR, where participants could receive participation payments [23]. Among these approaches, the RTP program might be the most direct and efficient DR program, a statement supported by by many economists [24]. Many works in the literature about the RTP-based DR program have investigated power systems to measure the economic benefit with a deterministic or stochastic price. A smart home energy system (HES) structure was proposed with a deterministic time-varying price signal in [25]. To realize the optimal energy management for residential appliances, the Monte Carlo simulation and robust optimization technique were adopted to address the uncertainties of RTP in PBDR in [26]. Because of the successful implementation of DR in the power system, some farsighted scholars introduced it into an MECS to study their interaction. Three scenarios were conducted—PBDR programs (TOU, CPP, and TOU and CPP), IBDR programs (EDRP and CAP) and a combination between the PBDR and IBDR programs (TOU and CAP)—to simultaneously minimize EH operation cost and maximize customer benefit, according to [27]. In [28], an RTP model was developed that was calculated by the predicted electricity demand and TOU pricing, and it was implemented in a micro energy grid which was described by the EH concept. However, DR programs are not adequately considered in the operation strategy with the exergetic objective, and the interaction between DR and multi-objectives has been less investigated by most research. In this paper, the main goal was to realize rational energy distribution with the principles of simultaneously decreasing the operation cost and minimizing the exergy input for an MECS. A grid-connected MECS was built to meet the diverse energy needs of loads. To obtain an optimal energy management strategy, an optimal MO scheduling model considering both the economic and exergetic objectives was introduced from the local perspectives with the implementation of the RTP-based DR scheme, which could flatten the electricity load curve. The Pareto front was obtained by the ε-constraint approach to the optimal MO scheduling model, and the compromise solution on the Pareto front was acquired by the approach of linear programming technique for a multidimensional analysis of preference (LINMAP). Therefore, the novelty and contribution of this paper could be summarized as follows:

1. The MO optimal scheduling model of an MECS with economic and exergetic objectives are conducted and formulated into an MILP problem. 2. To evaluate the performance of the DR program in an MECS, an RTP-based DR model is introduced with a dynamic pricing rate. 3. An MO operation strategy is developed where the ε-constraint method is utilized to obtain the Pareto front, and the LINMAP approach is introduced to make a trade-oﬀ between the economic and exergetic objectives.

In the following, the economic–exergetic scheduling model of an MECS is introduced, and the RTP-based DR scheme is formulated in Section2. In Section3, the MO operation strategy is presented to obtain the Pareto front, and a trade-oﬀ is conducted between these two diﬀerent objectives. Case studies are conducted in Section4, and the conclusion for this paper is summarized in Section5.

2. Problem Formulation The energy generation, transmission, coupling and consumption of diﬀerent energy carriers such as coal, solar, natural gas, electricity, cooling, and heat are described in Figure1 to fulﬁll the diverse energy demands of this paper. Traditionally, primary energy resources have been utilized through Energies 2019, 12, 3995 4 of 21

Energiesenergy 2019 generation, 12, x FOR andPEERenergy REVIEWtransmission, and the energy coupling process have not really4 beenof 21 taken into account in the energy supply chain. Therefore, the energy coupling segment, which is whichregarded is regarded as an MECS, as an is focusedMECS, is on focused in this paper.on in this Some paper. energy Some conversion energy conversion devices are devices considered are consideredto realize the to transformationrealize the transfor betweenmation diff betweenerent energy different carriers, energy such ca asrriers, a photovoltaic such as a photovoltaic (PV), a solar (PV),thermal a solar collector thermal (STC), collector a combined (STC), heat a andcombined power (CHP),heat and a gas power boiler (CHP), (GB), an a electricalgas boiler chiller (GB), (EC), an electricalan absorption chiller chiller (EC), (AC) an andabsorption a heat exchanger chiller (AC) unit (HEU).and a heat In addition, exchanger energy unit storage (HEU). systems In addition, (ESSs), energysuch as storage cooling energysystems storage (ESSs), (CES) such andas cooling thermal energy energy storage storage (CES) (TES), and are wellthermal considered energy tostorage boost (TES),the flexibility are well of considered each energy to carrier.boost the The flexibilit MO optimaly of each operation energy modelcarrier. of The a grid-connected MO optimal operation MECS is modelformulated of a grid-connected in this section, which MECS is is divided formulated into threein this parts. section, In Section which 2.1 is ,divided the economic into three and exergeticparts. In Subsectionobjective models 2.1, the are economic formulated. and Theexergetic operation objective model mode for eachls are energy formulated. device, Th whiche operation contains model equality for eachconstraints energy and device, inequality which constraints,contains equality is developed constraints in Section and inequality 2.2. To measure constraints, the eff isect developed of DR for anin SubsectionMECS, the RTP2.2. modelTo measure for the the DR effect scheme of isDR introduced for an MECS, in Section the RTP2.3. model for the DR scheme is introduced in Subsection 2.3.

Figure 1. AnAn MECS MECS (multiple (multiple energy energy carrier carrier system system)) coupled with diverse energy carriers.

2.1. Objectives Objectives

2.1.1. Economic Economic Objective Objective In this study,study, thethe electricity electricity was was not not permitted permitted to to be be sold sold back back to theto the utility utility grid, grid, which which meant meant that thatbi-directional bi-directional power power ﬂow flow was was not considered.not considered. The Th energye energy cost cost of an of MECS an MECS is the is aggregationthe aggregation of two of twoterms: terms: the costthe ofcost electricity of electricity bought bought from thefrom grid the and grid the and cost the of cost natural of natural gas bought gas bought from the from natural the naturalgas pipeline; gas pipeline; this relationship this relationship is described is described as follows: as follows:

2424 X e fctPtctLHVGtt=+()()e () () ()/36 . Δ f1 =1 ce(egridgt)P (t) + cg(t)LHV gasggasGg(t)/3.6 ∆ t (1)(1) t=1 grid t=1

ct() ctg () where f1 is the economic objective, ( e) and( ) are the electricity and gas pricing at periode ( t), where f 1 is the economic objective, ce t and cg t are the electricity and gas pricing at period t, Pgrid t e Ptgrid () G (t) Gtg () is the power is the purchased power purchased from the utilityfrom the grid, utilityg grid,is the natural is gasthe boughtnatural fromgas bought the gas from pipeline, the andgas LHVgas is the lowerLHV heat value (LHV) of natural gas. pipeline, and gas is the lower heat value (LHV) of natural gas. 2.1.2. Exergetic Objective 2.1.2. Exergetic Objective Due to the coupling of different energy carriers, energy efficiency might not be the best evaluation index,Due since to itthe emphasizes coupling energyof different quantities energy but carriers, neglects energy energy qualities.efficiency Tomight this end,not maximizingbe the best evaluationexergy efficiency index, of since the whole it emphasizes day is a better energy choice quanti forties an evaluationbut neglects index energy in the qualities. day-ahead To operationthis end, maximizingof an MECS, exergy which isefficiency formulated of as:the whole day is a better choice for an evaluation index in the day-ahead operation of an MECS, which is formulated as: X24 X24 24 24 ψex = Exout(t)∆t/ Exin(t)∆t (2) ψ =ΔEx() t t / Ex () t Δ t (2) ext=1 outt=1 in tt==11 where ψex is the exergy efficiency and Exout () t and Exin () t are the exergy output and input of an MECS, respectively. According to Figure 1, the exergy used is equal to the exergy needed from the energy consumption segments and is presented as:

Energies 2019, 12, 3995 5 of 21

where ψex is the exergy eﬃciency and Exout(t) and Exin(t) are the exergy output and input of an MECS, respectively. According to Figure1, the exergy used is equal to the exergy needed from the energy consumption segments and is presented as:

e c h Exout(t) = Exout(t) + Exout(t) + Exout(t) (3)

e c h where Exout(t), Exout(t), and Exout(t) are the electricity, cooling, and heat exergy output of an MECS, respetively. Electricity energy is pure exergy according to [16], because it can be completely converted to work. Cooling and heat exergy are tightly related to the required temperature of demands and reference temperature, which generally adopt ambient temperature. Energy flowing with different temperatures means different abilities to work; hence, the temperature factor must be taken into account to formulate cooling and heat exergy. The exergy formulation for each energy carrier is introduced as follows:

e Exout(t) = Le(t) (4)

c c Ex (t) = Ta(t)/T 1 Lc(t) (5) out req − h h Ex (t) = 1 Ta(t)/T L (t) (6) out − req h where Le(t), Lc(t), and Lh(t) are the load demands of electricity, cooling, and heating at period t, c h respectively; Ta(t) is the ambient temperature; Treq and Treq are the required temperatures of the cooling and heat loads, respectively. However, it must be noted that diverse energy demands could be obtained by load forecast techniques with high precision, and total exergy needs could be calculated by forecasted load proﬁles. Therefore, the exergetic objective could be converted from maximizing exergy eﬃciency to minimizing the exergy input of the MECS according to Equation (2), which is formulated as follows:

X24 = co( ) + s ( ) + g ( ) f2 Exin t Exin t Exin t (7) t=1

co( ) s ( ) g ( ) where f2 is the exergetic objective; Exin t , Exin t , and Exin t are the exergy input of each primary energy carriers of coal, solar and gas, respectively. In this paper, the electricity on the grid was assumed to be generated from the power plant, the energy of which was derived from coal, as shown in Figure1. The exergy in solar energy has been studied in many papers, including various approaches reported in [29,30]. Among these, in this paper, the exergy-to-energy ratio of radiation developed by Petela was adopted to describe solar exergy, which is regarded as thermal radiation at the sun temperature [15,17]. Chemical-specific exergy is often evaluated by the LHV and exergy factor of the energy carrier [20], and the exergy of natural gas could be formulated by specific exergies and mass flow rates. Therefore, the exergy rate for each primary energy carrier could be expressed as:

 co e p  Ex (t) = Ex (t)/ψex  in in (8)  e ( ) = e ( )  Exin t Pgrid t

( s ( ) = ( )( + ) ( ) Exin t I t SSTC SPV γ t /1000 1 4 4 (9) γ(t) = 1 + (Ta(t)/Tsun) (Ta(t)/Tsun) 3 − 3 ( πg = LHVgasξgas (10) g ( ) = ( ) Exin t πgGg t /3.6 Energies 2019, 12, 3995 6 of 21

p where ψex is the exergy eﬃciency of the power plant; I(t) is the solar radiation; SSTC and SPV are the installed area of the STC and the PV; γ(t) is the exergy-to-energy ratio of solar energy; Tsun is the temperature of the Sun; πg and ξgas are the speciﬁc exergy and exergy factor of natural gas, respectively.

2.2. Constraints The establishment of proper constraints is a vital issue to rationally allocate different energy flows on the premise of fulfilling diverse energy needs and guaranteeing the security of energy devices in an MECS. Therefore, relative equality and inequality constraints must be taken into account for each energy carrier and energy device, as shown in Figure1. Balancing electricity, heat, cooling and natural gas should be considered to meet energy demands and conduct energy distribution between each energy device in the MECS. For every energy equipment, balance and restrictions must be simultaneously involved. Moreover, for the sake of intuitively reflecting the coupling relationship between energy equipment and energy flow, the modeling approach was employed from local perspectives in this paper.

2.2.1. Energy Flows Constraints Balancing electricity, heat, cooling and natural gas should be considered to meet the energy demands and conduct energy distribution between each energy device in an MECS. The electricity needs of loads and the EC must be jointly satisfied by a PV, a CHP, and a grid at each time interval. Heat and cooling loads are satisfied by an HEU, chilling devices, and ESS, which could adjust the balance of supply and demand and increase the flexibility of the MECS. In addition, in the MECS, the heat flow distribution among the STC, the CHP, the GB, the AC, and the HEU should be considered. Since there is no natural gas need in the energy demands, only the allocation relationship between the CHP and the GB just be taken into account. Therefore, the energy balance model for each energy carrier is addressed as follows:

( ) + e ( ) = e ( ) + e ( ) + e ( ) Le t PEC t Pgrid t PPV t PCHP t (11)

L (t) = Ph,out (t) + Ph,dch(t) Ph,ch (t) (12) h HEU TES − TES c c c,dch c,ch Lc(t) = P (t) + P (t) + P (t) P (t) (13) EC AC CES − CES h ( ) + h,in ( ) = h ( ) + h ( ) + h ( ) PAC t PHEU t PSTC t PCHP t PGB t (14)

Gg(t) = GGB(t) + GCHP(t) (15) e ( ) e ( ) e ( ) where PEC t is the electricity need of the EC at period t; PPV t and PCHP t are the electricity generation h,in ( ) h,out h,dch( ) of the PV and the CHP; PHEU t and PHEU are the heat input and output of the HEU; PTES t and h,ch ( ) c ( ) c ( ) PTES t are the heat discharg and charge of TES; PEC t and PAC t are the cooling generated by the EC c,dch( ) c,ch ( ) h ( ) and the AC; PCES t and PCES t are the cooling discharged and charged of CES; PAC t is the heat h ( ) h ( ) h ( ) consumption of the AC; PSTC t , PCHP t and PGB t are the heat generated by the STC, the CHP and ( ) ( ) the GB, respectively; and GGB t and GCHP t are the gas demands of the GB and the CHP, respectively.

2.2.2. Energy Device Constraints In the MECS shown in Figure1, the PV and the STC are regarded as uncontrollable devices, and their energy output rates depend on solar irradiance and other conditions, such as the conversion eﬃciency of the device and time-varying ambient temperature. Therefore, according to [13,15], the electricity and heat output rates of the PV and the STC could be formulated as:

e P (t) = ηPVSPVI(t)(1 0.005(Ta(t) 298.15))/1000 (16) PV − − h PSTC(t) = ηSTCSSTCI(t)/1000 (17) Energies 2019, 12, 3995 7 of 21

where ηPV and ηSTC are the energy conversion efficiency of the PV and the STC, respectively. The CHP unit is employed to realize the coupling of electricity, heat, and natural gas, which is vital in the optimal operation of an MECS due to its controllability. Hence, the capacity and ramp limit issues must be taken into account in the operating model. The constant efficiency model was adopted to describe the coupling relation of different energy carriers according to [13,14], and the 0–1 variable was introduced to describe the start/stop state because the electricity rate of the CHP could not be decreased to zero due to the technique restrictions, as shown in the following:

 e e  P (t) = η GCHP(t)LHVgas/3.6  CHP CHP  Ph (t) = h G (t)LHV  CHP ηCHP CHP gas/3.6  e e e (18)  uCHP(t)P P (t) uCHP(t)P  CHP,min ≤ CHP ≤ CHP,max  RPe Pe (t + ∆t) Pe (t) RPe CHP,min ≤ CHP − CHP ≤ CHP,max

e h e where ηCHP and ηCHP are the electricity and heat efficiency of the CHP, respectively; PCHP,min and e ( ) PCHP,max are the lower and upper limits of capacity, respectively; uCHP t is a binary variable; e e 1/0 denotes the start/stop state; RPCHP,min and RPCHP,max represent the ramp-down and ramp-up bounds, respectively. The GB consumes natural gas to generate heat which could be utilized in the AC and the HEU to meet the cooling and heat demands, and the energy balance and capacity restrictions should be considered:  h  P (t) = η GGB(t)LHVgas/3.6  GB GB (19)  u (t)Ph Ph (t) u (t)Ph GB GB,min ≤ GB ≤ GB GB,max ( ) h h where ηGB is the heat efficiency of the GB; uGB t represents the start/stop state; PGB,min and PGB,max are, respectively, the lower and upper limits of the GB. High-temperature heat flow from the STC, the CHP, and the GB need to be exchanged with low-temperature heat in the HEU according to the required temperature of the user. In this paper, it was assumed that the HEU had enough capacity to consume heat to avoid heat waste. Due to heat loss in the heat exchange process, the efficiency of the HEU was developed to establish the conversion model, i.e.: h,out ( ) = h,in ( ) PHEU t ηHEUPHEU t (20) where ηHEU is the efficiency of the HEU. To fulfill the cooling demands, the EC and AC devices are included in this paper by consuming different energy carriers. The EC needs electricity, which is a kind of high-quality energy to drive, and the AC takes advantage of low-quality heat to generate cooling energy; their energy conversion relationship can be formulated by the coefficient of performance (COP), which is regarded as a constant in this paper, as follows:   Pc (t) = COP Pe (t)  EC EC EC (21)  u (t)Pc Pc (t) u (t)Pc EC EC,min ≤ EC ≤ EC EC,max   Pc (t) = COP Ph (t)  AC AC AC (22)  u (t)Pc Pc (t) u (t)Pc AC AC,min ≤ AC ≤ AC AC,max c c where COPEC and COPAC are the COP of the EC and the AC, respectively; PEC,min and PEC,max are the c c lower and upper bounds of the EC; PAC,min and PAC,max are the lower and upper bounds of the AC; and uEC(t) and uAC(t) denote the on/off states of the EC and the AC, respectively, at period t. To further improve the flexibility of the MECS, TES and CES were introduced to achieve the energy shift during the day with three modes of operation: charging, discharging and still. The general model for ESS was adopted. The energy stored in ESS at time interval t depends on the energy stored at the previous time step as well as the energy variation, both of which directly relate to the mode selection. The energy dissipation rate and charging/discharging efficiency were applied to the ESS Energies 2019, 12, 3995 8 of 21 model. Furthermore, the capacity of ESS and the charging/discharging restrictions must be taken into account. Therefore, the models for TES and CES were concluded as:   E (t) = (1 δ )E (t ∆t) + ηch Pch (t) Pdch (t)/ηdch ∆t  TES TES TES TES TES TES TES  min − max − −  E ETES(t) E  TES ≤ ≤ TES  uch (t)Pch Pch (t) uch (t)Pch (23)  TES TES,min TES TES TES,max  dch dch ≤ dch ≤ dch dch  u (t)P P (t) u (t)P  TES TES,min ≤ TES ≤ TES TES,max  uch (t) + udch (t) 1 TES TES ≤   E (t) = (1 δ )E (t ∆t) + ηch Pch (t) Pdch (t)/ηdch ∆t  CES CES CES CES CES CES CES  min − max − −  E ECES(t) E  CES ≤ ≤ CES  uch (t)Pch Pch (t) uch (t)Pch (24)  CES CES,min CES CES CES,max  dch dch ≤ dch ≤ dch dch  u (t)P P (t) u (t)P  CES CES,min ≤ CES ≤ CES CES,max  uch (t) + udch (t) 1 CES CES ≤ where ETES(t) and ECES(t) are the energy stored in TES and CES at time t, respectively; δTES and δCES ch dch are the energy dissipation rates; ηTES and ηTES are the charging and discharging efficiency of TES, ch dch min respectively; ηCES and ηCES are the charging and discharging efficiencies of CES, respectively; ETES and max min max ETES are the lower and upper limits of TES, respectively; ECES and ECES are the lower and upper limits ch ch ch ch of CES, respectively; PTES,min, PTES,max, PCES,min, and PCES,max are the charging bounds for TES and dch dch dch dch ch ( ) CES; PTES,min, PTES,max, PCES,min, and PCES,max are the discharging bounds for TES and CES; uTES t , dch ( ) ch ( ) dch ( ) uTES t , uCES t and uCES t are all binary variables which denote the operation state for TES and CES.

2.3. RTP-Based DR Model As described in Section1, the DR program could achieve the interaction between the demand side and the MECS, which could benefit from a decreasing electricity cost and avoidance of capacity investment. The RTP-based DR program as one kind of PBDR is employed in this section. The RTP modeling is introduced by the day-ahead forecasted electricity load and TOU pricing to acquire the dynamic RTP tariff [28]. Furthermore, the electricity load model was conducted based on the RTP tariff. To acquire the day-ahead RTP tariff, the day-ahead average electricity demand needs to be calculated first, and then the floating factor of RTP could be obtained. At last, the RTP tariff could developed by TOU price rate and floating factor, as shown in the following:

X24 av Le = Le(t)/24 (25) t=1

av θ(t) = Le(t)/Le (26)

cRTP(t) = cTOU(t)θ(t) (27) In addition, the upper and lower limit of the RTP tariﬀ should be considered, i.e.:

c cRTP(t) cRTP (28) RTP,min ≤ ≤ ,max Based on the RTP tariff, the new load model could be formulated based on self-elasticity coefficient and cross-elasticity coefficient. The self-elasticity factor is always negative, while the cross-elasticity Energies 2019, 12, 3995 9 of 21 factor is always positive. By evaluating the ability of the RTP-based DR, the new load profile could be acquired according to [23,31], which is shown as follows:            X24  RTP  cRTP(t) cTOU(t) cRTP(k) cTOU(k)  Le (t) = Le(t)1 + e(t, t) − + e(t, k) −  (29)  c (t) c (k)   TOU TOU   k = 1     k , t 

av where Le is the average electricity load; θ(t) is the floating factor; cRTP(t) and cTOU(t) are the electricity price based on the RTP and TOU schemes, respectively; cRTP,min and cRTP,max are the lower and upper RTP bounds of the RTP tariff, respectively; Le (t) is the new load model based on RTP pricing; e(t,t) is the self-elasticity coefficient; and e(t,k) represents the cross-elasticity coefficient.

3. MO Operation Strategy

3.1. MO Optimization Approach Based on the optimal scheduling model formulated in Section2, objectives (Equations. (1) and Equations. (7)) and constraints (Equations (11)–(24)) were all MILP modelled. Therefore, some heuristic evolutionary algorithms were not suitable for this paper’s model, which might suﬀer from computational demands and inconsistent solutions owing to the high-dimension and dynamic variables in the MECS [12]. In this paper, the ε-constraint method, as a mathematical programming technique, was implemented to cope with the MO optimization problem, as it converts the MO model into a series of single-objective models. In this approach, one of the objective functions was regarded as an inequality constraint, and the other one was optimized, as shown below:

min f2 s.t. ( (30) f ε 1 ≤ Eqs. (11) (24) − where ε varies from f 1,max to f 1,min by setting the values of i and n according to:

ε = f ( f f )(i 1)/(imax 1), i = 1, 2 ... imax (31) 1,max − 1,max − 1,min − −

By solving the single-objective optimization problem by minimizing f 1 with Equations (11)–(24), the optimal solution X1 could be acquired, and then the values of f 1 and f 2 could be obtained, which are indicated by f 1,min and f 2,max. In addition, X2, f 2,min and f 1,max could be obtained by minimizing f 2 with Equations (11)–(24). Finally, the ε-constraint approach could be applied to the optimal scheduling model, and the Pareto front could then be acquired.

3.2. Preferred Solution Selection Approach It must be noted that the points on the Pareto front are all optimal values, and their main differences are the preferences from the operator. Therefore, in this study, the LINMAP is introduced to assist the operator to make a trade-off among the economic and exergetic objectives. According to [32], the concepts of the Utopia point and the Nodir point can be introduced to conduct the most preferred solution among Pareto set points. The Utopia point corresponds to the fact that all objectives are at their best values f k,min, and the Nodir point is a specific point where all objectives are at their worst values f k,max. Hence, the distance between each Pareto point and Utopia point is employed in the LINMAP, which means that a shorter distance represents a better solution. However, due to the distinct magnitude of the two objective functions, normalization is required before Energies 2019, 12, x FOR PEER REVIEW 10 of 21 Energies 2019, 12, 3995 10 of 21 before the implementation of this principle, and then the distance between each normalized point and (0,0) can be calculated, as shown below: the implementation of this principle, and then the distance between each normalized point and (0,0) can be calculated, as shown below: 1 ffj ≥  kk,max j  ff− j jj=≤≤ 1 kk,min f f ρkkkk k fff,mink,max ,max (32)  ffj − ≥ j  kk,maxf fk,min ,min j ρ =  k − (32) k  fk,minj f fk,max  0f k ,max f k ,min ff≤ k ≤  − j kk,min  0 f f k ≤ k,min 2 d =s ()ρj (33) jkX j 2 = k dj ρk (33) k j where fk,min and fk,max are the minimum and maximum value for k-th objective, respectively; fk j j where fkj,min and fk,max are the minimum and maximum value for k-th objective, respectively; f and ρ and ρk are the optimal value and optimal normalized value of j-th Pareto point for k-th objective,k k are the optimal value and optimal normalized value of j-th Pareto point for k-th objective, respectively; respectively; and dj represents the evaluation index of j-th Pareto point. and dj represents the evaluation index of j-th Pareto point.

3.3.3.3. MO MO Operation Operation Rules Rules BasedBased on on the the solution solution for MOfor optimalMO optimal scheduling scheduling model inmodel Sections in Subsections3.1 and 3.2, the3.1 ﬂowchart and 3.2, forthe theflowchart MO optimal for the operation MO optimal of the MECSoperation is developed of the MECS in this is part, developed as shown in inthis the part, following as shown (Figure in2 the): following (Figure 2):

MO optimization Preferred solution selection approach approach

ε-constraint approach LINMAP Start Start Solve the model

min f1 Obtain the ρ2 st.. values of f1,min 1 − f2 Eqs. (11 ) ( 24 ) and f2,max Normalization for f2,max Pareto front Nodir point each point on the Pareto front Solve the model Obtain the Pareto front values of f2,min min f2 1,max 0 and f f2,min 1 ρ1 st.. Uptopia point − Eqs. (11 ) ( 24 ) f1,min f1,max f1 Calculate the Set imax, i=1 distance between Pareto front each normalized Solve the model point and (0,0)

min f2 ρ2 st.. Calculate the 1 value of ε ≤ Select the point with Preferred solution  f1 ε minimum distance  i=i+1 Eqs. (11 )− ( 24 ) as the preferred Y solution i

End End

FigureFigure 2. 2.Flowchart Flowchart of of multi-objective multi-objective (MO) (MO) operation operation strategy. strategy.

4.4. Case Case Study Study InIn order order to to demonstrate demonstrate the the rationality rationality of the of economic–exergeticthe economic–exergetic scheduling scheduling model model with a MILPwith a descriptionMILP description and to showand to the show eﬀectiveness the effectiveness of the MO of operation the MO strategy,operation case strategy, studies case were studies simulated were forsimulated this section. for this Two section. cases (with Two cases and without (with and RTP-based without RTP-based DR) are investigated DR) are investigated to highlight to the highlight merit ofthe the merit DR program of the DR in theprogram MECS, in and the theMECS, results and are the discussed results are in detail discussed to compare in detail the to distinction compare ofthe thedistinction scheduling of the results. scheduling In addition, results. sensitivity In addition, experiments sensitivity were experiments conducted were to measure conducted the to objective measure variationthe objective caused variation by diverse caused implementation by diverse implemen strengthstation of the strengths DR scheme. of the Furthermore,DR scheme. Furthermore, due to the MILPdue to formulation the MILP offormulation the optimal of scheduling the optimal model sche ofduling the MECS, model theof state-of-the-artthe MECS, the CPLEXstate-of-the-art solver wasCPLEX employed solver towas address employed the MILP to address issue underthe MILP the GAMSissue under environment. the GAMS environment.

Energies 2019, 12, 3995 11 of 21

Energies 2019, 12, x FOR PEER REVIEW 11 of 21 4.1. System Description 4.1. System Description The energy supply chain of the MECS (described in Figure1) fulfills the secondary energy carrier needsThe (electricity energy supply load, cooling chain load,of the and MECS hot water (described load) fromin Figure the energy 1) fulfills user sidethe secondary by the coupling energy of carrierdifferent needs energy (electricity carriers load, and the cooling interaction load, and of diverse hot water energy load) conversion from the unitsenergy and user energy side storageby the couplingdevices. However,of different it mustenergy be notedcarriers that, and in the this inte paper,raction the MECSof diverse is not energy an energy conversion coupling units system and in energyreality. storage We just demonstratedevices. However, the optimal it must economic–exergetic be noted that, in schedulingthis paper, modelthe MECS and theis not MO an operation energy couplingstrategy bysystem way ofin simulation.reality. We Thejust detaileddemonstrate energy the topology optimal and economic–exergetic energy device model scheduling were discussed model andin Section the MO2; therefore,operation instrategy this subsection, by way of we simula mainlytion. focus The on detailed the energy energy device topology configuration and energy and devicedata sets model for the were optimal discussed energy in management Section 2; therefore, of the MECS. in this Some subsection, literature we have mainly conducted focus optimalon the energyoperation device of an configuration MECS for di andfferent data seasonal sets for days,the optimal such as energy in [12 ,14management]. However, of in the this MECS. paper, Some data literaturerequired forhave a typicalconducted day inoptimal summer operation were implemented of an MECS as for a testdifferent example seasonal to verify days, the such rationality as in [and12,14 eff]. ectivenessHowever, ofin thethis optimal paper, schedulingdata required model for anda typical solving day algorithm, in summer both were og whichimplemented are not onlyas a testsuitable example for summers to verify daythe rationality but also for and each effectiveness seasonal day. of the optimal scheduling model and solving algorithm,According both og to [which28], the are demands not only of suitable cooling, fo heat,r summers and electricity day but withoutalso for each the consideration seasonal day. of the DR programAccording are to given [28], inthe Figure demands3. The of electricitycooling, heat, loads and with elec thetricity DR schemewithout implementation the consideration for of each the DRinterval program which are were given calculated in Figure by 3. theThe RTP-based electricity loads DR model with withthe DR a self-elasticity scheme implementation coefficient -0.2 for andeach a intervalcross-elasticity which were coeffi cientcalculated 0.01, according by the RTP-based to [31], and DR are model presented with ina self-elasticity Figure3a. There coefficient are some -0.2 similar and aand cross-elasticity distinct features coefficient by the 0.01, comparison according of to these [31], two and electricity are presented loads in curves. Figure Two3a. There peaks are and some two similarvalleys and exist distinct for both features cases at by the the same comparison periods in of the these summer two typicalelectricity day, loads while curves. the DR Two implementation peaks and twocould valleys smooth exist the electricityfor both cases demand at the curve same to realizeperiods peak in the shaving summer and valleytypical filling. day, while Moreover, the DR the implementationcooling and heat could energy smooth are required the electricity to meet the dema needsnd ofcurve space to cooling realize and peak domestic shaving hot and water valley and filling.the energy Moreover, demands the forcooling each intervaland heat are energy illustrated are required in Figure to3 b.meet In addition,the needs to of evaluate space cooling the exergy and domesticefficiency hot of the water MECS, and the the temperature energy demands parameters for ofeach cooling interval and heatare energyillustrated were in considered Figure 3b. to In be addition,299.15 and to 333.15evaluate K, respectivelythe exergy efficiency [14]. of the MECS, the temperature parameters of cooling and heat energy were considered to be 299.15 and 333.15 K, respectively [14]. 1000 2500 Electricity without DR Electricity with DR Heat Cooling 800 2000

600 1500

400 1000

200 500 Elect r icity demand (kW h) Heat andcooling demand (kWh) 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Time (h) Time (h)

(a) (b)

FigureFigure 3. 3. EnergyEnergy demands demands of a of typical a typical summer summer day: day:(a) electricity (a) electricity demand demand and (b) and heat ( band) heat cooling and demands.cooling demands.

ToTo evaluateevaluate thethe distinction distinction between between exergy exergy efficiency efficiency and energyand energy efficiency efficiency of the whole of the system, whole a system,mathematical a mathematical model of themodel energy of the effi energyciency efficien was requiredcy was and required could and be obtained could be in obtained many references, in many references,such as [18 ,such20]. Theas [18 electricity,20]. The from electricity the utility from grid the wasutility assumed grid was to beassumed generated to be by generated power plants by powerand supply plants to and the supply MECS, andto the the MECS, exergy and effi ciencythe exergy of the efficiency power plant of the was power employed plant was to calculate employed the toexergy calculate input the of exergy the coal, input which of wasthe coal, considered which towas be considered 33.5% [16]. to Additionally, be 33.5% [16 the]. Additionally, energy efficiency the energyof the powerefficiency plant of was the assumedpower plant to be was 32.4% assumed in this paper,to be 32.4% a val uewhichin this paper, was applied a val uewhich to the energy was appliedefficiency to calculationthe energy accordingefficiency tocalculation [20]. The according dynamic electricity to [20]. The TOU dynamic tariff was electricity applied TOU to the tariff scenario was appliedwithout to the the DR scenario scheme without in this the example DR scheme day and in this was example divided day into and three was parts: divided peak, into mid-peak three parts: and peak,off-peak, mid-peak a division and which off-peak, was deriveda division from which [33]. was Furthermore, derived from the RTP [33]. was Furthermore, obtained according the RTP to was the obtainedproposed according model with to the the minimum proposed $0.05 model and maximumwith the minimum $0.25. The $0.05 electricity and pricingmaximum of the $0.25. TOU The and electricityRTP schemes pricing for eachof the period TOU areand described RTP schemes in Figure for each4. period are described in Figure 4.

Energies 2019, 12, 3995 12 of 21 Energies 2019, 12, x FOR PEER REVIEW 12 of 21

0.30 TOU RTP 0.25

0.20

Energies 2019, 12, x FOR PEER 0.15REVIEW 12 of 21 0.10

Grid price ($/kWh) 0.050.30 TOU RTP 0.000.25 123456789101112131415161718192021222324 0.20 Time (h) 0.15 Figure 4. TOU (time of use) and RTP (real-time pricing) tariﬀs. Figure0.10 4. TOU (time of use) and RTP (real-time pricing) tariffs.

Solar energy is utilizedGrid price ($/kWh) 0.05 in the MECS extensively because it is environmentally friendly and free of charge. To acquire the energy output of the PV and STC systems, which are regarded as uncontrollable 310 0.00 1000 units, ambient temperature andTemperature solar radiationSolar were radiation retrieved from a building management system 123456789101112131415161718192021222324900 308 installed in Hong Kong and are illustrated in FigureTime5[ 11 (h)] in this typical day, and) the installed area of 800 2 2 the PV and the STC were306 set to be 500 and 300 m , respectively. The ambient700 temperature varied from 2 300 to 308 K, and the maximum solar radiation was about 900 W/m at 12:00.600 Additionally, to evaluate 304Figure 4. TOU (time of use) and RTP (real-time pricing) tariffs. the exergy in solar energy, the temperature of the sun was typically regarded500 as 6000 K [34]. 302 400

Temperature (K) 300 300310 1000 Temperature Solar radiation 200 Solar radiation (w/m 900 298308

100 )

800 2 296 0 306 700 1 2 3 4 5 6 7 8 9 101112131415161718192021222324 600 304 Time (h) 500

302 400

Temperature (K) 300 300 Figure 5. Ambient temperature and solar radiation.

200 Solar radiation (w/m 298 100 Solar energy is utilized in the MECS extensively because it is environmentally friendly and free 296 0 of charge. To acquire the 1energy 2 3 4 5 output 6 7 8 9 101112131415161718192021222324of the PV and STC systems, which are regarded as uncontrollable units, ambient temperature andTime solar (h) radiation were retrieved from a building management system installed in Hong Kong and are illustrated in Figure 5 [11] in this typical day, Figure 5. Ambient temperature and solar radiation. and the installed area of the PV and the STC were set to be 500 and 300 m2, respectively. The ambient Figure 5. Ambient temperature and solar radiation. temperatureNatural varied gas can from be utilized300 to 308 in K, the and CHP the unit maximum to generate solar electricity, radiation whilewas about waste 900 heat W/m can2 at be 12:00. Additionally, to evaluate the exergy in solar energy, the temperature of the sun was typically recoveredSolar toenergy meet is the utilized cooling in and the heatMECS demands extensivel of they because AC and it the is environmentally HEU, respectively. friendly It was assumedand free regarded as 6000 K [34]. ofthat charge. the states To ofacquire natural the gas, energy such as output temperature, of the pressure, PV and andSTC heat systems, value werewhich not are changed regarded during as Natural gas can be utilized in the CHP unit to generate electricity, while waste heat can be uncontrollablethe operation of units, the MECS. ambient The exergytemperature factor ofand the solar natural radiation gas was were set to retrieved be 1.04, and from its LHVa building was 50 recovered to meet the cooling and heat demands of the AC and the HEU, respectively. It was managementMJ/kg, according system to [14 installed,35]. However, in Hong owing Kong to and the are strong illustrated coupling in between Figure 5 electricity [11] in this and typical heat in day, the assumed that the states of natural gas, such as temperature, pressure, and heat value were not andCHP, the the installed operation area of of which the PV lacks and flexibility, the STC were the GB set was to be considered 500 and 300 to bem2 the, respectively. adjustable heatThe ambient unit and changed during the operation of the MECS. The exergy factor of the natural gas was set to be 1.04, temperatureTES was integrated varied tofrom realize 300 theto 308 heat K, charging and the and maximum discharging solar during radiation the operation.was about 900 W/m2 at and its LHV was 50 MJ/kg, according to [14,35]. However, owing to the strong coupling between 12:00.To Additionally, fulfill the cooling to evaluate demand, the the exergy AC and in sola ther EC energy, are employed the temperature in the MECS of the with sun complementary was typically electricity and heat in the CHP, the operation of which lacks flexibility, the GB was considered to be regardedadvantages. as 6000 The ACK [34 could]. utilize low-quality heat to generate cooling energy, while its COP was low the adjustable heat unit and TES was integrated to realize the heat charging and discharging during comparedNatural to EC.gas Ancan ECbe withutilized high in COP the consumesCHP unit high-qualityto generate electricity.electricity, Inwhile addition, waste CES heat saves can thebe the operation. recoveredreceived cooling to meet energy the cooling and then and releases heat thedemands stored of energy the AC to jointly and the meet HEU, the coolingrespectively. demand It withwas To fulfill the cooling demand, the AC and the EC are employed in the MECS with assumedthe AC and that the the EC. states of natural gas, such as temperature, pressure, and heat value were not complementary advantages. The AC could utilize low-quality heat to generate cooling energy, while changedThe during efficiency, the COP, operation and energy of the lossMECS. rate The of relative exergy devicesfactor of are the shown natural in Tablegas was1. The set restrictionto be 1.04, its COP was low compared to EC. An EC with high COP consumes high-quality electricity. In andconditions its LHV are was summarized 50 MJ/kg, inaccording Table2. The to [ simulation14,35]. However, interval owing∆t was to defined the strong as 1 coupling hour, and between the loop addition, CES saves the received cooling energy and then releases the stored energy to jointly meet electricitycount imax andwas heat setto in be the 20 CHP, in the theε-constraint operation method.of which lacks flexibility, the GB was considered to be the cooling demand with the AC and the EC. the adjustable heat unit and TES was integrated to realize the heat charging and discharging during the operation. Table 1. Efficiency, COP (coefficient of performance), and energy loss rate of relative devices. To fulfill the cooling demand, the AC and the EC are employed in the MECS with complementary advantages. The AC could utilize low-quality heat to generate cooling energy, while its COP was low compared to EC. An EC with high COP consumes high-quality electricity. In addition, CES saves the received cooling energy and then releases the stored energy to jointly meet the cooling demand with the AC and the EC.

Table 1. Efficiency, COP (coefficient of performance), and energy loss rate of relative devices.

Energies 2019, 12, x FOR PEER REVIEW 13 of 21

Parameters Value Parameters Value Parameters Value dch ηPV 0.157 ηHEU 0.9 ηTES 0.98 Energies 2019, 12, 3995 13 of 21 ηSTC 0.8 COPEC 4.2 δCES 0.02 e ch ηCHP 0.3 COPAC 0.8 ηCES 0.97 Table 1. Efficiency,h COP (coefficient of performance), and energy lossdch rate of relative devices. ηCHP 0.45 δTES 0.02 ηCES 0.95 Parameters Value Parametersch Value Parameters Value ηGB 0.88 ηTES 0.98 dch ηPV 0.157 ηHEU 0.9 ηTES 0.98 ηSTC Table0.8 2. RestrictionCOP conditionsEC of4.2 relative devices.δCES 0.02 e COP ch ηCHP 0.3 AC 0.8 ηCES 0.97 Parametersh Value/kW Parameters Value/kW Parametersdch Value/kW ηCHP 0.45 δTES 0.02 ηCES 0.95 e c ch dch PCHPη,minGB 10000.88 PACη,minTES 1000.98 PCES,min 0 e c dch PCHP,max 200 PAC,max 1000 PCES,max 300 Table 2. Restriction conditions of relative devices. e ch RPCHP,min 500 PTES,min 0 ETES ()0 250 Parameterse Value/kW Parametersch Value/kW Parameters Value/kW RPCHP,max -500 PTES,max 200 ECES ()0 400 Pe 1000 Pc 100 Pdch 0 CHPPh ,min PdchAC,min minCES,min Pe GB,min 30 PTESc ,min 0 ETESdch 50 CHP,max 200 AC,max 1000 PCES,max 300 h dch RPPe PPch max ( ) CHPGB,max,min 300500 TESTES,max,min 2000 EETESTES 0 500 250 e ch RP c ch min ( ) CHP,max 500 PTES,max 200 ECES 0 400 PEC,min −150 PCES,min 0 ECES 80 Ph 30 Pdch 0 Emin 50 GBc,min TESch ,min maxTES hPEC,max 1500 PCESdch,max 300 ECESmax 800 PGB,max 300 PTES,max 200 ETES 500 Pc Pch min The efficiency,EC,min COP, and150 energy lossCES ,minrate of relative0 devices EareCES shown in 80Table 1. The Pc ch max restriction conditionsEC,max are summarized1500 inP TableCES,max 2. The simulation300 intervalECES Δt was defined800 as 1 hour, and the loop count imax was set to be 20 in the ε-constraint method. 4.2. Results Analyses and Discussion 4.2. Results Analyses and Discussion In Case 1, the economic–exergetic scheduling model was established with the TOU tariff, and DR wasIn Case not considered.1, the economic–exergetic In Case 2, the MOscheduling scheduling model model was wasestablished integrated with with the theTOU DR tariff, program and DRand was the RTPnot considered. tariff, and both In Case models 2, the were MO solved schedu byling the εmodel-constraint was methodintegrated to obtainwith the the DR Pareto program front, andas shown the RTP in Figuretariff, 6and. To both determine models a were trade-o solvedff between by the the ε-constraint economic method and exergetic to obtain objectives, the Pareto the front,LINMAP as shown approach in Figure was then6. To applied,determine and a trade-off the results between are illustrated the economic in Table and3 .exergetic The details objectives, of the theε-constraint LINMAP method approach and was LINMAP then applied, approach and werethe results illsustrated are illustrated in Figure in2. Table For Case 3. The 1, Solutiondetails of #17the εwas-constraint the compromise method scheme,and LINMAP while Solutionapproach #16 were was illsustrated the preferred in Figure point for 2. CaseFor Case 2, both 1, Solution of which #17 are wasmarked the incompromise red on each scheme, Pareto carvewhile in Solution Figure6 .#16 Detailed was the analyses preferred of these point two for selected Case 2, solutionsboth of which were areconducted, marked thein red results on each of which Pareto are carve shown in Figure below. 6. Detailed analyses of these two selected solutions were conducted, the results of which are shown below. 77,000 76,500 Without DR With DR 76,000 75,500 75,000 74,500 74,000 73,500 Daily exergy input (kWh) input exergy Daily 73,000 72,500 3400 3450 3500 3550 3600 3650 3700 3750 Daily energy cost ($)

Figure 6. Pareto front with and without demand response (DR). Figure 6. Pareto front with and without demand response (DR).

Table 3. Pareto optimal solutions by the ε-constraint method.

Energies 2019, 12, 3995 14 of 21

Table 3. Pareto optimal solutions by the ε-constraint method.

Case 1: Without DR Case 2: With DR

i f 1($) f 2(kWh) ρ1 ρ2 d f 1($) f 2(kWh) ρ1 ρ2 d 1 3704.41 73574.05 1.000 0.000 1.000 3676.15 72847.62 1.000 0.000 1.000 2 3692.63 73618.71 0.947 0.014 0.947 3663.75 72882.55 0.947 0.014 0.947 3 3680.84 73667.91 0.895 0.029 0.895 3651.35 72923.42 0.895 0.029 0.895 4 3667.59 73727.98 0.836 0.048 0.837 3638.95 72959.54 0.842 0.044 0.843 5 3657.27 73766.30 0.789 0.060 0.792 3626.55 73000.49 0.789 0.059 0.792 6 3645.49 73816.30 0.737 0.076 0.741 3614.15 73042.87 0.737 0.076 0.741 7 3633.70 73864.69 0.684 0.091 0.690 3601.75 73086.82 0.684 0.093 0.690 8 3621.91 73913.88 0.632 0.107 0.641 3589.35 73132.19 0.632 0.111 0.641 9 3610.13 73963.08 0.579 0.122 0.592 3576.95 73180.72 0.579 0.129 0.593 10 3598.34 74012.27 0.526 0.137 0.544 3564.55 73229.62 0.526 0.148 0.547 11 3586.56 74061.46 0.474 0.153 0.498 3552.15 73281.93 0.474 0.169 0.503 12 3574.77 74110.66 0.421 0.168 0.453 3539.75 73337.80 0.421 0.191 0.462 13 3562.98 74159.85 0.368 0.184 0.412 3527.35 73401.26 0.368 0.215 0.427 14 3551.20 74209.47 0.316 0.199 0.373 3514.95 73464.81 0.316 0.240 0.397 15 3539.41 74263.74 0.263 0.216 0.341 3502.55 73546.73 0.263 0.272 0.378 16 3527.63 74318.07 0.211 0.233 0.314 3490.15 73644.09 0.211 0.310 0.374 17 3515.84 74389.45 0.158 0.256 0.301 3477.74 73793.13 0.158 0.367 0.400 18 3504.05 74731.77 0.105 0.363 0.378 3465.34 74162.12 0.105 0.511 0.522 19 3492.27 75732.11 0.053 0.677 0.679 3452.94 74642.98 0.053 0.698 0.700 20 3480.48 76761.51 0.000 1.000 1.000 3440.54 75420.63 0.000 1.000 1.000

The energy operation costs for Cases 1 and 2 were, respectively, $3515.84 and 3490.15 with corresponding exergy inputs of 74,389.45 and 73,644.09 kWh. The exergy efficiency for each case was 21.91% and 21.86%, and the energy efficiency for each case was 92.98% and 93.71%, respectively. It was found that the energy cost and exergy input both decreased with DR, which means that the more preferred operation strategy was arranged to simultaneously boost these two different goals after the implementation of the RTP-based DR, and the exergy or energy efficiencies had few variations between these two cases. However, it was noted that the energy efficiency for each case was more than 90% which was because the EC unit possessed a high COP. Therefore, the energy efficiency index only denotes the quantity performance, which neglects the quality evaluation of different energy carriers. If the energy efficiency index is adopted, the energy-saving performance might not be reasonably evaluated. Hence, exergy efficiency might be more rational for the assessment of an MECS compared to the energy efficiency index. The purchased electricity from the grid is described in Figure7a, and the optimal scheduling results for electricity are presented in Figure8 for both cases at each time interval. In Case 1, the electricity was not purchased during the periods of 8:00–20:00 when the TOU tariffs were at peak or mid-peak, as shown in Figure4. However, the electricity was purchased not only during the o ff-peak periods 1:00–7:00 and 23:00–24:00 but also during the time interval of 21:00–22:00, even if the price rate was at mid-peak, which was caused by the electricity needs of the EC and demand that could not solely be fulfilled by the CHP unit. Therefore, more electricity was required through the grid when the PV output was zero during the night. Compared to Case 1, the electricity was not bought from the grid when the RTP tariff was high in Case 2, especially during the period of 20:00–22:00. This contributes to the fact that the electricity needs and demand of the EC could be met by the CHP unit alone, as shown in Figure8b. The electricity demand was reduced under the implementation of DR, which is illustrated in Figure3. Hence, electricity consumption was more rationally and economically arranged. By comparing Figure8a,b, it can be seen that there were many di fferences at 3:00–4:00 and 23:00 other than the 21:00–22:00 time periods. The electricity demands varied slightly, while the RTP rates were obviously less than TOU tariffs during 3:00–4:00, as illustrated in Figure4. Therefore, the power energy purchased from the utility grid in Case 2 was more than in Case 1 during 3:00–4:00. Different results Energies 2019, 12, x FOR PEER REVIEW 14 of 21

Case 1: Without DR Case 2: With DR i f1($) f2(kWh) ρ1 ρ2 d f1($) f2(kWh) ρ1 ρ2 d 1 3704.41 73574.05 1.000 0.000 1.000 3676.15 72847.62 1.000 0.000 1.000 2 3692.63 73618.71 0.947 0.014 0.947 3663.75 72882.55 0.947 0.014 0.947 3 3680.84 73667.91 0.895 0.029 0.895 3651.35 72923.42 0.895 0.029 0.895 4 3667.59 73727.98 0.836 0.048 0.837 3638.95 72959.54 0.842 0.044 0.843 5 3657.27 73766.30 0.789 0.060 0.792 3626.55 73000.49 0.789 0.059 0.792 6 3645.49 73816.30 0.737 0.076 0.741 3614.15 73042.87 0.737 0.076 0.741 7 3633.70 73864.69 0.684 0.091 0.690 3601.75 73086.82 0.684 0.093 0.690 8 3621.91 73913.88 0.632 0.107 0.641 3589.35 73132.19 0.632 0.111 0.641 9 3610.13 73963.08 0.579 0.122 0.592 3576.95 73180.72 0.579 0.129 0.593 10 3598.34 74012.27 0.526 0.137 0.544 3564.55 73229.62 0.526 0.148 0.547 11 3586.56 74061.46 0.474 0.153 0.498 3552.15 73281.93 0.474 0.169 0.503 12 3574.77 74110.66 0.421 0.168 0.453 3539.75 73337.80 0.421 0.191 0.462 13 3562.98 74159.85 0.368 0.184 0.412 3527.35 73401.26 0.368 0.215 0.427 14 3551.20 74209.47 0.316 0.199 0.373 3514.95 73464.81 0.316 0.240 0.397 15 3539.41 74263.74 0.263 0.216 0.341 3502.55 73546.73 0.263 0.272 0.378 16 3527.63 74318.07 0.211 0.233 0.314 3490.15 73644.09 0.211 0.310 0.374 17 3515.84 74389.45 0.158 0.256 0.301 3477.74 73793.13 0.158 0.367 0.400 18 3504.05 74731.77 0.105 0.363 0.378 3465.34 74162.12 0.105 0.511 0.522 19 3492.27 75732.11 0.053 0.677 0.679 3452.94 74642.98 0.053 0.698 0.700 20 3480.48 76761.51 0.000 1.000 1.000 3440.54 75420.63 0.000 1.000 1.000 The energy operation costs for Cases 1 and 2 were, respectively, $3,515.84 and 3,490.15 with corresponding exergy inputs of 74,389.45 and 73,644.09 kWh. The exergy efficiency for each case was 21.91% and 21.86%, and the energy efficiency for each case was 92.98% and 93.71%, respectively. It was found that the energy cost and exergy input both decreased with DR, which means that the more preferred operation strategy was arranged to simultaneously boost these two different goals Energies 2019, 12, 3995 15 of 21 after the implementation of the RTP-based DR, and the exergy or energy efficiencies had few variations between these two cases. However, it was noted that the energy efficiency for each case waswere more obtained than during90% which 21:00–23:00, was because when the the EC electricity unit possessed demands a werehigh atCOP. peak Therefore, and curtailed the energy by the efficiencyDR scheme index with only higher denotes RTP tari theﬀs quantity in Case 2. perfor Hence,mance, less power which energy neglects was the purchased quality evaluation from the grid, of differentwhich was energy because carriers. of the If fact the that energy the newefficiency price mechanismindex is adopted, more rationally the energy-saving modiﬁed theperformance electricity mightconsumption, not be andreasonably the operation evaluated. strategy Hence, was optimized exergy efficiency in this paper. might be more rational for the assessment of an MECS compared to the energy efficiency index.

700 300 Without DR With DR Without DR With DR 600 250

500 200 400 150

300 (kg)

(kWh) 100 200 100 50 0 Purchased gas from pipeline 0

Purchased electricity fromgrid 123456789101112131415161718192021222324 1 2 3 4 5 6 7 8 9 101112131415161718192021222324 Time (h) Time (h)

(a) (b)

EnergiesFigureFigure 2019 7., 12 The 7., xThe FORpurchased purchased PEER REVIEW electricity electricity and and natural natural gas: gas: (a) (purchaseda) purchased electricity; electricity; (b) (purchasedb) purchased gas. gas. 15 of 21

1500 1500 Grid PV CHP EC Demand without DR Grid PV CHP EC Demand with DR 1000 1000

500 500

0 0 (kWh) (kWh) -500 -500

-1000 -1000 Electricity energybalance with DR -1500 -1500 Electricity energy balance without DR 123456789101112131415161718192021222324 123456789101112131415161718192021222324 Time (h) Time (h)

(a) (b)

FigureFigure 8. 8.Optimal Optimal scheduling scheduling results results for for electricity: electricity: ( a(a)) without without DR; DR; ( b(b)) with with DR. DR.

TheThe purchased purchased natural electricity gas fromfrom the the pipeline grid is isdescribed presented in inFigure Figure 77a,b, and and the the optimal distribution scheduling ratios forresults the CHP for electricity unit and the are GB presented are illustrated in Figure in Figure 8 for9 both for each cases case at each at each time hour. interval. The purchased In Case 1, gas the forelectricity both cases was was not distinctpurchased in someduring periods, the periods owing of to8:00–20:00 the DR program. when the TheTOU amount tariffs were of gas at needed peak or inmid-peak, Case 1 was as moreshown than in inFigure Case 24 during. However, 3:00–4:00, the electricity and more was natural purchased gas was purchasednot only during in Case the 2 atoff-peak 23:00, as periods described 1:00–7:00 in Figure and7 b.23:00–24:00 Hence, the but scheduling also during results the fortime natural interval gas of and 21:00–22:00, grid electricity even if werethe price complementary rate was at atmid-peak, these periods which because was caused more by natural the electricity gas are purchased needs of andthe EC less and electricity demand were that neededcould not when solely the RTPbe fulfilled tariffs were by the high CHP at someunit. periods,Therefore, and more vice electricity versa. From was Figure required9, it can through be seen the thegrid GB when served the as PV an auxiliaryoutput was heat zero source during to adjust the nigh thet. imbalance Compared between to Case heat 1, /thecooling electricity demands was and not electricitybought from demands the grid due when to the strongthe RTP coupling tariff was of the high CHP. in ForCase example, 2, especially when during the heat the and period cooling of demands20:00–22:00. were This at contributes peak and the to electricitythe fact that demands the electricity were needs low during and demand 14:00–15:00, of the theEC heatcould energy be met producedby the CHP by the unit CHP alone, could as notshown meet in the Figure thermal 8b. requirements The electricity of thedemand AC and was the reduced HE; therefore, under thethe GBimplementation started to make of up DR, for which the heat is deficiencyillustrated for in bothFigure cases, 3. Hence, as shown electricity in Figure consumption9. However, it was must more be notedrationally that theand GB economically operated not arranged. only during By comparing 14:00–15:00 Figure but also 8a,b, at it 13:00 can be in seen Case that 1 due there to thewere lack many of thedifferences filling valley at 3:00–4:00 effect of and the DR23:00 scheme. other than In Case the 2,21 the:00–22:00 electricity time demands periods. wereThe electricity increased demands at 13:00 comparedvaried slightly, to Case while 1; hence, the theRTP electricity rates were generated obviou bysly the less CHP than is higherTOU thantariffs that during in Case 3:00–4:00, 1, and the as heatillustrated energy in produced Figure 4 by. Therefore, the CHP couldthe power fulfill energy the thermal purchased needs from of the the AC utility and the grid HE. in To Case this 2 end, was themore GB than unit didin Case not have1 during to start 3:00–4:00. up at 13:00 Different in Case results 2. were obtained during 21:00–23:00, when the electricity demands were at peak and curtailed by the DR scheme with higher RTP tariffs in Case 2. Hence, less power energy was purchased from the grid, which was because of the fact that the new price mechanism more rationally modified the electricity consumption, and the operation strategy was optimized in this paper.

CHP GB CHP GB 100% 100%

80% 80%

60% 60%

40% 40%

20% 20% Gas distributionGas withDR 0% 0% Gas distribution without DR without distribution Gas 123456789101112131415161718192021222324 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Time (h) Time (h)

(a) (b)

Figure 9. Natural gas distribution ratio: (a) without DR; (b) with DR.

The purchased natural gas from the pipeline is presented in Figure 7b, and the distribution ratios for the CHP unit and the GB are illustrated in Figure 9 for each case at each hour. The purchased gas for both cases was distinct in some periods, owing to the DR program. The amount of gas needed in Case 1 was more than in Case 2 during 3:00–4:00, and more natural gas was purchased in Case 2 at 23:00, as described in Figure 7b. Hence, the scheduling results for natural gas and grid electricity were complementary at these periods because more natural gas are purchased and less electricity were needed when the RTP tariffs were high at some periods, and vice versa. From Figure

Energies 2019, 12, x FOR PEER REVIEW 15 of 21

1500 1500 Grid PV CHP EC Demand without DR Grid PV CHP EC Demand with DR 1000 1000

500 500

0 0 (kWh) (kWh) -500 -500

-1000 -1000 Electricity energybalance with DR -1500 -1500 Electricity energy balance without DR 123456789101112131415161718192021222324 123456789101112131415161718192021222324 Time (h) Time (h)

(a) (b)

Figure 8. Optimal scheduling results for electricity: (a) without DR; (b) with DR.

The purchased electricity from the grid is described in Figure 7a, and the optimal scheduling results for electricity are presented in Figure 8 for both cases at each time interval. In Case 1, the electricity was not purchased during the periods of 8:00–20:00 when the TOU tariffs were at peak or mid-peak, as shown in Figure 4. However, the electricity was purchased not only during the off-peak periods 1:00–7:00 and 23:00–24:00 but also during the time interval of 21:00–22:00, even if the price rate was at mid-peak, which was caused by the electricity needs of the EC and demand that could not solely be fulfilled by the CHP unit. Therefore, more electricity was required through the grid when the PV output was zero during the night. Compared to Case 1, the electricity was not bought from the grid when the RTP tariff was high in Case 2, especially during the period of 20:00–22:00. This contributes to the fact that the electricity needs and demand of the EC could be met by the CHP unit alone, as shown in Figure 8b. The electricity demand was reduced under the implementation of DR, which is illustrated in Figure 3. Hence, electricity consumption was more rationally and economically arranged. By comparing Figure 8a,b, it can be seen that there were many differences at 3:00–4:00 and 23:00 other than the 21:00–22:00 time periods. The electricity demands varied slightly, while the RTP rates were obviously less than TOU tariffs during 3:00–4:00, as illustrated in Figure 4. Therefore, the power energy purchased from the utility grid in Case 2 was more than in Case 1 during 3:00–4:00. Different results were obtained during 21:00–23:00, when the electricity demands were at peak and curtailed by the DR scheme with higher RTP tariffs in Case 2. Hence, less power energy was purchased from the grid, which was because of the fact that the new priceEnergies mechanism2019, 12, 3995 more rationally modified the electricity consumption, and the operation strategy16 of 21 was optimized in this paper.

CHP GB CHP GB 100% 100%

80% 80%

60% 60%

40% 40%

(a) (b)

Figure 9. NaturalNatural gas gas distribution distribution ratio: ratio: ( (aa)) without without DR; DR; ( (bb)) with with DR. DR.

TheThe optimalpurchased scheduling natural resultsgas from of heatthe energypipeline for is demands presented and in devices Figure are 7b, presented and the indistribution Figures 10 ratiosand 11 for, and the the CHP cooling unit energy and the balance GB are graphs illustrated are described in Figure in Figure 9 for 12 each. Though case theat each DR scheme hour. The was purchasedapplied to thegas electricityfor both cases demands, was distinct the heat in andsome cooling periods, operation owing to strategies the DR program. were affected The amount due to the of gasenergy needed coupling. in Case In 1 Case was 2,more TES than released in Case the thermal2 during energy 3:00–4:00, during and 3:00–4:00, more natural while gas it waswas stillpurchased in Case in1, asCase illustrated 2 at 23:00, in Figureas described 10. This in was Figure owed 7b. to Hence, the fact the that scheduling the CHP unit results generatds for natural less electricitygas and grid and electricitythermal energy, were whilecomplementary the utility gridat these supplied periods more because electricity more in Casenatural 2, asgas shown are purchased in Figures 8and and less 11. EnergieselectricityHence, 2019 the , were12 heat, x FOR needed energy PEER vacancywhenREVIEW the was RTP compensated tariffs were byhigh TES. at some In addition, periods, the and heat vice energy versa. demands From16 Figure of were 21 fulfilled not only by the HE but also by TES discharging in Case 1 at 23:00, while more heat energy was 9 generated, it can be by seen theCHP the unitGB inserved Case 2,as soan TES auxiliar neededy heat charge source from theto surplusadjust the thermal imbalance energy between as shown heat/coolingin Figures 10 demands and 11. Thisand distinctionelectricity demands of TES operation due to the mode strong is obviously coupling dependentof the CHP. on For the example, RTP-DR whenprogram. the heat Moreover, and cooling the cooling demands energy were operation at peak strategies and the have electricity been modified demands in casewere 2 duringlow during some 14:00–15:00,periods as well. the heat Due energy to the low produced RTP electricity by the CH tariPff scould at 3:00 not and meet 4:00 the in casethermal 2, more requirements electricity of energy the ACare and purchased the HE; from therefore, the grid the and GB are started converted to make to cooling up for energythe heat by deficiency the EC unit for and both then cases, are as charged shown in inthe Figure CES device9. However, so that it the must surplus be noted cooling that energy the GB could operated be utilized not only at the during peak 14:00–15:00 of cooling demands. but also at In 13:00addition, in Case in Figure 1 due 12to itthe can lack be summarizedof the filling thatvalley the effect EC module of the DR is the scheme. main cooling In Case device 2, the toelectricity fulfill the demandscooling demand were increased compared at with13:00 the compared AC unit, to which Case contributes1; hence, the to electricity the fact that generated the EC device by the possesses CHP is higherhigher than COP that by consumingin Case 1, and high-quality the heat energy electricity produced energy by which the CHP means could lower fulfill energy the thermal cost. needs of the AC and the HE. To this end, the GB unit did not have to start up at 13:00 in Case 2. 1500 1500 Demand without DR HE Output TES charge TES discharge Demand with DR HE output TES charge TES discharge 1000 1000

500 500

0 0

-500 with DR (kWh) -500 without DR (kWh)without

-1000 -1000 Heat Heat energy balance for demand Heat energy balancefor energy demandHeat

-1500 -1500 1 2 3 4 5 6 7 8 9 101112131415161718192021222324 1 2 3 4 5 6 7 8 9 101112131415161718192021222324 Time (h) Time (h)

(a) (b)

FigureFigure 10. 10. OptimalOptimal scheduling scheduling results results for for heat: heat: (a (a) )without without DR; DR; (b (b) )with with DR. DR.

1800 1800 CHP STC GB AC HE 1600 CHP STC GB AC HE 1600 1400 1400 1200 1200 1000 1000 800 800 600 600 with DR (kWh) without DR (kWh) without 400 400 200 200 Heat energy balance forbalance devices energy Heat Heat Heat energy balance for devices 0 0 123456789101112131415161718192021222324 123456789101112131415161718192021222324 Time (h) Time (h)

(a) (b)

Figure 11. Heat balance for each device: (a) without DR; (b) with DR.

2500 2500 Demand without DR EC AC CES charge CES discharge Demand with DR EC AC CES charge CES discharge 2000 2000 1500 1500 1000 1000 500 500 0 0 (kWh) (kWh) -500 -500 -1000 -1000 -1500 -1500

-2000 Cooling energy balance with DR -2000 Cooling energy balance withour DR -2500 -2500 1 2 3 4 5 6 7 8 9 101112131415161718192021222324 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Time (h) Time (h)

(a) (b)

Figure 12. Optimal scheduling results for cooling: (a) without DR; (b) with DR.

The optimal scheduling results of heat energy for demands and devices are presented in Figures 10 and 11, and the cooling energy balance graphs are described in Figure 12. Though the DR scheme was applied to the electricity demands, the heat and cooling operation strategies were affected due to the energy coupling. In Case 2, TES released the thermal energy during 3:00–4:00, while it was still in Case 1, as illustrated in Figure 10. This was owed to the fact that the CHP unit generatds less electricity and thermal energy, while the utility grid supplied more electricity in Case

Energies 2019,, 12,, xx FORFOR PEERPEER REVIEWREVIEW 16 of 21

9,, itit cancan bebe seenseen thethe GBGB servedserved asas anan auxiliarauxiliary heat source to adjust the imbalance between heat/cooling demands and electricity demands due to the strong coupling of the CHP. For example, when the heat and cooling demands were at peak and the electricity demands were low during 14:00–15:00, the heat energy produced by the CHP could not meet the thermal requirements of the AC and the HE; therefore, the GB started to make up for the heat deficiency for both cases, as shown inin FigureFigure 9.. However,However, itit mustmust bebe notednoted thatthat thethe GBGB operatedoperated notnot onlyonly duringduring 14:00–15:0014:00–15:00 butbut alsoalso atat 13:00 in Case 1 due to the lack of the filling valley effect of the DR scheme. In Case 2, the electricity demands were increased at 13:00 compared to Case 1; hence, the electricity generated by the CHP is higher than that in Case 1, and the heat energy produced by the CHP could fulfill the thermal needs of the AC and the HE. To this end, the GB unit did not have to start up at 13:00 in Case 2. 1500 1500 1500 1500 Demand without DR HE Output TES charge TES discharge Demand with DR HE output TES charge TES discharge Demand without DR HE Output TES charge TES discharge Demand with DR HE output TES charge TES discharge 1000 1000 1000 1000

500 500 500 500

0 0 0 0

-500 with DR (kWh) -500 -500 with DR (kWh) -500 without DR (kWh)without without DR (kWh)without -1000 -1000 -1000 -1000 Heat energy balance for demand Heat energy balancefor energy demandHeat Heat Heat energy balance for demand Heat energy balancefor energy demandHeat -1500 -1500 -1500 -1500 1 2 3 4 5 6 7 8 9 101112131415161718192021222324 1 2 3 4 5 6 7 8 9 101112131415161718192021222324 1 2 3 4 5 6 7 8 9 101112131415161718192021222324 1 2 3 4 5 6 7 8 9 101112131415161718192021222324 Time (h) Time (h) Time (h) Time (h)

((a)) ((b)) Energies 2019, 12, 3995 17 of 21 Figure 10. Optimal scheduling results for heat: (a)) withoutwithout DR;DR; ((b)) withwith DR.DR.

1800 1800 1800 1800 CHP STC GB AC HE 1600 CHP STC GB AC HE 1600 CHP STC GB AC HE 1600 CHP STC GB AC HE 1600 1400 1400 1400 1400 1200 1200 1200 1200 1000 1000 1000 1000 800 800 800 800 600 600 600 with DR (kWh) 600 with DR (kWh) without DR (kWh) without

without DR (kWh) without 400 400 400 400 200 200 Heat energy balance forbalance devices energy Heat Heat Heat energy balance for devices 200 200 Heat energy balance forbalance devices energy Heat Heat Heat energy balance for devices 0 0 0 0 123456789101112131415161718192021222324 123456789101112131415161718192021222324 123456789101112131415161718192021222324 123456789101112131415161718192021222324 Time (h) Time (h) Time (h) Time (h)

((a)) ((b)) Figure 11. Heat balance for each device: (a) without DR; (b) with DR. FigureFigure 11. 11.Heat Heat balance balance for for each each device: device: (a ()a without) without DR; DR; (b ()b with) with DR. DR.

2500 2500 2500 Demand without DR EC AC CES charge CES discharge 2500 Demand with DR EC AC CES charge CES discharge 2000 Demand without DR EC AC CES charge CES discharge 2000 Demand with DR EC AC CES charge CES discharge 2000 2000 1500 1500 1500 1500 1000 1000 1000 1000 500 500 500 500 0 0 0 0 (kWh) (kWh) (kWh) (kWh) -500 -500 -500 -500 -1000 -1000 -1000 -1000 -1500 -1500 -1500 -1500 Cooling energy balance with DR

-2000 Cooling energy balance with DR -2000

Cooling energy balance withour DR -2000 -2000 Cooling energy balance withour DR -2500 -2500 -2500 -2500 1 2 3 4 5 6 7 8 9 101112131415161718192021222324 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Energies 2019,1 12 2, 3x 4FOR 5 6 PEER 7 8 9 101112131415161718192021222324REVIEW 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 2117 22 23of 24 21 Time (h) Time (h) Time (h) Time (h)

2, as shown in Figures((a ))8 and 11. Hence, the heat energy vacancy was ((compensatedb)) by TES. In addition, the heat energy demands were fulfilled not only by the HE but also by TES discharging in Figure 12. Optimal scheduling results for cooling: (a) without DR; (b) with DR. Case 1 at 23:00,Figure whileFigure 12. more12.Optimal Optimal heat scheduling energyscheduling was results results gene for ratedfor cooling: cooling: by the (a ()a without )CHP without unit DR; DR; in (b ( )bCase with) with DR.2, DR. so TES needed charge from the surplus thermal energy as shown in Figures 10 and 11. This distinction of TES TheThe state optimal variations scheduling of TES results and CES of forheat the ener wholegy for day demands are described and indevices Figure are 13 .presented For TES in in operationThe modeoptimal is obviouslyscheduling dependent results of on heat the ener RTP-DRgy for program. demands Moreover, and devices the arecooling presented energy in FigureFigures 13 a,10 itand releases 11, and the the heat cooling energy energy at the balance peak during graphs 12:00–14:00 are described and in 21:00–22:00 Figure 12. Though for both the cases DR operationFigures 10strategies and 11, haveand the been cooling modified energy in casebalance 2 during graphs some are describedperiods as in well. Figure Due 12 to. Though the low the RTP DR toscheme compensate was applied for the heat to the deficits. electricity In addition, demands, TES the discharged heat and the cooling energy operation during 3:00–4:00 strategies in Case were electricityscheme wastariffs applied at 3:00 toand the 4:00 electricity in case 2, demands, more electricity the heat energy and arecooling purchased operation from strategies the grid andwere 2,affected which contributeddue to the energy to the factcoupling. that more In Case electricity 2, TES wasreleased purchased the thermal from theenergy grid during because 3:00–4:00, of the areaffected converted due toto coolingthe energy energy coupling. by the InEC Case unit 2,an TESd then released are charged the thermal in the energyCES device during so that3:00–4:00, the lowwhile RTP it tariwasff s,still and in the Case CHP 1, as unit illustrated generated in lessFigure electricity 10. This and was heat owed energy. to the Similarly, fact that thethe CESCHP unit unit surpluswhile coolingit was still energy in Case could 1, asbe illustratedutilized at thein Figure peak of 10 cooling. This was demands. owed to In the addition, fact that in theFigure CHP 12 unit it releasedgeneratds cooling less electricity energy during and thermal 12:00–14:00 energy, and while 20:00–21:00 the utility when grid the supplied cooling more demands electricity were in at Case the canpeak, be as summarized illustrated in that Figure the 13ECb. module With the is coordination the main cooling of TES device and CES to systems,fulfill the the cooling MECS demand could be comparedeconomically with and the e ffiACciently unit, operatedwhich contributes to meet diverse to the energyfact that needs. the EC device possesses higher COP by consuming high-quality electricity energy which means lower energy cost. 600 1000 Without DR With DR Without DR With DR 500 800 400 600 300 400 200

State of TES (kWh) TES of State 200 100 (kWh) CES of State

0 0 123456789101112131415161718192021222324 1 2 3 4 5 6 7 8 9 101112131415161718192021222324 Time (h) Time (h)

(a) (b)

FigureFigure 13. 13. StateState variation variation of of energy energy storage storage devices: devices: ( (aa)) state state of of TES; TES; ( (bb)) state state of of CES;. CES.

4.3. SensitivityThe state variations Experiments of TES and CES for the whole day are described in Figure 13. For TES in FigureIn 13 thisa, it section, releases the the e ffheatect ofenergy the DR at the scheme peak isduring investigated 12:00–14:00 with and the 21:00–22:00 consideration for ofboth diff caseserent toelasticity compensate coeffi cients.for the Twenty-oneheat deficits. configurations In addition, TE wereS discharged developed the with energy different during combinations 3:00–4:00 in of Case self- 2,and which cross-elasticity contributed coe toffi thecients, fact asthat presented more electric in Tableity 4was. In purchased addition, to from more the intuitively grid because investigate of the low the RTPeffect tariffs, of DR program,and the CHP Equation unit (29)generated could beless reshaped electricity as: and heat energy. Similarly, the CES unit released cooling energy during 12:00–14:00 and 20:00–21:00 when the cooling demands were at the RTP peak, as illustrated in FigureL 13e b.( Witht) = Lthee(t )coordination1 + e(t, t)Ps (oft) TES + e( tan, kd)Pc CES(t) systems, the MECS could(34) be economically and efficiently operated to meet diverse energy needs.

4.3. Sensitivity Experiments In this section, the effect of the DR scheme is investigated with the consideration of different elasticity coefficients. Twenty-one configurations were developed with different combinations of self- and cross-elasticity coefficients, as presented in Table 4. In addition, to more intuitively investigate the effect of DR program, Equation (29) could be reshaped as: RTP =+{ + } Lee() t L () t1 ettPst (,) () etkPct (, ) () (34)

ctct()− () Ps() t = RTP TOU (35) ctTOU () 24 ckck()− () Pc() t = RTP TOU  (36) k=1 ckTOU () kt≠

Table 4. Self- and cross-elasticity coefficients for each configuration.

# e(t,t) e(t,k) # e(t,t) e(t,k) 1 0 0.01 12 −0.20 0 2 −0.04 0.01 13 −0.20 0.005 3 −0.08 0.01 14 −0.20 0.015

Energies 2019, 12, 3995 18 of 21

cRTP(t) cTOU(t) Ps(t) = − (35) cTOU(t) 24 X cRTP(k) cTOU(k) Pc(t) = − (36) cTOU(k) k = 1 k , t

Table 4. Self- and cross-elasticity coeﬃcients for each conﬁguration.

# e(t,t) e(t,k)# e(t,t) e(t,k) 1 0 0.01 12 0.20 0 − 2 0.04 0.01 13 0.20 0.005 − − 3 0.08 0.01 14 0.20 0.015 − − 4 0.12 0.01 15 0.20 0.020 − − 5 0.16 0.01 16 0.20 0.025 − − 6 0.20 0.01 17 0.20 0.030 − − 7 0.24 0.01 18 0.20 0.035 − − 8 0.28 0.01 19 0.20 0.040 − − 9 0.32 0.01 20 0.20 0.045 − − 10 0.36 0.01 21 0.20 0.050 − − 11 0.40 0.01 Energies 2019, 12, x FOR PEER REVIEW − 18 of 21

The values for Ps and Pc 4at each−0.12 hour 0.01 are illustrated 15 −0.20 in Figure 0.020 14 , where it can be seen that the values of Pc were always negative5 −0.16 while Ps0.01was positive16 −0.20 or negative 0.025 at different time intervals. In Figure 15, the electricity demands6 with−0.20 DR 0.01 implementation 17 −0.20 for each0.030 configuration are illustrated. The electricity loads were shifted from7 peak−0.24 to valley,0.01 and the18 demand−0.20 curves0.035 were more smoothly modified with the increase of the absolute8 value−0.28 of0.01 the self-elasticity 19 −0.20 coe ffi0.040cient, as shown in Figure 15a. In Figure 15b, the peak loads are distinctly9 −0.32 curtailed0.01 while 20 the−0.20 valley 0.045 loads are changed little, a difference caused by the fact that the value10 of−0.36Pc was 0.01 away from 21 zero−0.20 at peak0.050 while Pc was close to zero at the valley, as shown in Figure 14. 11 −0.40 0.01

0.6 0.4 Ps Pc 0.2 0 -0.2 Value -0.4 -0.6 -0.8 -1 123456789101112131415161718192021222324 Time (h)

Figure 14. Values of Ps and Pc at each hour. Figure 14. Values of Ps and Pc at each hour.

1100 Config.1 Config.2 Config.3 Config.4 1100 Config.12 Config.13 Config.14 Config.15 1000 Config.5 Config.6 Config.7 Config.8 1000 Config.16 Config.17 Config.18 Config.19 900 Config.9 Config.10 Config.11 900 Config.20 Config.21 800 800 700 700 600 600 500 500 400 400 300 300 200 200 Electricity demand (kWh) Electricity demand (kWh) 100 100 0 0 123456789101112131415161718192021222324 123456789101112131415161718192021222324 Time (h) Time (h)

(a) (b)

Figure 15. Load profile of DR for each configuration: (a) Configurations 1–11 and (b) Configurations 12–21.

The values for Ps and Pc at each hour are illustrated in Figure 14, where it can be seen that the values of Pc were always negative while Ps was positive or negative at different time intervals. In Figure 15, the electricity demands with DR implementation for each configuration are illustrated. The electricity loads were shifted from peak to valley, and the demand curves were more smoothly modified with the increase of the absolute value of the self-elasticity coefficient, as shown in Figure 15a. In Figure 15b, the peak loads are distinctly curtailed while the valley loads are changed little, a difference caused by the fact that the value of Pc was away from zero at peak while Pc was close to zero at the valley, as shown in Figure 14. 3540 74400 3540 74000 Energy cost Exergy input Energy cost Exergy input 3520 74000 3520 73600 3500 3500 73600 3480 3480 73200 73200 3460 3460 Energy cost ($) Energy cost ($) cost Energy 72800 Exergy input (kWh) input Exergy 3440 72800 3440 Exergy input (kWh)

3420 72400 3420 72400 1234567891011 12 13 14 15 16 17 18 19 20 21 Configuration number Configuration number

Energies 2019, 12, x FOR PEER REVIEW 18 of 21

4 −0.12 0.01 15 −0.20 0.020 5 −0.16 0.01 16 −0.20 0.025 Energies 2019, 12, x FOR PEER REVIEW 6 −0.20 0.01 17 −0.20 0.030 18 of 21 7 −0.24 0.01 18 −0.20 0.035 4 8 −0.12−0.28 0.010.01 15 19− 0.20−0.20 0.020 0.040 5 9 −0.16−0.32 0.010.01 16 20− 0.20−0.20 0.025 0.045 6 10− 0.20−0.36 0.010.01 17 21− 0.20−0.20 0.030 0.050 7 11− 0.24−0.40 0.010.01 18 −0.20 0.035 8 −0.28 0.01 19 −0.20 0.040 9 −0.32 0.01 20 −0.20 0.045 0.6 10 −0.36 0.01 21 −0.20 0.050 0.4 Ps Pc 11 −0.40 0.01 0.2

0 0.6 -0.2

0.4Value Ps Pc -0.4 0.2 -0.6 0 -0.8 -0.2 -1 Value -0.4 123456789101112131415161718192021222324 -0.6 Time (h) Energies 2019, 12, 3995 -0.8 19 of 21 -1 Figure 14. Values of Ps and Pc at each hour. 123456789101112131415161718192021222324

1100 Config.1 Config.2 Config.3 Config.4 Time (h)1100 Config.12 Config.13 Config.14 Config.15 1000 Config.5 Config.6 Config.7 Config.8 1000 Config.16 Config.17 Config.18 Config.19 900 Config.9 Config.10 Config.11 900 Config.20 Config.21 800 800 700 Figure 14. Values of Ps and Pc at700 each hour. 600 600 500 500 400 400 1100 Config.1 Config.2 Config.3 Config.4 1100 300 300 Config.12 Config.13 Config.14 Config.15 1000 Config.5 Config.6 Config.7 Config.8 1000 Config.16 Config.17 Config.18 Config.19 200 200

900 Electricity demand (kWh) Electricity demand (kWh) Config.9 Config.10 Config.11 900 Config.20 Config.21 100 800 100 800 0 700 0 700 123456789101112131415161718192021222324 600 123456789101112131415161718192021222324 600 500 Time (h) 500 Time (h) 400 400 300 300 200 (a) 200 (b) Electricity demand (kWh) Electricity demand (kWh) 100 100 0 FigureFigure 15. 15. LoadLoad profile profile of DR of for DR each for configuration each configuration:: (a0) Configurations (a) Configurations 1–11 and (b 1–11) Configurations and (b) 123456789101112131415161718192021222324 123456789101112131415161718192021222324 Configurations12–21. 12–21.Time (h) Time (h)

(a) (b) AfterThe the values MO operationfor Ps and strategy Pc at each was hour applied are illustrated to each configuration, in Figure 14 the, where energy it cost,can be exergy seen input,that the exergyvaluesFigure effi ofciency 15. Pc Load were and profile energyalways of eDRnegativeffi ciencyfor each ofwhile configuration the preferredPs was : positive(a solution) Configurations or for negative each 1–11 case at and are different illustrated(b) Configurations time in intervals. Figure 16 .In TheFigure economic12–21. 15 , the and electricity exergetic objectivesdemands werewith allDR reduced implementation with the absolute for each value configuration increase of are the elasticityillustrated. coeThefficients, electricity as illustrated loads were in Figure shifted 16 froma,b, which peak meansto valley, that and a better the demand Pareto front curves was were obtained more with smoothly the strongermodifiedThe implementationvalues with for the Ps increase and of Pc the atof DR eachthe scheme. absolute hour are In value addition,illustrated of the the inself-elasticity changingFigure 14 trends, where coefficient, of itexergy can as be shown eseenfficiency that in Figure andthe valuesenergy15a. ofIn effi PcFigureciency were 15 were alwaysb, the completely peak negative loads opposite, while are distinctly Ps as was seen positive incurtailed Figure or 16while negativec,d. However,the valleyat different theloads variation timeare changed intervals. amplitudes little, In a Figurefordifference both 15 exergy, the caused electricity and by energy the demands fact effi thatciency with the were valueDR veryimplementation of Pc small, was away which for from means each zero configuration that at peak the DR while scheme are Pc illustrated. was had close less to Theinfluencezero electricity at the on exergyvalley, loads wereas and shown energy shifted in e Figurefromfficiency. peak 14. Therefore, to valley, and to distinctly the demand promote curves the were effect more of RTP-based smoothly modified3540 with the increase of the absolute value74400 of the self-elasticity3540 coefficient, as shown in Figure74000 DR, a better identificationEnergy cost of self-Exergy andinput cross-elasticity coefficients mustEnergy cost be conductedExergy input since the pricing 15modelsa. In 14 of3520. RTP play an important role. 74000 3520 73600 3500 3500 73600 3540 74400 3540 74000 3480 Energy cost Exergy input 3480 Energy cost Exergy input 73200 3520 73200 3460 74000 35203460 Energy cost ($) Energy cost ($) cost Energy 7360072800 Exergy input (kWh) input Exergy 3500 72800 3500 Exergy input (kWh) 3440 73600 3440 3480 3480 73200 3420 72400 3420 72400 73200 3460 1234567891011 3460 12 13 14 15 16 17 18 19 20 21 Energy cost ($) Configuration number ($) cost Energy Configuration number 72800 Exergy input (kWh) input Exergy 3440 72800 3440 Exergy input (kWh) 3420 72400 3420 72400 1234567891011 12 13 14 15 16 17 18 19 20 21 Configuration number Configuration number

Energies 2019, 12, x FOR PEER REVIEW(a) (b) 19 of 21

21.9 94.2 21.9 94.4 Exergy efficiency Energy efficiency Exergy efficiency Energy efficiency 21.88 94 94.2 21.85

21.86 93.8 94 21.8 21.84 93.6 93.8

21.75 93.6 21.82 93.4 Exergy efficiency (%) Exergy efficiency (%) Energy efficiency (%) Energy efficiency (%) efficiency Energy

21.8 93.2 21.7 93.4 1234567891011 12 13 14 15 16 17 18 19 20 21 Configuration number Configurarion number

FigureFigure 16. 16. EnergyEnergy cost, cost, exergy exergy input, input, exergy exergy efficiency efficiency and and energy energy efficiency efficiency for for each each configuration. configuration. (a()a )Energy Energy cost cost and and exergy exergy input input for for Configurations Configurations 1–11; 1–11; (b (b) )energy energy cost cost and and exergy exergy input input for for ConfigurationsConfigurations 12–21; 12–21; (c ()c )exergy exergy efficiency efficiency and and energy energy effi efficiencyciency for for Configurations Configurations 1–11; 1–11; (d (d) )exergy exergy efficiencyefficiency and and energy energy efficien efficiencycy for for Configurations Configurations 12–21. 12–21.

5. ConclusionsAfter the MO operation strategy was applied to each configuration, the energy cost, exergy input,In exergy this paper, efficiency an MECS and energy was employed efficiency to of fulﬁll the preferred the diverse solution energy for demands each case with are the illustrated coupling in of Figurediﬀerent 16. energyThe economic carriers and byenergy exergetic conversion objectives devices were all and reduced energy with storage the instruments.absolute value In increase addition, of the elasticity coefficients, as illustrated in Figure 16a,b, which means that a better Pareto front was obtained with the stronger implementation of the DR scheme. In addition, the changing trends of exergy efficiency and energy efficiency were completely opposite, as seen in Figure 16c,d. However, the variation amplitudes for both exergy and energy efficiency were very small, which means that the DR scheme had less influence on exergy and energy efficiency. Therefore, to distinctly promote the effect of RTP-based DR, a better identification of self- and cross-elasticity coefficients must be conducted since the pricing models of RTP play an important role.

5. Conclusions In this paper, an MECS was employed to fulfill the diverse energy demands with the coupling of different energy carriers by energy conversion devices and energy storage instruments. In addition, the economic–exergetic optimal scheduling model with an RTP-based DR scheme for an MECS was introduced to decrease operation costs and achieve energy savings by considering different qualities of energy. The ε-constraint approach was applied to this MO scheduling model to obtain the Pareto front, and a trade-off was conducted between these two objectives by the LINMAP approach. Moreover, sensitivity analyses of self- and cross-elasticity coefficients were explored and showed that the electricity load curve was modified to be smoother with the increase of the absolute value of the elasticity coefficient, and the economic and exergetic objectives were simultaneously decreased, while the exergy and energy efficiencies were not distinctly changed. Therefore, with the consideration of the RTP-based DR program in the optimal scheduling model, the MECS operator could more rationally allocate different energy carriers and simultaneously decrease energy costs and exergy inputs. However, to successfully implement the DR scheme, elasticity coefficients need to be more accurately identified, a need which will be investigated in the future.

Author Contributions: Y.H. formulated the research problem; S.L. reviewed the manuscript; Y.H., S.L., P.D. and Y.Z. modified the revised version; K.Y. proposed the mathematical model and algorithms; K.Y. and Y.Z. wrote the code and addressed the manuscript; W.Z. conducted the analysis of the conclusion; all of the authors were involved in writing this paper.

Funding: This work was funded by the Fundamental Research Funds for the Central Universities (Grant No. 2019MS099).

Conflicts of Interest: The authors declare no conflict of interest.

Reference

1. Steven, C.; Arun, M. Opportunities and challenges for a sustainable energy future. Nature 2012, 488, 294–303.

Energies 2019, 12, 3995 20 of 21 the economic–exergetic optimal scheduling model with an RTP-based DR scheme for an MECS was introduced to decrease operation costs and achieve energy savings by considering different qualities of energy. The ε-constraint approach was applied to this MO scheduling model to obtain the Pareto front, and a trade-off was conducted between these two objectives by the LINMAP approach. Moreover, sensitivity analyses of self- and cross-elasticity coefficients were explored and showed that the electricity load curve was modified to be smoother with the increase of the absolute value of the elasticity coefficient, and the economic and exergetic objectives were simultaneously decreased, while the exergy and energy efficiencies were not distinctly changed. Therefore, with the consideration of the RTP-based DR program in the optimal scheduling model, the MECS operator could more rationally allocate different energy carriers and simultaneously decrease energy costs and exergy inputs. However, to successfully implement the DR scheme, elasticity coefficients need to be more accurately identified, a need which will be investigated in the future.

Author Contributions: Y.H. formulated the research problem; S.L. reviewed the manuscript; Y.H., S.L., P.D. and Y.Z. modified the revised version; K.Y. proposed the mathematical model and algorithms; K.Y. and Y.Z. wrote the code and addressed the manuscript; W.Z. conducted the analysis of the conclusion; all of the authors were involved in writing this paper. Funding: This work was funded by the Fundamental Research Funds for the Central Universities (Grant No. 2019MS099). Conflicts of Interest: The authors declare no conflict of interest.

References

1. Steven, C.; Arun, M. Opportunities and challenges for a sustainable energy future. Nature 2012, 488, 294–303. 2. Geidl, M.; Koeppel, G.; Favre-Perrod, P.; Klöckl, B.; Andersson, G.; Fröhlich, K. Energy hubs for the future. IEEE Power Energy Mag. 2007, 5, 24–30. [CrossRef] 3. Krause, T.; Andersson, G.; Fröhlich, K.; Vaccaro, A. Multiple-energy carriers: Modeling of production, delivery, and consumption. Proc. IEEE 2011, 99, 15–27. [CrossRef] 4. Geidl, M.; Andersson, G. Optimal Power Flow of Multiple Energy Carriers. IEEE Trans. Power Syst. 2007, 22, 145–155. [CrossRef] 5. Wang, Y.; Zhang, N.; Kang, C.; Kirschen, D.S.; Yang, J.; Xia, Q. Standardized Matrix Modeling of Multiple Energy Systems. IEEE Trans. Smart Grid 2019, 10, 257–270. [CrossRef] 6. Liu, T.; Zhang, D.; Wang, S.; Wu, T. Standardized modelling and economic optimization of multi-carrier energy systems considering energy storage and demand response. Energy Convers. Manag. 2019, 182, 126–142. [CrossRef] 7. Moeini-Aghtaie, M.; Abbaspour, A.; Fotuhi-Firuzabad, M.; Hajipour, E. A decomposed solution to multiple-energy carriers optimal power flow. IEEE Trans. Power Syst. 2014, 29, 707–716. [CrossRef] 8. Moeini-Aghtaie, M.; Dehghanian, P.; Fotuhi-Firuzabad, M.; Abbaspour, A. Multiagent genetic algorithm: An online probabilistic view on economic dispatch of energy hubs constrained by wind availability. IEEE Trans. Sustain. Energy 2014, 5, 699–708. [CrossRef] 9. Yi, J.-H.; Ko, W.; Park, J.-K.; Park, H. Impact of carbon emission constraint on design of small scale multi-energy system. Energy 2018, 161, 792–808. [CrossRef] 10. Khorsand, H.; Seifi, A.R. Probabilistic energy flow for multi-carrier energy systems. Renew. Sustain. Energy Rev. 2018, 94, 989–997. [CrossRef] 11. Lu, Y.; Wang, S.; Sun, Y.; Yan, C. Optimal scheduling of buildings with energy generation and thermal energy storage under dynamic electricity pricing using mixed-integer nonlinear programming. Appl. Energy 2015, 147, 49–58. [CrossRef] 12. Li, Z.; Xu, Y. Optimal coordinated energy dispatch of a multi-energy microgrid in grid-connected and islanded modes. Appl. Energy 2018, 210, 974–986. [CrossRef] 13. Brahman, F.; Honarmand, M.; Jadid, S. Optimal electrical and thermal energy management of a residential energy hub, integrating demand response and energy storage system. Energy Build. 2015, 90, 65–75. [CrossRef] Energies 2019, 12, 3995 21 of 21

14. Di Somma, M.; Yan, B.; Bianco, N.; Graditi, G.; Luh, P.; Mongibello, L.; Naso, V. Operation optimization of a distributed energy system considering energy costs and exergy efficiency. Energy Convers. Manag. 2015, 103, 739–751. [CrossRef] 15. Torio, H.; Angelotti, A.; Schmidt, D. Exergy analysis of renewable energy-based climatisation systems for buildings: A critical view. Energy Build. 2009, 41, 248–271. [CrossRef] 16. Krause, T.; Kienzle, F.; Art, S.; Andersson, G. Maximizing exergy efficiency in multi-carrier energy systems. In Proceedings of the IEEE Power and Energy Society General Meeting, Providence, RI, USA, 25–29 July 2010; pp. 1–8. 17. Hepabsli, A. A Key Review on Exergetic Analysis and Assessment of Renewable Energy Resources for a Sustainable Future. Renew. Sustain. Energy Rev. 2008, 12, 593–661. [CrossRef] 18. Wang, J.; Yang, Y. Energy, exergy and environmental analysis of a hybrid combined cooling heating and power system utilizing biomass and solar energy. Energy Convers. Manag. 2016, 124, 566–577. [CrossRef] 19. Ahmadi, P.; Rosen, M.A.; Dincer, I. Multi-objective exergy-based optimization of a polygeneration energy system using an evolutionary algorithm. Energy 2012, 46, 21–31. [CrossRef] 20. Elizondo, L.M.R.; Paap, G.; Ammerlaan, R.; Negenborn, R.R.; Toonssen, R. On the energy, exergy and cost optimisation of multi-energy-carrier power systems. Int. J. Exergy 2013, 13, 364. [CrossRef] 21. Yan, B.; Di Somma, M.; Bianco, N.; Luh, P.B.; Graditi, G.; Mongibello, L.; Naso, V. Exergy-based operation optimization of a distributed energy system through the energy-supply chain. Appl. Therm. Eng. 2016, 101, 741–751. [CrossRef] 22. Luis, A.A.; Edwin, R.; Francisco, S.; Victor, H. A review and analysis of trends related to demand response. Energies 2018, 11, 1617. 23. Aalami, H.; Moghaddam, M.P.; Yousefi, G. Demand response modeling considering Interruptible/Curtailable loads and capacity market programs. Appl. Energy 2010, 87, 243–250. [CrossRef] 24. Albadi, M.; El-Saadany, E.; Albadi, M. A summary of demand response in electricity markets. Electr. Power Syst. Res. 2008, 78, 1989–1996. [CrossRef] 25. Erdinc, O.; Paterakis, N.G.; Mendes, T.D.P.; Bakirtzis, A.G.; Catalão, J.P.S. Smart Household Operation Considering Bi-Directional EV and ESS Utilization by Real-Time Pricing-Based DR. IEEE Trans. Smart Grid 2015, 6, 1281–1291. [CrossRef] 26. Zhi, C.; Lei, W.; Yong, F. Real-Time Price-Based Demand Response Management for Residential Appliances via Stochastic Optimization and Robust Optimization. IEEE Trans. Smart Grid 2012, 3, 1822–1831. 27. Aghamohamadi, M.; Hajiabadi, M.E.; Samadi, M. A novel approach to multi energy system operation in response to DR programs; an application to incentive-based and time-based schemes. Energy 2018, 156, 534–547. [CrossRef] 28. Ma, T.; Wu, J.; Hao, L. Energy flow modeling and optimal operation analysis of the micro energy grid based on energy hub. Energy Convers. Manag. 2017, 133, 292–306. [CrossRef] 29. Petela, R. Exergy of undiluted thermal radiation. Sol. Energy 2003, 74, 469–488. [CrossRef] 30. Candau, Y. On the exergy of radiation. Sol. Energy 2003, 75, 241–247. [CrossRef] 31. Aalami, H.; Yousefi, G.R.; Parsa Moghadam, M. A MADM-based support system for DR programs. In Proceedings of the 43rd Internatioal Universities Power Engineering Conference, Padova, Italy, 1–4 September 2008; pp. 1–7. 32. Aghaei, J.; Baharvandi, A.; Rabiee, A.; Akbari, M.A. Probabilistic PMU Placement in Electric Power Networks: An MILP-Based Multi-objective Model. IEEE Trans. Ind. Inform. 2015, 11, 332–341. [CrossRef] 33. Ma, T.; Wu, J.; Hao, L.; Li, D. Energy flow matrix modeling and optimal operation analysis of multi energy systems based on graph theory. Appl. Therm. Eng. 2019, 146, 648–663. [CrossRef] 34. Chow, T.; Pei, G.; Fong, K.; Lin, Z.; Chan, A.; Ji, J. Energy and exergy analysis of photovoltaic–thermal collector with and without glass cover. Appl. Energy 2009, 86, 310–316. [CrossRef] 35. Ahmadi, P.; Dincer, I. Exergoenvironmental analysis and optimization of a cogeneration plant system using Multimodal Genetic Algorithm (MGA). Energy 2010, 35, 5161–5172. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).