Regionalized LCA-Based Optimization of Building Energy

Article

pubs.acs.org/est

Regionalized LCA-Based Optimization of Building Energy Supply: Method and Case Study for a Swiss Municipality Dominik Saner, Carl Vadenbo, Bernhard Steubing,* and Stefanie Hellweg Group for Ecological Systems Design, Institute of Environmental Engineering, ETH Zurich, John-von-Neumann-Weg 9, 8093 Zurich, Switzerland

*S Supporting Information

ABSTRACT: This paper presents a regionalized LCA-based multiobjective optimization model of building energy demand and supply for the case of a Swiss municipality for the minimization of greenhouse gas emissions and particulate matter formation. The results show that the environmental improvement potential is very large: in the optimal case, greenhouse gas emissions from energy supply could be reduced by more than 75% and particulate emissions by over 50% in the municipality. This scenario supposes a drastic shift of heat supply systems from a fossil fuel dominated portfolio to a portfolio consisting of mainly heat pump and woodchip incineration systems. In addition to a change in heat supply technologies, roofs, windows and walls would need to be refurbished in more than 65% of the municipality’s buildings. The full potential of the environmental impact reductions will hardly be achieved in reality, particularly in the short term, for example, because of financial constraints and social acceptance, which were not taken into account in this study. Nevertheless, the results of the optimization model can help policy makers to identify the most effective measures for improvement at the decision making level, for example, at the building level for refurbishment and selection of heating systems or at the municipal level for designing district heating networks. Therefore, this work represents a starting point for designing effective incentives to reduce the environmental impact of buildings. While the results of the optimization model are specific to the municipality studied, the model could readily be adapted to other regions.

1. INTRODUCTION support decisions to lower the impact of the building stock, Environmentally extended input-output studies (EE-IOA) have regionalized bottom-up studies are needed, which model the heat shown that housing energy demand, in particular heating, makes and electricity demand and supply for individual buildings and up a large share of the overall environmental footprint of municipalities within their local context. 1,2 A suitable method for such analyses is life cycle assessment households, exceeding 20% of global greenhouse gas emissions 5 caused by households.1 These results illustrate the need to (LCA). Several LCA studies have been conducted to analyze the reduce the environmental impact, in particular climate change environmental impacts of households or buildings. There seems eﬀects, of the housing sector. Switzerland is an example of a to be a general consensus that the environmental performance of country where the housing sector contributes about 25% of the buildings is related to a number of interdependent factors, such as the construction typology, geographic location, and user overall CO2-footprint, which corresponds to more than 3 tonnes 3 behavior, and that the operational heat demand is usually the per person and year. This exceeds the long-term goal of the − “ ” main driver of environmental impacts.6 12 Suggested measures so-called 1 tonne CO2-society (i.e., reducing overall CO2-emissions to 1 t/capita/year)4 to combat climate change and hence calls for to decrease the environmental impact of buildings relate mainly action to reduce this impact. to lowering energy consumption, for example, through thermal While EE-IOA helps to target relevant consumption areas at insulation of the building envelope and reducing the impact of ﬃ larger scales such as nations, it does not usually contain enough heat supply, for example, through e ciency measures or 7,9,10 detail to capture the situation at the level of municipalities or renewables. Since measures to reduce the energy demand individual buildings. It is, however, exactly at these levels that of buildings are increasingly required by regulations and may many important decisions are taken, concerning, for example, come at the cost of higher embodied energy in building materials, building design and refurbishment, the choice of heating systems or the construction of district heating networks. Other local Received: January 13, 2014 factors also play a role in reducing the environmental impact of Revised: May 23, 2014 the building stock, such as the climate and the availability of Accepted: May 27, 2014 renewable energy sources. Therefore, in order to adequately Published: May 27, 2014

© 2014 American Chemical Society 7651 dx.doi.org/10.1021/es500151q | Environ. Sci. Technol. 2014, 48, 7651−7659 Environmental Science & Technology Article several authors emphasize that the whole life cycle should be households in 1332 buildings. Geo-referenced energy demands − considered when designing or retrofitting buildings.7,8,10 12 for individual households for the reference year 2010 were While analyses and comparisons of alternative scenarios are calculated in a previous study.32 For the case study presented commonly performed in LCA studies, this does not ensure that here, these household demands were aggregated to the building optimal solutions are identified, especially when there is a large level. In order to quantify potential savings in environmental number of possible scenarios resulting from different technology impacts by refurbishment and thus a reduction of heating combinations and, for example, building-specific and local demand, four building renovation measures (floor, roof, wall, and constraints. To overcome this problem, LCA has been combined window refurbishments) were included in the model. The actual with mathematical optimization techniques already in the mid- renovation status of each building was available from Saner 1990s.13,14 Since then, studies have addressed environmental et al.32 The space heat demand was calculated for each individual optimization problems related to scheduling and process building with and without each refurbishment measure (i.e., − − design15 18 as well as planning of supply chain networks19 25 refurbished windows, refurbished roof, etc.), assuming that all (see also26 for a review of the integration of process optimization refurbishments would meet best available standards. The techniques with the LCA methodology). difference corresponded to the amount of space heat that Several studies have combined LCA and optimization could be saved (i.e., supplied) from each renovation measure. approaches on a building level, for example, to optimize building Obviously, for buildings that had recently been renovated the envelopes27 or building energy supply.28,29 Other authors have potential savings were close to zero, while they were substantial presented approaches on how to quantify, with high geographic for old buildings with poor insulation. The environmental impact 30 31 resolution, the energy demand and CO2 emissions of cities. from the production of insulation materials for the refurbishment However, to the best of our knowledge, an LCA-based optimization were calculated per year (considering the lifetime of the material) of building energy supplies within the context of real settlements and as reported in the Supporting Information (SI) (section S1.2.2) for an entire municipality has not been presented previously. of Saner et al.32 This paper presents an LCA based optimization approach with 2.2. Energy Supply. The possible space heat and hot water the goal to minimize building related environmental impacts by supply technologies were oil, gas, and wood heaters (chips, logs, optimizing the energy supply through alternative technologies pellets), heat pumps (brine−water, air−water), district heat and refurbishment measures. For the case of a Swiss municipal- (woodchips), and polymer electrolyte membrane (PEM) fuel ity,32 two geographical scales are considered simultaneously: the cell systems (cogeneration of heat and electricity fueled by gas). level of individual buildings and the level of the municipality. Hot water could additionally be partly supplied by solar collector This disaggregated system perspective enables consideration of panels. Electricity demand could be supplied by electricity from system-wide constraints such as capacity limits and resource- local photovoltaic (PV) panels (ribbon-Si) mounted on roof supply constraints, as well as site-specific factors such as the tops, PEM fuel cells or electricity from the Swiss grid. For possibility to join a district heating network, to exploit electricity from PV, it was assumed that 50% of the rooftops in groundwater heat pumps or to install solar panels or collectors. the case study municipality could be equipped with panels. It was assumed that PV panels could only occupy the residual roof area 2. GOAL AND SCOPE DEFINITION after subtracting the area of the solar collectors from the total The aim of the study was to identify the theoretic environmental roof area. PEM fuel cells could only be run with natural gas or fi improvement potential that can be achieved by changing energy puri ed biogas and thus only in buildings with a connection to supply systems for space heat, hot water, and electricity and by the gas network. Ground source heat pump systems were only refurbishing the building stock in a Swiss municipality. As a case allowed for buildings situated in designated areas where the risk study we chose the municipality of Wattwil because another of groundwater contamination by leaking borehole heat recent study already estimated current household energy exchangers was small. A certain area was suitable for a district demands and supplies for the building stock of this municipality,32 heat grid, and all buildings situated in this area had the possibility which was an ideal point of departure for the present paper. We to choose district heat as a supply option. The supply of locally available woodchips within the municipality was 10 000 m3/year35 assumed that the population of the municipality would not grow 3 (as projected for the specific municipality studied), the building and that of wood logs 2500 m /year with shares of 28% hard wood park would remain constant (since there were no new building and 72% soft wood. Both chips and logs were assumed to be areas planned) and that the demand for per-capita living space residues from round wood production from sustainably managed forests. Potential long-term effects of removing wood from the would remain constant. The results of the study should serve as an 36 overall benchmark and to pinpoint specific measures for the forest, for example, on soil fertility, were not assessed in this study, partly because there is not yet a consistent methodology reduction of environmental impact. 37 The functional unit was defined as providing a heated living implemented in standard life cycle impact assessment methods. space as well as hot water and electricity for each building in the The supply with wood from outside the municipality was municipality. Environmental impacts were assessed for two assumed to be zero as other regions may need their local midpoint categories from the ReCiPe method33 with particular renewable energy sources for themselves. All background life cycle inventories except for ground source heat pumps38 were relevance for energy supply systems: climate change (kg CO2-eq ) 39 and particulate matter formation (kg PM10-eq ). While climate taken from the ecoinvent database (v2.2). The complete list of change is currently the environmental dimension receiving the life cycle inventories for the included energy technologies is highest political attention and an indicator of the fossil fuel provided in Table S4 in the SI. intensity of products, we chose particulate matter formation as a second indicator as we expected trade-offs with climate change, 3. OPTIMIZATION MODEL especially in the context of using wood energy.34 3.1. Distinction of Foreground and Background 2.1. Energy Demand. The case study municipality is Systems through Matrix Extension. Since the building situated in the eastern part of Switzerland. It consists of 3328 energy supply and refurbishment decisions involved are made on

7652 dx.doi.org/10.1021/es500151q | Environ. Sci. Technol. 2014, 48, 7651−7659 Environmental Science & Technology Article a local level, an explicit distinction was made between a local Table 2. Parameters and Variables Used in the Optimization foreground system, which could be influenced by the decision makers, and the background system, which consisted of those Parameter Description upstream processes that could not be influenced by the decision A technology matrix ̆ makers.40 In mathematical terms, this was implemented by using â element of the inverted background technology matrix A ̆ two distinct technology matrices: a square A matrix representing ă element of the foreground technology matrix A the background system (the ecoinvent technology matrix39) and b element of the environmental intervention matrix B ’ f element of the total final demand matrix F a rectangular A matrix representing the (underdetermined) ̆ ̆ foreground system. The foreground system matrix defined the f element of the total final demand matrix F L L competing technologies, energy carriers and refurbishment h element of the lower bound characterization result vector h U U options used in the municipality as well as interactions between h element of the upper bound characterization result vector h ren ren these. A’ can be considered an extension of the A matrix. A h element of the renovation characterization result matrix h m sufficiently large constant used to control the scaling vector based detailed description of the implementation and an example are on the binary decision variables provided in the SI. pdyn element of the time-specific constraint matrix pdyn 3.2. Mathematical Formulation. − Equations 1 18 describe pind element of the building-specific constraint matrix pind the mathematical formulation of the optimization (the pind,dyn element of the building- and time-specific constraint matrix pind,dyn nomenclature used is summarized in Table 1 and Table 2). To psys element of the process-specific constraint matrix psys q element of the characterization matrix Q Table 1. Indices and Subsets of Indices Used in the Variable Description Optimization λ auxiliary scalar variable describing degree of satisfaction (to be maximized) in the fuzzy LP formulation number of index description elements s element of the background scaling matrix S ̆ ̆ i full set of processes 4243 s element of the foreground scaling matrix S ind j economic flows in the background system 4243 p element of the binary decision variable for decisions on the level of individual buildings fl k economic ows in the foreground system (space heat, 6 sys hot water, electricity, natural gas or biogas, p Binary decision variable for decisions on system level, i.e. district woodchips, and logs) heating network l set of buildings 1332 t time steps (months) 12 below. The building-specific scaling vector s̆of the foreground e environmental exchanges 4373 system and the resulting global scaling vector s represent two of c characterization factors of life impact assessment 2 the decision variables to be optimized in order to supply the final methods demand f while minimizing the environmental impacts from a iU subset of i for processes with an upper limit on total 9 supply systems perspective. We also introduce two binary decision ind sys fl iL subset of i for processes for which a minimum 2 variables (u and u ) in order to re ect investment decisions of (threshold) demand has to be reached, i.e. heat and individual buildings and the municipality, respectively. This hot water supplied over a district network addition converted the problem into a mixed-integer program iLheating subset of iL representing heat supply over a district 1 (MIP). The optimization was performed for the total supply with heating network energy (i.e., space heat, hot water, and electricity) over one year. iLhotwater subset of iL representing hot water supply over a district 1 water network The year was divided into 12 time steps t, that is, months, to be able iheating subset of i for processes providing space heat 13 to account for temporal constraints (e.g., unusable surplus heat from ihotwater subset of i for processes providing hot water 13 solar collector systems in summer). The mathematical formulation iremovation subset of i for renovation measures 4 of the optimization problem is outlined below and presented in iil subset of i for processes with limited spatial availability 32 detail in Section S1.4 in the SI. The optimization was performed 41 iit subset of i for processes with limited temporal 3 using the General Algebraic Modeling System (GAMS). availability maximize λ (1) iilt subset of i for processes with limited spatial and 2 temporal availability Subject to ··+ind ·ren ∑∑∑qbsuhce, ej,, jl ∑∑ilcirenovation, ,, renovation l be able to minimize several life cycle impact categories at once in e jl ilrenovation a multiobjective optimization, we apply the fuzzy LP extension 25 U U L of the general matrix-based LCA developed by Tan et al. ≤−hhhcc λ()c −c ∀ (2) Compared to general LP formulation of matrix-based LCA (as described in SI), the original objective function (minimize h = 01≤≤λ (3) QBs, SI eq S.2) is replaced by eqs 1−3. Equation 4 represents the as̆ · ̆ ≥ f̆ ∀ klt,, inequality for which the final demand needs to be satisfied by the ∑ kit,, ilt ,, klt,, (4) product of the (rectangular) technology matrix and the scaling i fi vector, which should satisfy the unique nal demand vector of AA−1 = ̂ (5) each building (As ≥ f, SI eq S.3). Considering the entire building = ̂ · ̆ +∀ stock results in a scaling matrix where the number of columns is sjl,,,,∑∑asfjji (),ilt ilt,, l equal to the total number of buildings in the municipality. i t (6) Equations 5−6 replace SI eq S.4 (s = A−1 (s̃+ f)), and eqs 7−18 ̆ ≤∀∈sys U represent the constraints imposed on the system (SI eq S.5). ∑∑spiiilt,, i These constraints are described in more detail in the sections lt (7)

7653 dx.doi.org/10.1021/es500151q | Environ. Sci. Technol. 2014, 48, 7651−7659 Environmental Science & Technology Article

32 ̆ ≤∀∈dyn it calculated individually for each building according to Saner et al. ∑ spiililt,, it, , l (8) These impacts are accounted for in the second term on the left- hand side of eq 2. The binary variable uind reflects the decision of ̆ ≤∀∈ind il whether a possible renovation measure is chosen for each building ∑ spiitilt,, il, , t (9) considered. 3.2.2. System Constraints. Some of the systems constraints in ̆ ≤∀∈ind,dyn ilt eqs 7−18 are related and can be divided into blocks of equations silt,, piiltilt,, ,, (10) that impose limitations on the decision space for the optimal ind solution. Equations 7−10 represent different types of capacity uil, =∀{0; 1}i , l (11) constraints in terms of process availability based on spatial sys ui=∀{0; 1} (12) and/or temporal aspects. The subsets of the process index i i (summarized in Table 1) were used to address constraints fi ff U ∑ smuiĭ ≤·ind ∀∈ heating , l speci c for di erent processes. In eq 7, i is used to introduce an ilt,, il , upper limit (psys) for the system-wide supply by processes that t (13) are neither building- nor time-specific (e.g., the amount of locally 3 ulind ≤∀1 available hard and soft woodchips: 10 000 m ). Other subsets ∑ il, fi i∈iheating (14) were de ned to address process constraints that are only time- specific(pdyn) in eq 8 (subset iit, e.g., constraint for the temporal s̆ − s̆ ≥∀∈0,,, iihotwater i ′∈ i heating ii =′l availability of electricity from photovoltaic panels), specificto ∑∑ilt,,ilt′ ,, ind il t t the individual buildings (p ) as in eq 9 (subset i , e.g., constraint (15) for the spatial restriction for the use of ground source heat), or building- and time-specific(pind,dyn) as in eq 10 (subset iit, e.g., ind sys L uil, ≤∀∈u, i i, i l (16) constraint for the spatial and temporal availability of heat from solar collectors). − ∑∑ssilt̆,,+ ∑∑ ilt̆′ ,, The constraints in eqs 11 15 relate to process selection lt lt (based on binary decision variables) for the individual buildings and for the municipality. Buildings usually only have one major ≥·∀∈pusys sys iiLL heating,, ii ∈ hotwater ii =′ i i (17) hydronic energy supply system. It is, for example, almost never

ind renovation the case that energy is supplied by both an oil boiler and by a ∑ smuiiilt̆,,≤· il , ∀∈ , l natural gas furnace. Thus, the decision space is restricted to t (18) only one major energy system by introducing the binary variable ind A list of the indices and subsets used in eqs 1−18 is provided in u . However, there is still the possibility that part of the heat Table 1, and the parameters and variables used are summarized in for hot water could be delivered by an auxiliary device (i.e., solar collectors) (see eqs 13−15). The parameter m is a sufficiently Table 2. ind 3.2.1. Objective Functions. There are a wide range of different large constant, which allows the binary decision variable u to š methods available to solve multiobjective or multicriteria control the corresponding element of the foreground system optimization problems.42 The solution to these problems scaling matrix. typically consists of a set of Pareto-optimal alternatives that The dependencies between the demand and availability of the ffi ff district heating or hot water network are reflected in eqs 16 and 17. represent the e cient trade-o s between the objectives L considered.43 The concept of Pareto-optimality implies that it The index i is used in eq 16 to represent processes with a threshold sys fi is not feasible to improve the performance in one objective demand (p ) that had to be achieved to make a speci cenergy without compromising another objective. An alternative strategy supply option available for use (e.g., threshold for energy supply via ff a district heat system, assumed to be 1.8 GWh per year). The subsets is o ered by the fuzzy LP extension for matrix-based LCA Lheating Lhotwater developed by Tan et al.25 Following this approach several impact i and i include processes that provide energy for space categories are linearly related to a scalar denoted λ. It expresses heat and hot water, respectively, over a district heat network in U eq 17. In eq 16, uind are controlled by usys, that is, whether a district the degree of satisfaction based on the estimated upper (hc ) and L heating network is made available in the municipality or not, and lower bounds (hc ) for the objectives. The upper and lower eq 17 enforces the energy demand threshold constraints upon the bounds might, for example, represent conservative and ambitious sys environmental targets respectively for each objective. The decision to invest in a district heating system. The binary variable u optimization problem is thus converted to the maximization of describes whether a district heating system is an optimal option for the degree of satisfaction that can be obtained for all objectives the municipality. simultaneously. In the case study, the lower bound for each impact category (global warming potential and particulate 4. RESULTS emissions) was determined by solving the optimization problem Figure 1 shows two maps of the case study municipality, separately for each objective. During the single-objective depicting the annual life cycle greenhouse gas (GHG) emissions optimizations, the impact results for the other impact categories per hectare grid cell and year (including indirect emissions that were simultaneously calculated. The maximum value for each occur, that is, during the production and end-of-life of energy category over all generated solutions provided the upper bounds, carriers and technologies). Impact results are aggregated per that is, any constraints in terms of the required performance level hectare (100 × 100 m) to ensure data privacy of single buildings. of the heating technologies were disregarded for this study. For the The left map shows the impact results for the reference case, minimization of the impact results, λ was maximized (see eqs 1−3). whereas the right map shows the results of the multiobjective The environmental impacts for the four renovation measures optimization. The scale ranges from less than 0.5 to more than 32 ren fl h (i.e., oor, roof, walls, and windows refurbishment) were tons CO2-eq per hectare (colors encode quintiles).

7654 dx.doi.org/10.1021/es500151q | Environ. Sci. Technol. 2014, 48, 7651−7659 Environmental Science & Technology Article

Figure 1. Results for the energy supply of buildings in a Swiss municipality for the reference case (left) and the multiobjective optimal case (right), aggregated per hectare. Top: annual life cycle GHG emissions. Middle: distribution of hydronic energy supply systems. Bottom: solar collectors as auxiliary hot water supply devices (In the gray shaded areas only air-sourced heat pumps are allowed, borehole heat exchangers are prohibited due to groundwater protection).

In both cases, the highest GHG emissions are in the center of systems of the buildings in the city center rely mainly on fossil the municipality due to the higher population density and bigger fuels, whereas buildings in the outskirts of the municipality are buildings. Additionally, in the reference system, the heating mainly heated with wood. In the optimal case, GHG emissions

7655 dx.doi.org/10.1021/es500151q | Environ. Sci. Technol. 2014, 48, 7651−7659 Environmental Science & Technology Article are distributed more equally than in the reference case. GHG 41% decentralized woodchip incineration and 6% decentralized emissions above 32 tons CO2-eq per hectare are rare. Locally wood log incineration. On the contrary, in the case of particulate increased GHG emissions can be explained by a switch of the energy matter formation minimization, refurbishment rates are high supply system from a purely wood based system with low GHG (91% for roofs, 95% for walls, 89% for windows, and 51% for emissions to another system (e.g., heat pumps) in order to achieve a floors). The heat supply consists in this case of a mixture of heat better trade-off between GHG and particulate matter emissions. pump systems (21% ground sourced and 52% air sourced) and In the optimal case, GHG emissions are reduced by more than PEM fuel cells (27%). A trade-off therefore exists between the 75% compared to the reference case, from 11 824 to 2664 tons two optimization objectives: a minimization of GHG emissions CO2-eq. Also, particulate matter emissions are reduced by more by installing wood heating systems can only be realized at the than 50% from 10.7 to 5.12 tons PM10-eq This means that due to cost of higher particulate matter emissions. The single-objective the changing structure in energy supply, the impacts from both optimization results are provided in the SI. impacts categories decrease. The multiobjective optimization applied here searches for the best consensus between the two 5. DISCUSSION ff objectives and therefore leads to a fundamentally di erent The aim of this study was to analyze the potential of reducing portfolio of hydronic energy supply systems in the municipality environmental impacts related to the energy supply of 1332 than in the reference case. buildings in a Swiss municipality based on spatially explicit Figure 1 shows the distribution of energy supply systems, energy demand and infrastructure data, life cycle inventories for that is, space heat and hot water supply systems (middle) and energy technologies, and different constraints. auxiliary hot water supply with solar collectors (bottom) in the Compared to a standard LCA study, a major advantage of an case study municipality both for the reference case (left) and the optimization approach is that it is possible to take into account optimal case (right). Each pixel depicts the most frequent individual supply constraints (e.g., prohibition of ground source system per hectare. The gray area describes where borehole heat heat pump systems in certain areas) at the same time as systemic exchangers and thus brine-water heat pumps are prohibited due constraints that concern the energy supply of the whole to protection of groundwater. municipality. For instance, district heating was only an option In the reference situation, the most common supply systems in for individual buildings if the overall demand for a district heating the city center are oil boilers and gas furnaces, thereby 1156 t system was high enough (above threshold). In a standard LCA, of light fuel oil and 1418 Nm3 of gas are used per year. In the all of these constraints and local conditions would have to be outskirts, wood incineration is predominant (11 m3 of wood- addressed and compared in a (unfeasibly large) number of chips and 4583 m3 wood logs). Only in some cases the most alternative scenarios. In a linear programming extension of LCA, frequent systems are direct electric boilers and heat pumps. The only the constraints have to be defined while an algorithm looks total electricity use for housing purposes is 4.76 GWh per year. for the best solution among all possible scenarios allowing for Solar collector panel installations, delivering auxiliary heat for hot system-wide analyses to be performed. water generation, are very rare in the municipality (Figure 1). In The study highlights trade-offs between two different environ- the optimal case, however, solar collectors are installed to the mental impact categories with particular relevance for residential fullest possible extent. The distribution of the space heat supply heat systems. Policies aimed at minimizing only one indicator, for systems is rather homogeneous and heat pump systems (brine- example, GHG emissions, may lead to considerable impacts in water or air−water) as well as woodchips are preferred. Heat other categories, for example, PM emissions. Policies should pumps are predominant in the areas where drilling boreholes for therefore rely on studies that identify and present solutions for heat exchanger tubes is allowed. Woodchip incineration systems existing trade-offs from a system perspective. are predominant in the other areas. In total, 8452 m3 of An important finding of the study is that approximately 75% of woodchips and only 418 m3 of wood logs are used. The total the municipality’s GHG emissions and 50% of the PM emissions electricity demand for housing purposes almost quadruples due could be reduced simultaneously. A comparison to a recent study to the installation of heat pump systems and amounts to 16.4 GWh by Heeren et al.,44 who calculated a GHG reduction potential of per year (50% electricity from PV and 50% from the grid delivered 85% for the city of Zurich, suggests that the magnitude of these from outside the municipality). potential impact reductions is reasonable. At the same time, the Figure 2 shows whether each of the four possible refurbish- municipality could drastically minimize its dependency on fossil ment measures, respectively, is optimal for the majority of the energy resources to supply the energy demand of its building buildings located within a given hectare cell. The refurbishment stock. Although these figures are very encouraging in view of rate is the number of buildings that are refurbished divided by the sustainable future energy supplies, they should be seen as long- total number of buildings in the municipality. The multiobjective term targets for environmental improvement potentials rather optimization suggests that optimal refurbishment rates are 67% than realistic goals in the near future. Our underlying model for roofs, 81% for walls and 68% for windows, and only 5% for remains theoretical in the sense that neither economic aspects floors. However, refurbishment rates in the single-objective nor the willingness of households to take action regarding the optimizations are quite different: if only GHG emissions are installation of new technologies and the refurbishment of their minimized, refurbishment rates are low for all measures (8% for homes were taken into consideration. Further research is therefore roofs, 28% for walls, 3% for windows, and 0% for floors). This is necessary to take these factors into account and investigate, for because GHG emissions released during the production of example, cost-optimal options to reduce environmental impacts. building parts are higher than those of the additional heat Nevertheless, from an environmental perspective the study requirements. Due to the rural setting of the municipality with showed that energy technology replacement and high rates of surrounding forests, the wood supply is sufficient to provide heat refurbishment are desirable in spite of the environmental impacts and hot water to all buildings. Therefore, in the case of GHG associated with their production. emissions, heat is supplied entirely by wood-based systems: 53% The high reduction potential of GHG and PM emissions centralized woodchip incineration with district heat delivery, should be seen in the context of a rural Swiss setting where wood

7656 dx.doi.org/10.1021/es500151q | Environ. Sci. Technol. 2014, 48, 7651−7659 Environmental Science & Technology Article

Figure 2. Optimal decisions for four possible refurbishment options: roof, wall, floor, and window refurbishments. heating has a tradition and is, due to the used technology, living space. The results of the model can serve as a benchmark the source of a considerable share of total PM concentration.34 for policy makers, quantifying maximum reduction potentials in More stringent emission limits for PM-emissions would reduce environmental impact. They also pinpoint specific decisions that the feasible decision space in the optimization problem and can make a difference and thus help to design incentive systems presumably constrain the reliance on wood heating or require a for sustainable urban energy systems. Finally, the results can be shift toward “cleaner” wood furnaces. In an urban setting, bio- helpful to identify pathways to meet political goals, such as the mass availability can be assumed to be a limiting factor. energy turnaround in Switzerland (switching to a low-impact Moreover, PM emissions should be evaluated differently as the energy supply without nuclear power). population density is higher.45 Some of the results are therefore specific to the case study. However, given the availability of data, ■ ASSOCIATED CONTENT an adaptation of the model to other regions and constraints *S Supporting Information should be possible. Supporting Information describes the implementation of the Further model improvements are possible, for example, optimization extension including an illustrative example and considering future technology development, including cost provides additional results for the single-objective optimizations data, taking into account building park renewal rates and changes of the case study. This material is available free of charge via the in population number, and changed demand in, for example, Internet at http://pubs.acs.org/.

7657 dx.doi.org/10.1021/es500151q | Environ. Sci. Technol. 2014, 48, 7651−7659 Environmental Science & Technology Article ■ AUTHOR INFORMATION (17) Guillen-Gosá lbez,́ G.; Caballero, J. A.; Jimenez,́ L. Application of life cycle assessment to the structural optimization of process flowsheets. Corresponding Author Ind. Eng. Chem. Res. 2008, 47, 777−789. *Phone: +41 44 633 76 39; e-mail: [email protected]. (18) Gerber, L.; Gassner, M.; Marechal,́ F. Systematic integration of Notes LCA in process systems design: Application to combined fuel and fi electricity production from lignocellulosic biomass. Comput. Chem. Eng. The authors declare no competing nancial interest. 2011, 35, 1265−1280. (19) Hugo, A.; Pistikopoulos, E. N. Environmentally conscious long- ■ ACKNOWLEDGMENTS range planning and design of supply chain networks. J. Clean. Prod. 2005, 13, 1471−1491. Dominik Saner was funded by the Competence Center for (20) Guillen-Gosá lbez,́ G.; Grossmann, I. E. Optimal design and Energy & Mobility (CCEM) and swisselectric Research within planning of sustainable chemical supply chains under uncertainty. the THELMA project (www.thelma-emobility.net). Carl Vadenbo − fi AIChE J. 2009, 55,99 121. and Bernhard Steubing are grateful for the nancial support from (21) Guillen-Gosá lbez,́ G.; Grossmann, I. E. A global optimization the Swiss Competence Center for Energy Research (SCCER) strategy for the environmentally conscious design of chemical supply “Efficient Technologies and Systems for Mobility”.Wethank chains under uncertainty in the damage assessment model. Comput. Catherine Raptis and Justin Boucher for the English proofreading Chem. Eng. 2010, 34,42−58. and Alejandro Alonso for the design of the TOC art. The com- (22) Keckler, S. E.; Allen, D. T. Material reuse modelingA case study ments from two anonymous reviewers are gratefully acknowledged. in an industrial park. J. Ind. Ecol. 1999, 2,79−92. (23) Mellor, W.; Wright, E.; Clift, R.; Azapagic, A. A mathematical ■ REFERENCES model and decision-support framework for material recovery, recycling and cascaded use. Chem. Eng. Sci. 2002, 57, 4697−4713. (1) Hertwich, E. G.; Peters, G. P. Carbon footprint of nations: A global, (24) Nakatani, J.; Hirao, M. Multicriteria design of plastic recycling trade-linked analysis. Environ. Sci. Technol. 2009, 43 (16), 6414−6420. based on quality information and environmental impacts. J. Ind. Ecol. (2) Tukker, A.; Jansen, B. Environmental impacts of products: A 2011, 15, 228−244. detailed review of studies. J. Ind. Ecol. 2006, 10 (3), 159−182. (25) Tan, R. R.; Culaba, a.; Aviso, K. A fuzzy linear programming (3) Jungbluth, N.; Nathani, C.;Stucki,M.;Leuenberger,M. extension of the general matrix-based life cycle model. J. Clean. Prod. Environmental Impacts of Swiss Consumption and ProductionA 2008, 16, 1358−1367. Combination of Input-Output Analysis with Life Cycle Assessment; Federal (26) Pieragostini, C.; Mussati, M. C.; Aguirre, P. On process Office for the Environment (FOEN): Bern, Switzerland, 2011; p 171. optimization considering LCA methodology. J. Environ. Manage. (4) Swiss Federal Office of Environment, ETH-Energiestrategie - Die 2012, 96,43−54. 1-Tonne-CO2-Gesellschaft [Energy strategies - the one tonne CO2 (27) Stazi, F.; Mastrucci, A.; Munafo,̀ P. Life cycle assessment approach society]. Umwelt, 2009; pp 12−13. for the optimization of sustainable building envelopes: An application (5) ISO. International Standard 14044:2006. In Environmental on solar wall systems. Build. Environ. 2012, 58, 278−288.   Management Life Cycle Assessment Requirements and Guidelines.; (28) Jing, Y. Y.; Bai, H.; Wang, J. J. Multi-objective optimization design ISO (International Organisation for Standardisation): Geneva, Switzer- and operation strategy analysis of BCHP system based on life cycle land, 2006. assessment. Energy 2012, 37 (1), 405−416. (6) Iyer-Raniga, U.; Wong, J. P. C. Evaluation of whole life cycle (29) Fabrizio, E.; Corrado, V.; Filippi, M. A model to design and assessment for heritage buildings in Australia. Build.Environ. 2012, 47 − optimize multi-energy systems in buildings at the design concept stage. (1), 138 149. Renewable Energy 2010, 35 (3), 644−655. (7) Asdrubali, F.; Baldassarri, C.; Fthenakis, V. Life cycle analysis in the (30) Girardin, L.; Marechal, F.; Dubuis, M.; Calame-Darbellay, N.; construction sector: Guiding the optimization of conventional Italian − Favrat, D. EnerGis: A geographical information based system for the buildings. Energy Build. 2013, 64,73 89. evaluation of integrated energy conversion systems in urban areas. (8) Beccali, M.; Cellura, M.; Fontana, M.; Longo, S.; Mistretta, M. Energy 2010, 35 (2), 830−840. Energy retrofit of a single-family house: Life cycle net energy saving and (31) Gurney, K. R.; Razlivanov, I.; Song, Y.; Zhou, Y.; Benes, B.; Abdul- environmental benefits. Renewable Sustainable Energy Rev. 2013, 27, Massih, M. Quantification of fossil fuel CO emissions on the building/ 283−293. 2 street scale for a large U.S. City. Environ. Sci. Technol. 2012, 46 (21), (9) Gustavsson, L.; Joelsson, A.; Sathre, R. Life cycle primary energy 12194−12202. use and carbon emission of an eight-storey wood-framed apartment ̈ building. Energy Build. 2010, 42 (2), 230−242. (32) Saner, D.; Heeren, N.; Jaggi, B.; Waraich, R. A.; Hellweg, S. Housing and mobility demands of individual households and their life (10) Ardente, F.; Beccali, M.; Cellura, M.; Mistretta, M. Energy and − environmental benefits in public buildings as a result of retrofit actions. cycle assessment. Environ. Sci. Technol. 2013, 47 (11), 5988 5997. Renewable Sustainable Energy Rev. 2011, 15 (1), 460−470. (33) Goedkoop, M.; Heijungs, R.; Huijbregts, M. A. J.; Schryver, A. D.; Struijs, J.; Zelm, R. V. ReCiPe 2008; 2009. (11) Yung, P.; Lam, K. C.; Yu, C. An audit of life cycle energy analyses ́̂ of buildings. Habitat Int. 2013, 39,43−54. (34) Szidat, S.; Prevot, A. S. H.; Sandradewi, J.; Alfarra, M. R.; Synal, H. (12) Ramesh, T.; Prakash, R.; Shukla, K. K. Life cycle energy analysis of A.; Wacker, L.; Baltensperger, U., Dominant impact of residential wood buildings: An overview. Energy Build. 2010, 42 (10), 1592−1600. burning on particulate matter in Alpine valleys during winter. Geophys. (13) Azapagic, A.; Clift, R. Life cycle assessment and linear Res. Lett. 2007, 34, (5). programming environmental optimization of product system. Comput. (35) Aerne, E., Personal communication with E. Aerne, Försterbüro Chem. Eng. 1995, 19, 229−234. [Forestry Office] Nesslau. Waldregion 5 Toggenburg. In Nesslau, (14) Stefanis, S.; Livingston, A.; Pistikopoulos, E., Minimizing the Switzerland, 2009. environmental impact of process plants: a process systems method- (36)Janowiak,M.K.;Webster,C.R.Promotingecological ology. Comput. Chem. Eng. 1995, 19. sustainability in woody biomass harvesting. J. For. 2010, 108 (1), 16−23. (15) Stefanis, S. K.; Livingston, a. G.; Pistikopoulos, E. N. (37) Martínez-Blanco, J.; Lazcano, C.; Christensen, T. H.; Muñoz, P.; Environmental impact considerations in the optimal design and Rieradevall, J.; Møller, J.; Anton,́ A.; Boldrin, A. Compost benefits for scheduling of batch processes. Comput. Chem. Eng. 1997, 21, 1073− agriculture evaluated by life cycle assessment. A review. Agron. 1094. Sustainable Dev. 2013, 33 (4), 721−732. (16) Azapagic, A. Life cycle assessment and its application to process (38) Saner, D.; Juraske, R.; Kübert, M.; Blum, P.; Hellweg, S.; Bayer, P. selection, design and optimization. Chem. Eng. J. 1999, 73,1−21. Is it only CO2 that matters? A life cycle perspective on shallow

7658 dx.doi.org/10.1021/es500151q | Environ. Sci. Technol. 2014, 48, 7651−7659 Environmental Science & Technology Article geothermal systems. Renewable Sustainable Energy Rev. 2010, 14 (7), 1798−1813. (39) ecoinvent Centre ecoinvent data v2.2. http://www.ecoinvent. org/database/ (accessed 16 Dec, 2013). (40) Baumann, H.; Tillman, A.-M. The Hitch Hiker’s Guide to LCA.; Studentlitteratur AB: Lund (SE), 2004. (41) General Algebraic Modeling System (GAMS), 23.9;GAMS Development Corporation: Washington, DC, 2012. (42) Ehrgott, M. Multicriteria Optimization. 2nd ed.; Springer: Berlin/ Heidelberg, Germany, 2005; p 323. (43) Azapagic, A. Life cycle assessment and multiobjective optimization. J. Clean. Prod. 1999, 7, 135−143. (44) Heeren, N.; Jakob, M.; Martius, G.; Gross, N.; Wallbaum, H. A component based bottom-up building stock model for comprehensive environmental impact assessment and target control. Renewable Sustainable Energy Rev. 2013, 20,45−56. (45) Humbert, S.; Manneh, R.; Shaked, S.; Wannaz, C.; Horvath, A.; Deschenes,̂ L.; Jolliet, O.; Margni, M. Assessing regional intake fractions in North America. Sci. Total Environ. 2009, 407 (17), 4812−4820. ■ NOTE ADDED AFTER ASAP PUBLICATION This paper was published ASAP on June 13, 2014. Due to production error, not all of the author’s corrections were incorporated. The corrected version was reposted on June 17, 2014.

7659 dx.doi.org/10.1021/es500151q | Environ. Sci. Technol. 2014, 48, 7651−7659