Improving the Control Performance of an Organic Rankine Cycle System for Waste Heat Recovery from a Heavy-Duty Diesel Engine using a Model-Based Approach Johan Peralez, Paolino Tona, Olivier Lepreux, Antonio Sciarretta, Luce Voise, Pascal Dufour, Madiha Nadri To cite this version: Johan Peralez, Paolino Tona, Olivier Lepreux, Antonio Sciarretta, Luce Voise, et al.. Improving the Control Performance of an Organic Rankine Cycle System for Waste Heat Recovery from a Heavy- Duty Diesel Engine using a Model-Based Approach. 2013 IEEE Conference on Decision and Control (CDC), Dec 2013, Florence, Italy. 7 p. hal-00875469 HAL Id: hal-00875469 https://hal-ifp.archives-ouvertes.fr/hal-00875469 Submitted on 22 Oct 2013 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Improving the Control Performance of an Organic Rankine Cycle System for Waste Heat Recovery from a Heavy-Duty Diesel Engine using a Model-Based Approach Johan Peralez, Paolino Tona, Olivier Lepreux, Antonio Sciarretta, Luc Voise, Pascal Dufour, Madiha Nadri Abstract— In recent years, waste heat recovery in a turbine or an expander to produce mechanical power. (WHR) systems based on Rankine cycles have been Vapor is then cooled by a condenser which transfers heat the focus of intensive research for transport appli- to an external cold sink. cations, as they seem to offer considerable potential for fuel consumption reduction. Because of the highly Most Rankine systems are designed to produce elec- transient conditions they are subject to, control plays tricity via a generator connected to the auxiliary network a fundamental role to enable viability and efficiency and/or an energy storage system, even though, in mobile of those systems. The system considered here is an Organic Rankine applications, the expansion machine can also deliver Cycle (ORC) for recovering waste heat from a heavy- mechanical power directly to the transmission. The main duty diesel engine. For this system, a hierarchical and differences with stationary applications lie in the limited modular control structure has been designed, imple- capacity of the cold sink and in the highly transient mented and validated experimentally on an engine behaviour of the hot source, both depending on driving testbed cell. The paper focuses more particularly on improving conditions. the baseline control strategy using a model-based In this context, an effective control system is essential approach. The improvements come from an extensive to attain satisfactory performance over a broad range system identification campaign allowing model-based tuning of PID controllers and, more particularly, from of operating conditions. This is especially true when a dynamic feedforward term computed from a non- there are few available actuators and sensors, as it of- linear reduced model of the high-pressure part of the ten happens in the automotive industry. Despite that, system. literature on control design for Rankine-based WHR for Experimental results illustrate the enhanced per- mobile application is still very scarce, as can be observed, formance in terms of disturbance rejection. for instance, in the comprehensive overview on Rankine I. INTRODUCTION WHR from internal combustion engines presented in [1]. More than one third of the energy produced by internal The system considered here is an Organic Rankine combustion engines (ICE) is released into the environ- Cycle (ORC) for recovering waste heat from a heavy-duty ment in the form of exhaust gas waste heat. Among the diesel engine. For this system, a hierarchical and modular possible waste heat recovery (WHR) solutions, the Rank- control structure has been designed, implemented and ine cycle has drawn the attention of many car and truck validated experimentally on an engine testbed cell. manufacturers in recent years as it holds an interesting The focus is on improving the baseline control strategy potential for reducing vehicle fuel consumption. using a model-based approach. The improvements come WHR Rankine systems for automotive applications from an extensive system identification campaign allow- apply the same principle used worldwide in industry to ing model-based tuning of PID controllers and, more generate power, by converting heat into work. A pump particularly, from a dynamic feedforward term computed circulates a working fluid in a closed loop where an from a nonlinear reduced model of the high-pressure external heat source supplies heat, via a heat exchanger part of the system. The latter approach has first been (or a series of heat exchangers). Vaporized fluid expands proposed in [2] for a steam Rankine cycle, a different system where the presence of an additional actuator in- J. Peralez, P. Tona, O. Lepreux and A. Sciarretta are with the creases the degrees of freedom and makes model inversion Control, Signal and System Department at IFP Energies Nouvelles, France {johan.peralez, paolino.tona, olivier.lepreux, simpler. antonio.sciarretta}@ifpen.fr Luc Voise is with D2T Powertrain Engineering In this paper, on the one hand, we start studying the [email protected] properties of the model-based control structure and, on P. Dufour and M. Nadri are with the Université de Lyon, the other hand, through a set of experimental results, we F-69622, Lyon, France – Université Lyon 1, Villeurbanne, France – CNRS, UMR 5007, LAGEP, France {dufour, show that it actually works, achieving excellent perfor- nadri}@lagep.univ-lyon1.fr mance. II. STATE OF THE ART Symbols V Volume (m3) 2 Among the few publications of practical interest S Area (m ) ρ Density (kg/m3) discussing automatic control issues of Rankine WHR α Heat transfer coefficient (W/(m2K)) systems for automotive applications, we can cite [3] c Heat capacity (J/(kg K)) and, more recently, [4], both on steam processes for h Specific enthalpy (J/kg) m Mass (kg) spark-ignition engines. [5] provides a complete (manual) m˙ Mass flow rate (kg/s) startup and shut-down procedure for an ethanol-based p Pressure (P a) Rankine cycle system, for heavy-duty applications, un- T Temperature (K) L Normalized zone length (−) derlining the difficulty of controlling the working fluid SH Superheating (K) conditions at the evaporator outlet. N Rotation speed (rpm) On the more general topic of Organic Rankine Cycles Subscripts for waste heat recovery operating with variable heat evap evaporator sources (not necessarily for transport applications), [6], exp expander [7], [8] apply control strategies based on linear models exh exhaust gas f working fluid (LQR, MPC), validated on one operating point. [9] covers w wall dynamic modeling and control of an ORC system with i zone i R234fa as a working fluid over a broader operating range. in inlet out outlet However, hot source variations used for simulation are v saturated vapor much slower than those observed at the exhaust of an l saturated liquid automotive engine, especially in terms of mass flow rate. d desired A somewhat richer literature is available on dynamic Superscript modelling and control of vapor compression cycles, the SP Set Point “reverse” of Rankine cycles ( [10], [11], [12]). In both contexts, there is considerable interest in TABLE I the development of simplified (moving-boundary, MB) Nomenclature models to reproduce the two-phase behavior of heat exchangers without the complexity of the finite-volume approach ( [13], [14]). III. SYSTEM DESCRIPTION The system considered here is illustrated in Fig. 1. It is an Organic Rankine Cycle system for waste heat recovery from a heavy-duty Diesel engine using a turbine for the expansion of the working fluid. A description of the main system variables follows, using the nomenclature given in Tab. I. Fig. 2. ORC system inputs–outputs available (corresponding to the corner points of the ther- modynamic cycle, see Fig. 3). Exhaust gas and cooling fluid conditions (mass flow rate and temperature) can be considered as measured disturbances. The available mea- surements can be used to estimate key output variables such as superheating, subcooling or enthalpy, from fluid thermodynamic properties. Four actuators are available, in principle: • SP The evaporator by-pass setpoint V oevap, controlling Fig. 1. The Organic Rankine Cycle system under investigation the fraction of exhaust gas entering the evaporator. • SP The expander by-pass setpoint V oexp, controlling the fraction of fluid entering the expander. • SP A. Inputs–outputs The pump speed setpoint Npump allowing to control the fluid mass flowm ˙ entering the evaporator. As shown in Fig. 2, measurements of the pressure– evap,in • The turbine speed setpoint N SP . temperature pairs (p, T ) between each component are exp Fig. 4. Moving boundaries layout for the evaporator state depending on the cooling conditions and on the pressure imposed by the tank (separator). Applying the MB approach to the low-pressure (LP) part of the Fig. 3. Rankine cycle for “dry” fluid circuit generally yields a high-order hybrid model (the number of states changes depending on fluid conditions at condenser outlet). However, as explained in Sec. IV- B. Control requirements B, condenser modelling is not necessary for (our) control The main objective of the supervision and control purposes. system is to maximize, in the presence of varying exter- nal conditions, the production of electric energy during B. Pump and turbine vehicle usage (assuming that the load control system can Pump and turbine dynamics being very fast compared always make use of it, or dissipate it, otherwise). to exchangers dynamics, they are modelled by algebraic Several output constraints ensuring system safety must equations.