ACCEPTED VERSION Wu, Wenyan; Maier, Holger R.; Simpson, Angus Ross Single-objective versus multiobjective optimization of water distribution systems accounting for greenhouse gas emissions by carbon pricing Journal of Water Resources Planning and Management, 2010; 136(5):555-565 © 2010 ASCE PERMISSIONS http://www.asce.org/Content.aspx?id=29734 Authors may post the final draft of their work on open, unrestricted Internet sites or deposit it in an institutional repository when the draft contains a link to the bibliographic record of the published version in the ASCE Civil Engineering Database. "Final draft" means the version submitted to ASCE after peer review and prior to copyediting or other ASCE production activities; it does not include the copyedited version, the page proof, or a PDF of the published version 28 March 2014 http://hdl.handle.net/2440/62766 Journal of Water Resources Planning and Management. Submitted March 27, 2009; accepted December 16, 2009; posted ahead of print December 18, 2009. doi:10.1061/(ASCE)WR.1943-5452.0000072 Single-Objective versus Multi-Objective Optimization of Water Distribution Systems Accounting for Greenhouse Gas Emissions by Carbon Pricing Wenyan Wu School of Civil, Environmental and Mining Engineering, University of Adelaitde, p Adelaide, 5005, Australia i [email protected] r c Holger R. Maier School of Civil, Environmental and Mining Engineering, Unisversity of Adelaide, Adelaide, 5005, Australia u [email protected] Angus R. Simpson a d School of Civil, Environmental and Mining Engineering, University of Adelaide, M e Adelaide, 5005, Australia t asimpson@ civeng.adelaide.edu.au i d d e e Abstract: Previous research has demonstrated that there are significant trade-offs t y p between the competing objectives of minimizing costs and Greenhouse Gas (GHG) p e emissions for water distribution system (WDS) optimization. However, upon introduction o c of an emissions trading scheme,C GHG emissions are likely to be priced at a particular c level.A Thus, a monetary v alue can be assigned to GHG emissions, enabling a single- t objective optimizatioon approach to be used. This raises the question of whether the N introduction of carbon pricing under an emissions trading scheme will make the use of a multi-objective optimization approach obsolete or whether such an approach can provide additional insights that are useful in a decision-making context. In this paper, the above questions are explored via two case studies. The optimization results obtained for the two case studies using both single-objective and multi-objective approaches are analyzed. The 1 Copyright 2009 by the American Society of Civil Engineers Journal of Water Resources Planning and Management. Submitted March 27, 2009; accepted December 16, 2009; posted ahead of print December 18, 2009. doi:10.1061/(ASCE)WR.1943-5452.0000072 analyses show that the single-objective approach results in a loss of trade-off information between the two objectives. In contrast, the multi-objective approach provides decision makers with more insight into the trade-offs between the two objectives. As a result, a multi-objective approach is recommended for the optimization of WDSs accountting for p GHG emissions when considering carbon pricing. i r c Keywords: Water distribution systems; Multi-objective soptimization; Genetic u algorithms; Greenhouse gas emissions; Sustainability; Carbon dioxide. n a d Introduction M e t i Climate change, especially global wdarming caused by hduman activities, presents serious e global risks. Mitigating global warming by reducinge greenhouse gas (GHG) emissions is t y a unique challenge facping our generation. In order to tackle this challenge, many p measures, includineg emissions/carbon trading schemes, are being introduced. An o c emissions trading scheme can be implemented in many ways, amongst which, a cap and C c trade approach is a popular method. Under a cap and trade scheme, emitters of GHGs A t need to acquire a permit for every tonne of GHG they emit. These permits can be bought o and sold on a maNrket. Some businesses may need to buy permits to cover the GHGs they emit; while others may be able to sell any excess permits they own, if they can reduce their emissions by employing advanced technology, for example. As a result, many industries, including the water industry, will be affected by the price of carbon and the amount of GHGs they emit. This leads to a need to incorporate GHG emission 2 Copyright 2009 by the American Society of Civil Engineers Journal of Water Resources Planning and Management. Submitted March 27, 2009; accepted December 16, 2009; posted ahead of print December 18, 2009. doi:10.1061/(ASCE)WR.1943-5452.0000072 considerations into the optimal design and operation of water distribution systems (WDSs). GHG related issues, such as energy consumption, have been investigated in manyt studies p in WDS research. In the area of optimization, Sarbu and Borza (1998) investigated i r various solutions to increasing the power efficiency of pumping systems. Baran et al. c (2005), Lopez-Ibáñez et al. (2005) and Ulanicki et al. (2007) optismized the scheduling of u pumps to reduce electricity costs. In the planning and management area, Lundie et al. n (2004) developed a life cycle assessment approach for metropolitan water system a d planning, in which energy use and direct gaseous emissions are identified as two of the M e important environmental indicators of a sustainable metropolitan water systems. Filion et t i al. (2004) also employed a life cycled approach to quantifyd energy expenditures of pipes in e a WDS. More recently, Filion (2008) explored the econnections between the urban form t y and energy use of watepr distribution networks. In a study carried out by Dandy et al. p (2006), GHG emisesions resulting from pipe manufacturing were evaluated for a WDS. o c Following the Dandy study, Wu et al. (2010) considered the impact of GHG emissions on C c the optimal design of WDSs explicitly, by incorporating the minimization of life cycle A t GHG emissions, together with the minimization of system costs, into the optimal design o of WDSs via a mNulti-objective approach. It is now becoming increasingly common for carbon related emissions to be priced under an emissions trading scheme, yet the impact of carbon pricing on the optimal design and operation of WDSs has not been investigated thus far. 3 Copyright 2009 by the American Society of Civil Engineers Journal of Water Resources Planning and Management. Submitted March 27, 2009; accepted December 16, 2009; posted ahead of print December 18, 2009. doi:10.1061/(ASCE)WR.1943-5452.0000072 The present study aims to consider the inclusion of carbon pricing into both single- objective and multi-objective optimization approaches for WDS optimization. Wu et al. (2010) demonstrated that there are significant trade-offs between the competing objectives of minimizing costs and GHG emissions. However, upon introductiotn of an p emissions trading scheme with a cap and trade approach, a monetary value (referred to as i r the carbon price in this paper) is usually assigned to GHG emissions. This monetary c value of the carbon price can be determined by either evaluation smethods, as done by the u International Panel on Climate Change (IPCC), or a carbon market. The expression of n GHG emissions in monetary terms enables a single-objective optimization approach to be a d used. This raises the question of whether the introduction of carbon pricing under a M e possible emissions trading scheme will make use of a multi-objective optimization t i approach obsolete or whether such dan approach can prodvide additional insights that are e useful in a decision-making context. In this paper, tweo case studies were used to compare t y single and multi-objectipve approaches when considering both cost and carbon emission p objectives. Based oen the results obtained for the case studies, recommendations regarding o c the optimization of WDSs under a carbon pricing regime as determined by an emissions C c trading scheme are presented. A t o The remainder oNf the paper is organized as follows. The methods used to solve the proposed WDS optimization problem, including evaluation of the objective functions, the optimization approach adopted, carbon pricing and present value analysis, are introduced in the next section. Next, the two case studies are introduced, to which both single- objective and multi-objective optimization approaches are applied. Thereafter, the 4 Copyright 2009 by the American Society of Civil Engineers Journal of Water Resources Planning and Management. Submitted March 27, 2009; accepted December 16, 2009; posted ahead of print December 18, 2009. doi:10.1061/(ASCE)WR.1943-5452.0000072 optimization results obtained using the two approaches are presented and discussed. Finally, conclusions and recommendations are presented. Methods t p i r Objective Function Evaluation c s u The WDS optimization problem investigated in this paper is a multi-objective n optimization problem that accounts for two objectives: the minimization of system costs a d and the minimization of GHG emissions (vMia a price for carbon). When the single- e objective optimization approach is used, the total cost, which is the sum of the system t i d costs and the GHG costs expressed in terms of dollards for the cost of carbon related e e emissions, is minimized as the sole objective. In contrast, in the multi-objective approach, t y the system and GHG cospts are minimized as two separate objectives. p e o c Figure 1 shows the objective funCction evaluation process. The system cost considered in c this study is defined as the sum of the capital costs, operating costs for pumping and A t pump replacement/reofurbishment costs at regular intervals during the service or design life of the systemN. The capital cost is incurred due to the purchase and installation of network components (pipes and pumps) and construction of pump stations. This cost occurs at the beginning of a project. As the design life of a WDS is much longer than the service life of pumps, then pumps and electrical control equipment need to be replaced or refurbished periodically to ensure the performance of the system is maintained. In the 5 Copyright 2009 by the American Society of Civil Engineers Journal of Water Resources Planning and Management. Submitted March 27, 2009; accepted December 16, 2009; posted ahead of print December 18, 2009. doi:10.1061/(ASCE)WR.1943-5452.0000072 case studies in this paper, a 100-year pipe network service life and a 20-year pump service life are assumed. The operating cost is incurred mainly due to the system operation of pumping. The computation of the annual operating cost is taken as the annual energy consumption multiplied by an average electricity tariff. A motor effticiency p of 95% is assumed for each pump. In practice, electricity tariffs may vary across regions i r and with time. In this study, an electricity cost of 0.143 dollars per kWh is assumed, c which is an approximate average electricity tariff taking into accsount peak and off-peak u tariffs. As both pump replacement/refurbishment costs and operating electricity costs n occur during the life of the system, calculation of these two costs requires present value a d analysis. M e t i In calculating the annual energy cdonsumption, a 48-hdour extended period simulation e (EPS) has been used in the simulation model to eaccount for the diurnal variation in t y demand, the fluctuation pin tank water levels and the variation of the pump operating point p during the day, to perovide a realistic estimate of the operational behavior of the system. o c In the EPS, a diurnal demand curve presented in Figure 2 applied to the average flow C c during a year or the average-day flow (Water Services Association of Australia 2002) is A t used to estimate the average energy consumption of the system due to pumping during o the design perioNd (100 years). In addition, an average flow on the peak day is used to design the distribution systems upstream of the balancing storage tanks, as suggested by Water Services Association of Australia (2002). The average flow on the peak day is computed by multiplying the average-day flow by the Peak Day Factor (PDF). In this paper, a PDF of 1.5 obtained from the Water Services Association of Australia (2002) is 6 Copyright 2009 by the American Society of Civil Engineers Journal of Water Resources Planning and Management. Submitted March 27, 2009; accepted December 16, 2009; posted ahead of print December 18, 2009. doi:10.1061/(ASCE)WR.1943-5452.0000072 used. It should be noted that in designing distribution systems downstream of the balancing storage tanks, the average flow on the peak hour and fire loading cases would also be required to ensure an adequate design. In both case studies, an average pipe roughness value of ε=0.25mm was assumed for the first 50-year period and a vtalue of p ε=1.5mm for the second 50-year period in order to account for pipe aging. i r c GHG emission costs are obtained by multiplying the carbon psrice by the total GHG u emissions of the system. The total GHG emissions considered in this study consist of n capital emissions and operating emissions. Capital emissions are due to the manufacture a d and installation of network components, such as pipes, pumps, valves and tanks. In this M e study, pipes are the only source of capital emissions considered, as they represent the t i largest proportion of the impact (dFilion et al. 2004).d These emissions occur at the e beginning of a project. Similarly to the operating coest, operating GHG emissions are due t y to electricity consumptipon related to the operation of the system over time in regions p where it is assumeed that fossil fuels are used for electricity generation. Operating o c emissions occur over time during the service life of the system. Therefore, the estimation C c of total operating emissions over the service life of the network also requires present A t value analysis. o N In addition, in evaluating the capital emission costs, embodied energy analysis (EEA) is first applied to translate the material use of the pipes into an estimate of their embodied energy in MJ. Thereafter, emission factor analysis (EFA) is used to translate embodied energy use into a corresponding estimate of GHG emissions in kg of CO -e (carbon 2 7 Copyright 2009 by the American Society of Civil Engineers Journal of Water Resources Planning and Management. Submitted March 27, 2009; accepted December 16, 2009; posted ahead of print December 18, 2009. doi:10.1061/(ASCE)WR.1943-5452.0000072 dioxide equivalent). In practice, embodied energy values and emission factors may also vary across regions and with time, depending on the material excavation and extraction methods used and the makeup of electricity energy sources (for example, thermal, nuclear, wind, hydroelectric, etc.). In this study, a typical embodied energy of ducttile iron p cement mortar lined (DICL) pipes of 40.2 MJ/kg and a typical emission factor of 1.042 i r kg CO -e per kWh are used. The embodied energy value of DICL pipes has been 2 c obtained from Ambrose et al. (2002), and the emission factor selsected is a full fuel cycle u emission factor for end electricity users in South Australia (Australian Greenhouse Office n 2006). While the embodied energy and emission factor values are realistic estimates, and a d adequate for the purpose of this paper, they are likely to change with time in actual M e applications due to changes in the way electricity is being generated as governments t i respond to the threat of climate chdange (e.g. an increadse in wind power generation to e replace production from coal-fired power stations). e t y p p Optimization Approeach o c C c In this paper, a multi-objective genetic algorithm (GA) is used, as GAs have been shown A t to be effective for WDS optimization problems (Simpson et al. 1994). GAs are a global o optimization metNhod that belong to the class of evolutionary algorithms (Goldberg 1989). GAs differ from traditional optimization techniques in that the concept of GAs is inspired by natural phenomena of heredity. GAs use the “principle of survival of the fittest” to select more suitable trial solutions by dealing with a population of solutions simultaneously. Each solution is represented by a binary, integer or real valued string 8 Copyright 2009 by the American Society of Civil Engineers Journal of Water Resources Planning and Management. Submitted March 27, 2009; accepted December 16, 2009; posted ahead of print December 18, 2009. doi:10.1061/(ASCE)WR.1943-5452.0000072 called a chromosome. By applying three genetic operators: selection, crossover and mutation to the chromosomes, GAs maintain good solutions in the current population of solutions and explore the search space for better solutions. The search process terminates when the stopping criteria are met. t p i r Traditionally, GAs have generally been applied to optimization problems that have one c objective. However, most problems in the real world have mores than one objective that u needs to be satisfied. Therefore, a number of multi-objective genetic algorithms, n including the Vector Evaluated Genetic Algorithm by Schaffer (1984), the Weight-Based a d Genetic Algorithm by Hajela and Lin (1993), the Multi-Objective Genetic Algorithm by M e Fonseca and Fleming (1993) and the Strength Pareto Evolutionary Algorithm by Zitzler t i and Thiele (1998) have been develdoped to solve real dworld multi-objective problems e (Deb 2002). In this study, a multi-objective geneteic algorithm called WSMGA (Water t y System Multi-objectivep Genetic Algorithm) has been used to solve both the single- p objective and multie-objective problems presented in this paper. WSMGA is based on the o c state-of-the-art multi-objective generic algorithm NSGA-II (Deb et al. 2002) and is C c described in more detail in Wu et al. (2010). A t o Carbon Pricing N Emissions trading is one of the most popular schemes for controlling GHG emissions. In most emissions trading schemes, a cap and trade approach is used. Under a cap and trade approach, emission permits are usually issued by the government. Businesses must have 9 Copyright 2009 by the American Society of Civil Engineers