ASSESSING HYDROPOWER VULNERABILITY UNDER UNCERTAIN CLIMATE CHANGE IN DEVELOPING COUNTRIES: CASE OF ECUADOR

摘要

OverviewClimate change will impact natural and human systems, specially energy systems and the technologies that are mostdependant or exposed to climate variations (DOE, 2015). Hydropower is one of the most vulnerable systems toclimate change due to its dependence on the hydrological cycle in which temperature and precipitation are dominantdrivers. This leads to energy security concerns in countries that are highly dependant on this technology, specially indeveloping regions such as Africa, Asia and South America where there has been a recent upsurge in hydropowerinfrastructure (WEC, 2015). Energy system models that usually consider deterministic and constant hydro-climaticconditions collide with climate change uncertainty and the myriad of possible scenarios that result from differentclimate modelling exercises. This underlines the challenge of incorporating impacts of hydro-climatic change intothese models due to its inherent uncertainty (Shadman, Sadeghipour, Moghavvemi, & Saidur, 2016). This researchintends to present a framework for generation capacity planning that includes climate change uncertainty andtherefore delivers a power generation portfolio that is more robust or adapted to climate change variations. TheRepublic of Ecuador will be used as a case study to apply the framework, since hydropower currently dominates 46per cent of total electricity generation (2014) and there are ambitious investment plans to reach more than 90 percent of hydropower participation in the grid by 2017 (MICSE, 2015). The location of Ecuador also posits interest,i.e. the tropical Andes, which is one of the areas where impacts of climate change are still highly uncertain. Theresults presented in this extended abstract are initial and part of a wider doctoral research project, which will developa TIMES energy system model for Ecuador’s power sector and later on use a mean-variance portfolio theoryapproach to handle uncertainty.MethodsThe proposed methodology consists in a three-step approach, which is depicted in Figure 1. The first step, ‘HydroclimateAnalysis’, uses a statistical hydrological model to assess how inflow into hydropower stations will varyaccording to projections of a large General Circulation Model (GCM) ensemble, namely the Coupled ModelIntercomparison Project 5 (CMIP5), which was used in the latest Assessment Report of the Intergovernmental PanelClimate Change (IPCC-AR5) (Taylor, Stouffer, & Meehl, 2012). This statistical model will be ‘signalled’ by thechanges in temperature and precipitation of the CMIP5. The results of this step are a set of monthly inflow timeseries into Ecuador’s largest hydropower plants. The second step, ‘Energy Modelling’, consists on building aTIMES energy systems optimisation model at a plant-by-plant detail level that represents the current and futurepower system of Ecuador up to 2050 (Loulou & Labriet, 2008). The climate change signalled time series of inflowfrom step one will be used to characterize the availability of hydropower and assess changes in total systemgeneration cost, emissions, etc., according to different generation portfolio options. The optimisation approach of themodel and the time-resolution at a monthly level of hydropower availability will bring insights compared to modelsthat consider constant annual availability factors. The third and final step, ‘Uncertainty Analysis’, seeks to includeuncertainty of climate change projections as additional criteria for assessing the least-cost generation portfoliosobtained in step two. The range of inflow due to different climate change projections from the CMIP5 will be usedas a proxy to parameterise the scenario probability space for hydropower output. This information, additional to thehistoric volatility of fossil fuel prices, will be used as inputs for a mean-variance portfolio theory approach appliedto the power sector as has been initially done by (Awerbuch & Berger, 2003). The outcome of this step is acomparison of different generation portfolio options according not only to their least-cost but to their cost-risk, thisbeing defined as the standard deviation of the portfolio generation cost.ResultsThis first step of the proposed methodology has been applied to assess the inflow to Ecuador’s largest hydropowerstation: Paute-Molino, an 1100MW unregulated run-of-river facility in the southeast of the country. Results for 39GCM models for RCP8.5 and averaged for the last 30years of the century (2071-2100) have been plotted in Figure 2(left panel) and show that the mean of the CMIP5 ensemble (in red) projects an annual increase of inflow of 16%compared to the historic value (in solid black), however mostly during the wet season. The historic standarddeviation is plotted (in black dashed lines) and is much narrower than the projected standard deviation of the CMIP5ensemble (shaded area), meaning that so far there is a wider uncertainty about what the projected inflows could looklike by the end of the century. The individual results of 39 GCM have been also plotted (in light grey dotted lines) and show the discrepancies among the models, some accounting for as much as a 300 per cent increase while othersprojecting a decrease in 80 per cent, as can be seen in Figure 2 (right panel) where projected inflows for ensemblemean and individual GCM have been normalised compared to the historic baseline.Figure 1 Proposed methodology for assessing the impact of uncertain climate change on hydropower generation.Figure 2 Inflow regime for the baseline, each GCM (39) and the ensemble mean projection in the Paute HydropowerStation for the RCP8.6 and for 2071-2100. The left panel shows absolute values and the right panel percentagechanges from the baseline. Shaded area represents the ±1 Standard Deviation of the GCM CMIP5 ensemble.ConclusionsWorking with annual mean availability factors for hydropower in an energy system model to assess a power systemwith large shares of hydropower can mask monthly variations in inflow that can impact the system costsignificantly. Also, working only with the ensemble mean (in red in Figure 1) can mask divergent results fromdifferent GCM projections that are also likely to occur thus denoting the need to give a statistical meaning of theresults obtained from climate modelling exercises. This underlines the importance of including the abundant resultsobtained by climate modellers into the energy system modelling process to be able to assess how climate change canimpact energy systems and define generation portfolios that have lower cost-risk thus being better adapted forclimate change.