Development of Upper Euphrates Basin hydro-economic model and hydropower generation optimization

AKL

Asagi Kalekoy Dam

BGT

Bagistas Dam

BYH

Beyhan Dam

CALVIN

California Value Integrated Network

DAP

Eastern Anatolia Project

FEHEM

Upper Euphrates Basin Hydro-economic Model

GAP

Southeastern Anatolia Project

GWh

Gigawatt hour

KBN

Keban Dam

KGI

Kigi Dam

MW

Megawatt

MWh

Megawatt hour

M$

Million dollar (revenue)

OZL

Ozluce Dam

PMB

Pembelik Dam

SPI

Standardized Precipitation Index

SYR

Seyrantepe Dam

TTR

Tatar Dam

UZN

Uzuncayir Dam

Global water withdrawals have been increasing rapidly in recent years to sustain rising living standards with ever-increasing energy and food demands (Kummu et al. 2010; Mekonnen & Hoekstra 2016; Liu et al. 2017). Many basins around the world have experienced widespread water scarcity conditions and associated water management challenges as a manifestation of this situation (Wada et al. 2013; Kahil et al. 2015; Veldkamp et al. 2017). These challenges are expected to become even more critical in the coming years (Hanasaki et al. 2013; Kim et al. 2016). In addition, the discharge of wastewater degrades water quality, resulting in water that cannot be used for potable water and industrial applications (Panagopoulos & Giannika 2022a, 2022b, 2023). Therefore, policymakers in vulnerable basins need to be able to meet the demands of different sectors. To do so, they need to anticipate how to adapt their management practices to secure future water supply. However, the choice of water management options is often associated with trade-offs across space and time between multiple water-related systems such as food production, energy supply and ecosystem services (Banzhaf 2009; Hurford et al. 2014). Appropriate consideration of all these options requires the development of a systematic approach that describes the biophysical and socioeconomic factors that determine the future dynamics of river basins, including the fundamental interactions between water, energy and agricultural systems (Rogers & Fiering 1986; Brown et al. 2015; Wada et al. 2017).

Hydro-economic models combine water resource engineering and economics, where water allocation is governed by the economic value of water, while operating costs and hydrology influence water allocations (Cai 2008; Harou et al. 2009a, 2009b; Booker et al. 2012). Hydro-economic models are a mathematical expression of the water demand and allocation relationships of different water-using sectors (e.g. agriculture, industry, municipalities and hydropower generation) and the hydrological relationships in the water system (Bekchanov et al. 2015). Hydro-economic models are usually built around a river basin, because there is a strong relationship between water-dependent production and watersheds in terms of environmental systems (Keller & Keller 1995; Keller et al. 1996; Ringler et al. 2004).

Hydropower is an important renewable energy source in the Euphrates basin. Hydropower plants in the Upper Euphrates Basin Hydro-economic Model (FEHEM) are from large water storage or run-of-river plants. Hydropower plants with storage are advantageous because they add flexibility to water resources operations. This is because these plants store water at times when the demand for water is low and the supply is high. At times of high demand for water, stored water is released and power is generated to meet water and energy demands. Due to these characteristics, hydropower plants with large storage capacities can better adapt to changing climatic conditions with their operational flexibility. Since unit energy prices are key economic factors for hydropower decisions, hydro-economic models are suitable tools for representing hydropower planning and management decisions.

Despite hydro-economic models' importance in more efficient water allocation and hydropower decision-making, this study is the first attempt to develop an integrated hydro-economic optimization model for the Upper Euphrates Basin with 10 hydropower plants. The aim of this study is to analyze the Upper Euphrates Basin water supply system with different management and decision-making strategies and to contribute to management solutions with engineering alternatives. For this purpose, it has developed an efficient and flexible hydropower operation and planning model using state-of-the-art optimization modeling techniques. The FEHEM is a hydro-economic optimization model for the interconnected water supply system of the Upper Euphrates Basin. The FEHEM represents approximately 8.1% of Turkey's total drainage area. Using 45 years of historical hydrological data to represent hydrologic variability, the model determines the optimum hydropower generation decisions of the modeled reservoirs. Using the developed model, it will be possible to prepare water management plans according to different policies and future climate scenarios to determine adaptation strategies in advance.

The paper is organized as follows: Section 2 describes model development, including governing equations for optimizations, modeled reservoirs and their characteristics. Section 3 presents turbine flow, reservoir storage, and hydropower generation and revenue results. Section 4 presents a discussion of the paper and Section 5 provides concluding remarks.

Hydrological and economic data are used in hydro-economic optimization models. Generally, hydrological datasets consist of reservoir and tributary flows, and evaporation rate and infiltration and gains due to groundwater interaction. In addition, reservoir, canal and water treatment plant capacities, hydropower turbine capacity and agricultural, urban, industrial and, if available, environmental water demand constitute the model inputs. Economic data consist of sectoral supply-demand functions, unit operating costs and unit energy prices (Dogan 2015). The main model outputs are reservoir water storage and diversions, agricultural and urban water allocations, and energy production.

The FEHEM is a deterministic hydro-economic optimization model of the hydropower reservoir system of the Upper Euphrates Basin. The FEHEM is an adaptation of the California Value Integrated Network (CALVIN) model to the Upper Euphrates Basin. CALVIN is a hydro and economic optimization model for California's interconnected water supply system (Draper et al. 2003).

The FEHEM supports the quantitative understanding of the integrated water distribution and economic system of Turkey's largest installed capacity power dams in the Upper Euphrates Basin. Initially, 10 selected large surface reservoirs in the Upper Euphrates Basin are modeled with the FEHEM. With further development, the model can represent all groundwater reservoirs of the basin, mega agricultural irrigation projects such as Southeastern Anatolia Project (GAP) and Eastern Anatolia Project (DAP), water demand areas and urban needs.

With the FEHEM, a database consisting of 45 years of historical hydrological and meteorological data of the basin was first created. Following the creation of the database, the network flow model was developed using Pyomo, a high-level optimization modeling language in Python (Hart et al. 2017).

Hydropower plants operate depending on the availability of water. The reservoir in the flows is therefore important for modeling. In the deterministic case, streamflow observations upstream of rim reservoirs (called rim inflows) are used directly as model inputs. The reservoir inflows represent hydrological variability over a modeling period. Some simulation models use historical hydrology to represent this variability, while others use hydrological predictions. Planning models such as the FEHEM use long-term historical hydrology and operational models usually use short-term future forecasts. The model can be run at various time step lengths – hourly, daily, weekly, monthly and yearly, depending on data availability – and historical hydrology or future forecasts can be used as reservoir inputs. Table 1 shows the modeled hydropower plants and their characteristics such as storage and release capacity and dead volume storage. The value for overall efficiency η of 0.90 is assumed for all plants.

Using the polynomial parameters θ, α, β, γ and c, the drop height H is calculated as a function of storage. The polynomial parameters for each reservoir are shown in Figure 3.

Hydropower generation is modeled with penalty curves in the FEHEM, but power capacity use, energy production and revenue are calculated in a separate post-processor. A hydropower processor takes storage and release data from the FEHEM output file and provides time-series data for each FEHEM plant, such as power capacity, total monthly and annual energy production and revenue, amount of water spilled and total turbine capacity used.

Network flow optimization models allocate water that is transported in the network and are commonly used in transportation, transmission and water resources modeling (Draper et al. 2003, 2004; Bazaraa et al. 2010; Dogan et al. 2018). A typical hydropower network contains nodes and links, where nodes indicate power plants and links indicate streams, channels or pipelines. The overall representation of the hydropower network flow representation includes the objective function to be minimized (Equation (2)).

In Equation (2), z represents the cumulative cost. For each link, the index i represents the start node and the index j represents the end node. In the piecewise linear programming technique, k represents each linear segment. These parts are the result of the linearization of nonlinear functions. c represents the linear unit cost. In this equation, the independent variable is X and represents the flow from node i to node j. All functions used in the model must be concave in maximization problems or convex in minimization problems to guarantee the global optimum outcome (maximum or minimum). Equation (2) can be explained simply as minimizing the total cost by multiplying the unit cost by the flow rate.

Equation (2) is subject to the mathematical function of three constraints. The first limiter represents the lower current limit (l) for each connection (Equation (3)). The lower limit value is zero unless otherwise specified. This lower limit can also be used to represent minimum flow requirements for environmental purposes. The second limiting function (Equation (4)) represents the upper flow limit (u) and its value is plus infinity unless otherwise specified. This function can also be used to represent the flow-carrying capacity for reservoirs, canals and turbines. The last limiter (Equation (5)) represents the mass balance. For each connection, the incoming flow must be equal to the outgoing flow. In this function, a is used to represent loss factors such as evaporation. In all functions, the independent variable is X, which represents the flow carried in the links. All other parameters (c, l, u and a) are fixed. These four parameters are predefined for all links in the system. The default value of c and the unit benefit (or cost) are zero. However, this c value can take a positive or negative value in connections that require benefit or cost.

The objective of the FEHEM is to minimize costs and maximize benefits. All these objectives are expressed mathematically in Equation (2). Equations (2), (3), and (5) are defined in the format that Pyomo uses. The Pyomo model solves the optimization problem through preloaded solvers (such as GLPK), and the results are organized and analyzed as time series through postprocessors.

Reservoir storage increases when release is less than the reservoir inflow in the wet season and storage decreases in dry seasons with greater releases to generate energy. With monthly varying energy prices, the FEHEM tends to store more water in months when the unit price of energy is low. Reservoirs usually approach maximum levels with rainfall and melting snow in winter and spring. It is used for irrigation in summer and dry times to meet urban water demands and to generate hydropower. As a result of this use, the reservoir water level is lowered.

The hydropower facilities in the Upper Euphrates Basin are located on the main tributaries of the Euphrates River. The system is complex as it consists of reservoirs with large sizes and different characteristics. The annual average results obtained from the modeled dams are summarized in Table 2. The largest plant, KBN, generates an annual average energy of 5,239 Gigawatt hour (GWh) energy/year, with a revenue of 343 million$ per year, utilizing a turbine capacity of about 55%. Overall, the FEHEM plants generate about 9,481 GWh of energy with an average turbine capacity use of 36% and obtain a revenue of 620 million $ per year.

Water resources system models use different time scales, ranging from hourly to annual. To maximize revenue, the models assume that a hydropower plant with reservoir storage preferably allocates hydropower releases to peak price times when energy demand and prices are highest. The developed model presents the combination of hydrological differences and electricity price changes in the operations of the long-term FEHEM using a monthly time step over a 45-year hydrological period.

According to the model outputs, the lowest storage was realized in energy stages with series-connected reservoirs. The model shows such a trend because it focuses on energy generation in a series of successive dams. With monthly variable prices, the FEHEM tends to store more water in months when the unit price of energy is low. The model generally chooses to store water as there is no economic benefit to spill. The FEHEM stored an average of 23.43 × 10⁶ hm³ of water per year in reservoir storage operations. In some years, this value reaches 27.13 × 10⁶ hm³, while there are also years when it drops as low as 12.33 × 10⁶ hm³. Spills are penalized to minimize the energy cost in the model. The model spills only if a plant's reservoir capacity is full and inflows exceed the turbine release capacity. When the FEHEM was operated for a period of 45 years, according to the model outputs, there was no spill from the Keban reservoir. However, in reality, the Keban reservoir spilled three times in 1985, 2004 and 2019. Therefore, the model gains by converting spill, which is an economic loss, into production with its optimized decisions.

Limitations are inherent to all models. Piecewise linearization of nonlinear hydropower curves sacrifices some accuracy. Moreover, the model uses perfect hydrologic foresight, knows all hydrologic events, such as floods and droughts, and prepares for them in advance. Perfect foresight results in somewhat optimistic storage and release decisions.

Hydro-economic models provide engineering solutions to large-scale water management planning and policy problems. Models simulate likely scenarios to help decision-makers and stakeholders. They also offer an optimization opportunity to increase the benefit under various operating conditions. For this purpose, 10 large dams in the Upper Euphrates Basin with an installed capacity of about 3,255 Megawatt (MW) were modeled with a hydro-economic optimization model for hydropower operations, called the FEHEM. The current model uses 45 years of monthly historical data between 1971 and 2016 to represent hydrologic variability. The modeled plants generated about 9,481 GWh of energy and contributed to the economy with a revenue of 620 million $ per year. In further development, the FEHEM will be expanded to include more reservoirs. As more hydrologic observations become available, its hydrological time horizon will be extended. In addition, agricultural and urban demands and groundwater basins can be added for a better-integrated water system representation.

A.A. and M.C.T. analyzed the data and wrote the paper. M.S.D. performed research conceptualization and methodology, supervised the work, and reviewed and edited the manuscript.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.