Impact of different hydrological models on hydroelectric operation planning

The flow predicted through hydrological models becomes the main input for hydroelectric generation. Several categories of hydrological models have been implemented across the globe in order to achieve regional cleaner production goals. In this regard, our study analyzes the impact of employing three different conceptual hydrological models for streamflows forecasting in planning the operation of a hydroelectric plant (HPP). Génie Rural à 4 Paramètres Journalier (GR4J), Hydrologiska Byrans Vattenbalansavdelning (HBV), and Sacramento Soil Moisture Accounting (SAC) are the models chosen for our study. GR4J, HBV, and SAC employ four, fourteen, and sixteen parameters for calibration, respectively. The applied methodology consists of four steps. Firstly, a hydrological representation is set up for the chosen basin. Secondly, scenarios are identified, including typical and atypical periods obtained from historical rainfall and flow data for the study area, with the aim of defining calibration and validation periods. Thirdly, the forecast models are calibrated using the Shuffled Complex Evolution-University of Arizona (SCE-UA) optimization algorithm and subsequently validated. Simulated streamflows are evaluated using four performance indicators: Nash Sutcliffe Coefficient (Nash), Nash Coefficient for Logarithm Values (Nash-ln), Relative Root Mean Square Error (RRMSE), and Pearson Correlation Coefficient (R). Finally, a hydropower representation is developed for the study object. The hydrological-hydropower representation is simulated for the chosen scenarios and the impact of the GR4J, HBV, and SAC models on hydrological, hydroelectric, and financial results are evaluated. For each scenario, the level of water storage in the reservoir and the energy produced by the plant are analyzed. The methodology is applied to the Três Marias HPP, in the upper São Francisco River basin (USFB), Brazil. The simulations obtained by the HBV and SAC models present the closest daily forecasts and the smallest variations of water storage level and energy in relation to the reference. However, the results of the GR4J model tend to overestimate the daily streamflow forecasts and the variations in reservoir level and generated energy.