Real-time model predictive control study of run-of-river hydropower plants with data-driven and physics-based coupled model

Abstract

Real-time control of hydroelectric generating units and floodgates is the key to ensuring the economy and safety of hydroelectric production. Runoff hydropower plants require high real-time control accuracy because of the small regulation interval. The existing real-time control methods have certain shortcomings such as frequent load and gate adjustments and a wide range of water level fluctuations. Therefore, we propose a data-driven and physics-based coupled real-time model predictive control (MPC) method with the aim to develop a highly accurate control strategy. Compared to the traditional MPC model, the model proposed in this study presents three major improvements: (1) the data-driven and physics-based coupled model enhances the accuracy of water level prediction; (2) the corresponding decision framework is developed for the actual operation of the power station to reduce the decision space; and (3) a feedback correction mechanism that can convert the water level prediction error into flow error, which reflects the time-varying nature of the system. We demonstrate that the coupled prediction model combined with feedback correction improves the water level prediction accuracy by 60% on average. Moreover, the reduced decision space enables the optimization algorithm to reduce the optimization time by 83% on average and achieve a globally optimal solution in the decision space. The proposed data-driven and physics-based coupled real-time MPC method is thus useful for the real-time regulation of run-of-river power plants.