基于模型预测控制的梯级水电-光伏互补系统实时调度框架

Abstract The integration of photovoltaic (PV) power and hydropower has become an effective approach for facilitating the consumption of renewable energy. Improving the management and control of hydropower and PV power during real-time scheduling can help with meeting the grid's anticipated electricity demand. However, hedging against the inherent uncertainties of PV power output and runoff is a significant challenge. This paper proposes a real-time scheduling framework for cascade hydropower-photovoltaic power (CH-PVP) complementary systems based on model predictive control (MPC). A Wasserstein generative adversarial network (WGAN) is used to forecast the runoff and PV power output. On this basis, a real-time scheduling model for CH-PVP complementary systems is established which considers dynamic water delay time and hydropower unit vibration zones, and aims to minimize power deviation and reduce hydropower unit adjustments. This ensures optimal utilization of water resources and the safe operation of units, thereby improving system reliability. To ensure computational efficiency and meet the timeliness requirements of real-time scheduling, linearization techniques are employed to transform the model into a MILP model, enabling solution times within 1 min. The case study shows that: (1) The WGAN model achieves high prediction accuracy, effectively capturing the signature features of the PV power output and runoff processes. (2) Considering the dynamic water delay time can significantly increase system stability and prevent frequent unit adjustments. Compared to conditions when the water delay time is not considered or is simplified to a constant, the power deviation in CH-PVP complementary systems is reduced by 59.2 % and 49.2 %, respectively. (3) Sensitivity analysis indicates that slight increases in runoff and PV power output are beneficial for the coordinated operation of the CH-PVP system. However, system performance deteriorates significantly under extreme conditions where runoff and PV power output fluctuations exceed 30 %.