Short-term optimal coordinated operation of a wind-solar-hydro hybrid system based on deep learning and double-layer nesting solution algorithm

Abstract Due to its randomness, intermittence, and volatility, the high-proportional integration of wind and solar power presents considerable challenges for the traditional approaches to the safe and stable operation of power systems. Cascade hydropower stations have a high response speed, high adjustability, and stable output. They can be the optimum candidates for the regulation and compensation of wind and photovoltaic (PV) power generation instability. This study proposed a wind-solar-hydro hybrid system, and investigated its short-term optimal coordinated operation on the basis of deep learning and a double-layer nesting algorithm. A stochastic complementary scheduling model was constructed to maximize the cascade energy storage. To reduce the problem-solving complexity, particle swarm optimization algorithm-dynamic programming (PSO-DP) coupled with the inner and outer nesting optimization algorithm was proposed. The hybrid system was applied to a national comprehensive development base of renewable energy with integrated wind, solar, and hydro power in China. Studies have shown the following: The wind-solar-hydro hybrid system has a certain degree of scalability. The utilisation of deep learning methods can fully consider the uncertainty of wind and solar and fit output scenarios well, the correlation coefficient is above 0.9. The internal and external nested optimisation algorithm, which realises the reasonable and efficient distribution of water and electricity among cascade hydropower stations and improves the future power generation capacity of cascade hydropower stations, was used to solve the problem. This study provided an approach to optimise the operation plan of grid dispatching, including large-scale wind power and PV power, and a valuable reference for the large-scale utilisation of other renewable energy sources worldwide.