Optimization of large-scale hydropower system peak operation with hybrid dynamic programming and domain knowledge

Abstract

With the rapid economic growth in recent years, the power demands in China keep growing and the need for reducing peak loads is becoming more prominent. With the merits of fast startup and shutdown, hydropower is often used to respond the peak load. In order to meet the practical requirement of peak operation in electrical power system, a novel min-max dynamic programming model is formulated for the peak operation of hydropower system. Then, the hybrid dynamic programming method is presented to alleviate the dimensionality problem in large-scale hydropower system, where the dynamic programming successive approximation is employed to divide the complex multi-dimensional problem into a series of small subproblems, and then the discrete differential dynamic programming is adopted to sequentially solve these subproblems. In addition, inspired by domain knowledge, the initial solution generation method and feasible space identification method are designed to promote the convergence speed of algorithm. The proposed method is used to solve the peak operation problem of a large-scale hydropower system in China. The simulations with different load demands indicate that the hybrid dynamic programming can achieve satisfactory performance in reducing peak loads of power system.