Power prediction and operation scheduling strategy of pumped storage power station based on machine learning

Abstract

The power of a pumped storage power station depends on the water level of the upstream reservoir and the downstream demand. The change in water level is related to rainfall, evaporation, river inflow, and other factors. Therefore, it is difficult to accurately predict the generating power and its scheduling. A machine learning-based power prediction and operation scheduling strategy for pumped storage power plants is proposed. The relationship between the forgetting gate, input gate, output gate, and memory unit of the long and short-term memory network is redefined. The activation function between input data and output data is constantly updated, which is used as the power prediction function of the pumped storage power station. The power prediction model is built. The operation scheduling process of the pumped storage power station is designed. The operation scheduling of the pumped storage power station is completed by using the daily scheduling time scale. The experimental results show that the research method can accurately predict the power generation of a pumped storage power station. The power generation is more balanced, and the power scheduling effect is better under the application of the research method.