Multi-Reservoir Fully Connected Echo State Network for Wind Power Prediction

Optimized control of wind power plants has a high demand for wind power prediction accuracy. Accurate wind power prediction data enables wind power plants to better arrange production plans, improve energy utilization efficiency, reduce energy losses, and make the power grid operation stable. In this paper, considering the impact of wind speed instability and unpredictability on wind power generation prediction, a multi-reservoir fully connected echo state network (MRFC-ESN) is established. And the model is used to predict wind power generation. Firstly, this model draws on the fully-connected approach of neurons, and fully connects multiple reservoirs. Secondly, to demonstrate the stability and practicality of MRFC-ESN, the stability proof of MRFC-ESN is provided. Thirdly, MRFC-ESN uses gradient descent to optimize parameters. Finally, through the prediction of a set of wind power data, it is verified that this method achieves good results in wind power prediction.