Short-term wind power prediction model based on PCA-GA-LSTM neural network

In order to improve the prediction accuracy of short-term wind power, a PCA-GA-LSTM short-term wind power combination prediction algorithm based on NWP is proposed. First, principal component analysis is used to reduce the dimension and denoise NWP sample data, and then genetic algorithm is used to train and optimize the model parameters of short-term memory artificial neural network (LSTM), such as learning rate, number of hidden layer nodes, regularization coefficient, etc. Finally, NWP historical data after dimension reduction is used as learning samples to train the neural network, and real-time information is processed to obtain prediction results. The results show that the prediction accuracy based on PCA-GA-LSTM neural network model is good, which has certain reference value for short-term wind power prediction.

I. Introduction

With the increasing attention paid by the international community to environmental protection, energy security and other issues, reducing fossil fuel consumption and speeding up the development of new energy technologies have become the key tasks of all countries. Among them, wind power has become the main force in building a clean and low-carbon energy system and promoting green and low-carbon development because of its low cost, clean and safe, large reserves and other advantages. By the end of August 2022, the installed capacity of wind power in China is 344.5 million kilowatts, up 16.6% year on year; The installed capacity of wind power increased by 16.14 million kilowatts, an increase of 1.5 million kilowatts over the previous year [1]. However, due to the strong volatility, intermittency and regional characteristics of wind power, grid connected wind power generation has brought severe challenges to the reliable and stable operation of the power system while providing clean energy. Therefore, accurate wind power prediction is of great significance for maintaining the stable operation of new power systems.