Ultra-Short-Term Wind Speed Forecasting Based on Meta Learning with Signal Trend and Fluctuation Decomposition

Abstract

Accurate wind speed prediction for each wind turbine is a critical basis of information for intelligent management and control of wind power system. How to precisely realize rapid prediction with small sample size is a critical open problem. This paper proposes an ultra-short-term wind speed prediction method based on meta learning algorithm with trend and fluctuation decomposition of wind speed signal. A meta learning prediction model with long short-term memory (LSTM) and recurrent neural network (RNN) is constructed as the base-learner, where low-frequency signal trend and high-frequency signal fluctuation are taken as the input of LSTM and RNN respectively. The forecasting results show that the mean absolute percentage error (MAPE) of the wind speed prediction scheme proposed in this paper is about 2.54%, and the sample size and training time costed in training are about 2.5% and 1.7% of traditional LSTM network. Results indicate that this method realizes ultra-short-term wind speed prediction with high accuracy as well as high efficiency.