High-precision wind power prediction is important for the planning, economics, and security maintenance of a power grid. Meteorological features and seasonal information are strongly related to wind power prediction. This paper proposes a hybrid method for ultrashort-term wind power prediction considering meteorological features (wind direction, wind speed, temperature, atmospheric pressure, and humidity) and seasonal information. The wind power data are decomposed into stationary subsequences using the ensemble empirical mode decomposition (EEMD). The principal component analysis (PCA) is used to reduce the redundant meteorological features and the algorithm complexity. With the stationary subsequences and extracted meteorological features data as inputs, the long short-term memory (LSTM) network is used to complete the wind power prediction. Finally, the seasonal autoregressive integrated moving average (SARIMA) is innovatively used to fit seasonal features (quarterly and monthly) of wind power and reconstruct the prediction results of LSTM. The proposed method is used to predict 15-minute wind power. In this study, three datasets were collected from a windfarm in Laizhou to validate the prediction performance of the proposed method. The experimental results showed that the prediction accuracy was significantly improved when meteorological features were considered and further improved with seasonal correction.
Two-Step Wind Power Prediction Approach With Improved Complementary Ensemble Empirical Mode Decomposition and Reinforcement Learning