Short-term wind power prediction based on combined long short-term memory

Wind power is an exceptionally clean source of energy; its rational utilization can fundamentally alleviate the energy, environment, and development problems, especially under the goals of 'carbon peak' and 'carbon neutrality'. A combined short-term wind power prediction based on long short-term memory (LSTM) artificial neural network has been studied aiming at the non-linearity and volatility of wind energy. Due to the large amount of historical data required to predict the wind power precisely, the ambient temperature and wind speed, direction, and power are selected as model input. The Complete Ensemble Empirical Mode Decomposition with Adaptive Noise has been introduced as data preprocessing to decompose wind power data and reduce the noise. And the Particle Swarm Optimization is conducted to optimize the LSTM network parameters. The combined prediction model with high accuracy for different sampling intervals has been verified by the wind farm data of Chongli Demonstration Project in Hebei Province. The results illustrate that the algorithm can effectively overcome the abnormal data influence and wind power volatility, thereby providing a theoretical reference for precise short-term wind power prediction. In this paper, CEEMDAN is introduced as data preprocessing to decompose wind power data and reduce noise. For the non-linearity and volatility of wind energy, the combined short-term wind power prediction based on LSTM artificial neural network is investigated by optimizing the parameters of the LSTM model through PSO.image