Short-term Wind Power Prediction Based on SSA-BiLSTM-AT Model
To address the complexity and diversity of short-term wind power,a Bi-directional Long Short-Term Memory(BiLSTM)neural network containing AttentionMechanism and Sparrow Algorithm(SSA)tuned to the short-term wind power prediction model,SSA-BiLSTM-AT,is proposed.Firstly,the input data are processed and normalized with outliers,and the relationship between wind power and each feature is analyzed by the Pearson correlation coefficient method to analyze the relationship between wind power and each feature,and eliminate the features with low correlation in the data,so as to improve the predic-tion accuracy of the model;Aiming at the problem of BiLSTM hyperparameter selection difficulty,the four important parameters of learning rate,adaptation degree,and the number of nodes in the 1st and 2nd implicit layers in BiLSTM are optimized intelligently and iteratively by the sparrow algorithm,and the op-timal parameters are obtained and then the BiLSTM is used for prediction;finally the attention mechanism is introduced,and the attention weights are used to highlighting the influence of key factors and mining the internal laws of wind power data.The stability and effectiveness of the proposed model's linear regression fitting ability are verified by using the historical data of a wind power station in Xinjiang as a practical cal-culation example.
short-term wind power forecastBi-directional Long Short-Term MemoryAttention mecha-nismsparrow search algorithm
万方数据
维普
