An improved LSTM-Seq2Seq-based forecasting method for electricity load

Power load forecasting has gained considerable research interest in recent years. The power is vulnerable to randomness and uncertainty during grid operations. Therefore, it crucial effectively predict the electric improve accuracy of prediction. This study proposes a novel method based on an improved long short-term memory (LSTM) neural network. Thus, network model established for forecasting, which supports variable-length inputs outputs. conventional convolutional (CNN) recurrent (RNN) cannot reflect sequence dependence between output labels. LSTM-Seq2Seq prediction was by combining sequence-to-sequence (Seq2Seq) structure with that accuracy. Four models, i.e., memory, deep belief (DBN), support vector machine (SVM), LSTM-Seq2Seq, were simulated tested two different datasets. results demonstrated effectiveness proposed method. In future, this can be extended more application scenarios.