Multi-day Load Forecasting Method In Electricity Spot Markets Based on Multiple LSTMs

摘要

Multi-Day Load Forecasting is the basis of Electricity Spot Markets Analysis and operation. At present, the Research on Electricity Spot Markets Load Forecasting has not considered the influence of Day-ahead Market on Real-Time Market, which leads to deviation and consumes a lot of time and energy. Therefore, this paper presents a Multi-Day Load Forecasting model based on Multiple Long Short Term Memory(M-LSTM), Consider the effects of Day-Ahead Locational Marginal Price(LMP) and Day-Ahead Cleared. The proposed method solves the problem that the single deep learning network structure is difficult to keep the time sequence characteristics between samples in the training process. In this paper, Case studies on the New England Electricity Market (ISO-NE) show that our proposed method is superior to S-LSTM in the forecasting. The MAPE of the model is 14.01%, compared with Single Variable LSTM(S-LSTM), it decreased by 3.33%. The RMSE of the model is 457.22 MW, compared with Single Variable LSTM(S-LSTM), it decreased by 98.09 MW. Experiments show that the model has higher prediction accuracy and generalization ability.