Short-term Load Forecasting Model for Smart Grid

Abstract

In the economic dispatch of power system. How to reasonably use the past and present of power load to speculate its future value has very long-term socio-economic value. Short-term power load forecasting is mainly used to predict the power load in the next few hours or days. The relationship between weather factors and load changes is very important for short-term forecasting. Short-term power load data has obvious temporal characteristics, and the traditional RNN model is more and more applied in this field. However, the RNN model may have gradient explosion or gradient disappearance. Therefore, based on the long-term and short-term memory neural network (LSTM), an improved AM-LSTM short-term load forecasting model is presented. The model improves the activation function in LSTM unit into weighted activation function group, and adds attention mechanism to improve the prediction accuracy.