Short-term power forecasting model based on GWO-LSTM network

Abstract In view of the time-series characteristics grid load data, this paper proposes a method to predict electricity demand by optimizing long-and short-term memory (LSTM) neural network model using grey wolf optimization algorithm, taking into account effects time, weather conditions and holiday on loads. The overcomes disadvantage that backpropagation through time algorithm tends converge local optimum. experimental results show prediction outperform those traditional LSTM for loads, providing reference direction future forecasting models.