Using Gated Recurrent Unit Network to Forecast Short-Term Load Considering Impact of Electricity Price

Abstract The volatility of renewable energy and the time variability of load bring challenges to load forecasting. To improve the accuracy of prediction, the paper use gated recurrent unit (GRU) network to forecast short-term load considering the impact of electricity price. First, methods are proposed to group historical data based on the main features of input. Second, the rules of classification tree are established to judge which cluster the new data belongs to. The gated recurrent unit network is trained by features of input and power load from the selected group. Finally, the real-world data are used to analyses the impact of electricity price and demonstrate the validity of GRU network. The simulation study shows that the proposed approaches can improve the accuracy for forecasting short-term load and have better performance than traditional methods.