A CNN-LSTM Hybrid Model Based Short-term Power Load Forecasting
Chang Ren, Li Jia, Zhangliang Wang
Abstract
Power load forecasting is always one of the most important research focuses in power systems, which can assist power companies to optimize dispatching and reduce cost. The artificial intelligence models represented by neural network are widely used with the increasing demand of prediction accuracy. However, the potential characteristics of load series cannot be well captured based on single artificial intelligence model. To solve the problem, a CNN-LSTM hybrid model is proposed to further explore characteristics information contained in power load sequence. This model applies convolution layer of CNN to capture the features of power load data and makes power load forecasting using unique cellular structures of LSTM. This hybrid model is verified based on historical power load data acquired from Shanghai and Belgium. The prediction accuracy of CNN-LSTM outperforms other compared forecasting models after the observation of simulation results.