A Novel Deep Learning Ensemble Model with Secondary Decomposition for Short-Term Electricity Price Forecasting

Accurate electricity price forecasting plays a crucial role in the operation and development of the electricity market. In this paper, a novel hybrid model based on hybrid mode decomposition (HMD), convolutional long short term memory network (CNNLSTM), Elman neural network, and Bayesian optimization (BO) is proposed to forecast the electricity price. HMD is used to deeply decompose data into several subsequences, which consists of complete ensemble empirical mode decomposition with adaptive noise, sample entropy and empirical wavelet transform. CNNLSTM and Elman are adopted to forecast the subsequences. Besides, BO is introduced to optimize parameters. Finally, two case studies are taken to justify the effectiveness of the proposed forecasting model. The results show that the proposed model can possess significantly superior forecasting performance.