Short-Term Load Forecasting Based on TVF-EMD and CNN-GRU Optimized By DBO

The economic and stable functioning of power systems is reliant upon precise load forecasting in the short-term. A novel short-term electricity load forecasting model, TVFEMD-DBO-CNN-GRU, is proposed in this paper by combining time-varying filtering based empirical mode decomposition (TVF-EMD) and dung beetle optimization algorithm (DBO). A convolutional neural network (CNN) and a gated recurrent unit network (GRU) are both present. Firstly, addressing the nonlinearity and non-stationarity issues of electricity load data, the TVF-EMD algorithm decomposes the electricity load data into different sub-sequences. Subsequently, to avoid the randomness in parameter selection, the DBO is used to optimize the CNN-GRU model hyperparameters. Finally, the CNN-GRU combination network is utilized to predict individual components. The forecast result of each component are added together, a short-term electricity load forecast is ultimately achieved. Using publically accessible datasets from the Australian region, to confirm the effectiveness of the proposed paradigm, empircal research is done. The suggested model has been demonstrated to be more successful than other forecasting models, as evidenced by experimental results.