Research on Electricity Load Forecasting Method Based on GRU-LightGBM Combined Model with Bayesian Optimization

Abstract

Abstract Accurate power load forecasting is the key to the efficient operation of the power system and the optimal allocation of resources. Improving the accuracy and robustness of load forecasting can optimize the promotion of renewable energy integration and reduce operational costs, thus promoting green energy. In this paper, a combined model prediction method based on Bayesian optimization for gate control recurrent unit (GRU) and light gradient boosting machine (LightGBM) is proposed. First, the optimal parameters of the GRU and LightGBM models are found, and the corresponding training models are built by Bayesian optimization (BO), respectively. Then, using the least absolute deviation regression (LAD) weighting, the prediction results of the two single models are combined to build the final BO-GRU-LightGBM combined prediction model. Finally, the performance of the single and combined models and the performance of different optimization algorithms for optimizing GRU-LightGBM are compared through Matlab simulation experiments to verify the accuracy and robustness of the established models, respectively. Compared with the traditional load forecasting methods, the model has a root mean square error (RMSE) value of only 0.331, which is highly practical and provides an effective solution to improve the accuracy of power load forecasting, helps in the management and utilization of energy resources, and provides a strong support for building a more efficient, reliable, and sustainable energy system.