Optimization of Typical Industrial Load Forecasting Based on CNN-BILSTM-Attention-NGO

The randomness and fluctuation of short-term power load are strong, and the traditional load forecasting method is difficult to grasp the law of short-term load change. In order to improve the accuracy of power load forecasting, a CNN-BILSTM-Attention-NGO typical industrial load forecasting Optimization method was proposed based on the Northern Goshawk Optimization (NGO) algorithm and the traditional forecasting model. After feature extraction by CNN, the traditional joint prediction model uses BiLSTM to effectively capture the long-term dependence between sequences, and then uses the Attention algorithm to ensure that some important feature information of the original model is not ignored. In order to solve the problem of complex data of power system, this paper introduces the NGO optimization algorithm, adopts the gradient decline of adaptive learning rate, and establishes a new optimization framework through the normalization and balance of weights, and constructs a new activation function and forgetting gate mechanism to improve the performance of the model. The results before and after optimization are compared and analyzed by practical examples. The simulation results show that the proposed method can accurately grasp the law of load change, achieve more accurate load prediction, and the fitting effect is better.