Short-term Load Forecasting of Power System Based on Weighted Improved Adaptive Neuro-Fuzzy Inference System and Random Forest

Load forecasting of power system has become one of the important symbols to measure the modernization of power system operation and management, and it is very important for the safe and stable operation and economic dispatching of power system. Aiming at the problem that the accuracy of short term load forecasting of power system is reduced due to external interference, this paper first introduces the basic principle of related algorithms. Then, a short-term load forecasting method based on weighted adaptive neuro-fuzzy inference system (ANFIS) and random forest (RF) is proposed. A whale optimization algorithm (WOA) is employed to fine-tune the parameters within the membership function of the adaptive neuro-fuzzy inference system, while a random forest algorithm is utilized to identify the most important features. In conclusion, a comparison is made on the load data predictions using different algorithms. The adaptive neuro-fuzzy inference system is employed for short-term load forecasting in the power system, yielding experimental results with a maximum relative error of 1.6659%, a minimum relative error of 0.00768%, and an average absolute error of 64.352MW. By optimizing the membership function parameters of the adaptive neuro-fuzzy inference system using the whale optimization algorithm and weighting the results with random forest predictions, the experimental results showed a maximum relative error of 0.991%, a minimum relative error of 0.00008%, and an average absolute error of 24.84MW, further enhancing the accuracy of the predictions. Compared with the relevant algorithms, the proposed WOA-ANFIS-RF algorithm can further increase the nonlinear processing ability, reduce the error of load prediction and improve the accuracy of load prediction. In addition, the load prediction results obtained by the proposed algorithm meet the practical application requirements and have certain reference value.