Multi-user Load Forecasting Method Based on Multi-scenario Algorithm Models

With the continuous improvement of the national electricity marketization policy, it is particularly important for the generation measurement, sales side, operation side and management side to accurately predict the electricity load curve of each user in the next day and every hour of the day. In this paper, based on the historical electricity consumption data of users, cluster analysis and variational mode algorithm are first used to carry out targeted refinement and decomposition of the electricity load data, and then combined with weather data, user characteristics, electricity consumption characteristics and other data through feature fusion and feature dimension enhancement, a neural network algorithm model under multiple scenarios is established. Finally, an efficient model screening algorithm is designed to accurately predict the user’s electricity load. The results show that using the modified Imape deviation evaluation model as the benchmark, this strategy algorithm can greatly reduce the forecast deviation rate of single user’s electricity load, and improve the accuracy of the total electricity load of single user and all users.