Load Forecasting for Classified Agency Electricity Purchasing Customers Based on Electricity Usage Characteristics Analysis

In the electricity market, accurate forecasting of proxy power purchases is crucial for power agents to optimize their power purchase strategies, reduce costs, and improve the efficiency of the power system. Given that the loads of different agent power purchasers are affected by multiple factors such as weather to varying degrees, and the characteristics of their power consumption behaviors, traditional forecasting methods are no longer applicable when dealing with complex and variable volume environments. Therefore, this paper proposes a load forecasting for classified agency electricity purchasing customers based on electricity usage characteristics analysis, aiming to improve the forecasting accuracy and stability of the model. Firstly, user classification is realized by analyzing the electricity consumption characteristics of each proxy purchaser and identifying different typical electricity consumption patterns using the CLIQUE clustering algorithm. Then, the influence of meteorological factors on the power purchased by each type of agent is evaluated through correlation and characterization, and the input features are screened accordingly to enhance the adaptability of the model to different types of power consumption. In order to further improve the prediction accuracy, this paper introduces a temporal convolutional network (TCN) to effectively capture the nonlinear relationship between electricity consumption and input features. The experimental results demonstrate that the feature-screened TCN model performs well under different types of electricity consumption and significantly improves the prediction performance.