TCN Based on Charging Pile Load Prediction Considering Key Meteorological Factors

The charging load of electric vehicles(EVs) has strong randomness, which is not only related to historical charging behavior, but also influenced by meteorological factors. Changes in weather conditions may cause users to adjust their travel plans or adjust their charging behavior based on travel plans, resulting in a “time shift” in the charging load. Therefore, using the strong ability of time convolutional networks to explore the potential relationship between load and meteorological data, a time convolutional network EV charging load prediction method considering key meteorological factors is proposed. Firstly, analyze the impact of different meteorological factors such as temperature, rainfall, wind speed, snowfall, relative humidity, and total sunshine intensity on EV charging load, and extract key meteorological factors using the maximum correlation minimum redundancy criterion. Secondly, to simplify the process of selecting similar days and ensure the accuracy of classification, a method for selecting charging load similar days under different weather categories is constructed; Finally, key meteorological factors and historical loads of similar days are used as input vectors for EV charging load prediction in a time convolutional network. Comparative experiments show that considering meteorological influences can effectively improve the accuracy of load forecasting. This method can be applied to short-term load forecasting for different regions and types of users, demonstrating the universality of the forecasting method.