Prediction of User Electricity Consumption based on Adaptive K-Means Algorithm

When predicting the total power load of many users, the computing resources often can’t keep up with the growth rate of the total amount of data, and it is difficult to analyze effectively the data in the actual environment. This paper firstly considers clustering users, then predicts each cluster separately, and finally summarizes the results of each cluster to get the result. This paper firstly performs PCA dimension reduction on user data, and then uses the adaptive K-Means clustering method to determine the number of clusters and the initial cluster center, and then uses the determined parameters to cluster the users, and then builds a model for each cluster user and sum up the forecast results to get the total power load. In order to illustrate the effect of this method under different models, this paper establishes XGBoost, CatBoost and LightGBM models respectively and predicts the total power load of all users. From the experimental results, it can be seen that this method is consistent with the actual data trend, and the prediction effect is better than that of directly modeling all user data.