Investigation into a Short-term Electric Load Prediction Model Leveraging K-means Clustering Enhanced Convolutional Neural Networks

The K-means clustering and convolutional neural network (CNN) integrated model for short-term power load forecasting is designed to enhance the precision of predicting electricity consumption in power systems through the application of state-of-the-art machine learning methodologies. In this paper, K-means algorithm is used to cluster the historical power load data, in order to excavate the inherent structure and regularity of load patterns, and divide them into multiple clusters with similar characteristics. Then, the deep convolutional neural network is designed and trained according to the different load features obtained from each cluster, so as to better capture the local features and time series dependencies in the time series data. The model combines unsupervised preprocessing technology and supervised learning method to improve the accuracy of short-term power load forecasting and reduce the forecasting errors caused by the diversity and uncertainty of load changes. The practical effectiveness and superiority of the proposed model in tackling real-world power system load forecasting tasks are experimentally substantiated.