An Accurate Power Load Forecasting Technique Based on the Fusion of Clustering Algorithms

Abstract

An upgraded ANN model and clustering algorithm are coupled in a suggested combined model forecasting technique to increase the prediction accuracy of power system load forecasting. The samples are clustered using the clustering method, and the data with comparable properties are used as the input for the prediction, strengthening sample regularity and increasing prediction accuracy. The classic ANN prediction model is also enhanced by the multi-output technique, which increases the level of model fitting and brings the output closer to the actual value. The proposed clustering technique is integrated to create the combined prediction model. Finally, the combined model’s efficacy is evaluated using the data set from the 2012 Global Energy Competition load forecasting competition. The proposed method enhances prediction accuracy and has strong learning and adaptability capabilities when compared to the conventional ANN prediction model and conventional deep learning prediction model.