Abstract

Supplying power to regions and households via solar systems is a challenge, since energy production consistently varies owing to changing weather conditions and fluctuating consumer needs. Therefore, mechanisms must be established that accurately predict the user's expected needs and the energy produced. This article proposes a deep learning (DL) approach using a modified convolutional neural network (CNN) to evaluate the load response capability of residential energy customers by extracting typical load patterns. The precise evaluation of customer energy consumption patterns is essential for demand-side load management; nevertheless, it is affected by many variables, including load needs, weather conditions, and available power. A three-zone power usage, weather prediction, and time factor model are provided for the dataset (Tetouan City from 01/01/2017 to 30/12/2017). A modified CNN predicting accuracy is shown by the lowest mean absolute error of 0.1594, mean squared error of 0.04356, and the highest R-squared value of 0.9561. Thus, the suggested technique features a modified deep learning CNN model (Fig. 4) that exhibits considerable efficiency and stability in properly predicting power consumption, applicable in utility demand forecasting, intelligent building energy management (EM), and the integration of renewable energy (RE).