Development of an AI-Based Model with Low Computational Complexity for Accurate Load Demand Forecasting

Abstract

This research addresses the challenge of short-term load demand forecasting in microgrids, where renewable energy unpredictability destabilizes power systems. Current forecasting models often suffer from high computational complexity, resulting in increased power consumption and reduced real-time applicability. To overcome these limitations, this study develops and optimizes an Artificial Neural Network (ANN)-based shortterm forecasting model with significantly reduced computational demands. In this study, a model was constructed utilizing historical operational data from a microgrid system. To optimize the computational efficiency of the model, various techniques were applied to reduce its complexity. The model’s performance was systematically evaluated using appropriate performance metrics. The experimental results demonstrate that the proposed approach significantly decreases the computational complexity of the final model, while preserving an acceptable level of accuracy when compared to the original, unoptimized model. The practical implications of this research include enabling real-time demand forecasting on resource-constrained microgrid controllers and edge devices, facilitating more efficient energy management in sustainable power systems. Future work will focus on enhancing the model's generalization capabilities by incorporating additional geographical and climatic factors, enabling accurate demand forecasting across diverse microgrid environments beyond the specific conditions of the initial dataset.