AI-driven energy optimization in university microgrid: A hybrid LSTM-RNN forecasting consumption patterns approach

This paper presents a novel methodology for classifying and forecasting electric power consumption within a university microgrid. Our study utilizes one year of high-resolution data collected at 15-minute intervals from real energy consumption, capturing detailed temporal variations in usage to identify distinct consumption patterns and accurately predict future energy demand. In the first stage of the proposed approach, the K-means clustering algorithm is employed to identify representative load profiles and characterize distinct consumption patterns, such as peak, mid peak and off-peak periods. In the second stage, a hybrid deep learning framework is proposed for load forecasting, combining Long Short-Term Memory (LSTM) and Recurrent Neural Network (RNN) architectures to effectively capture both short-term fluctuations and long-term dependencies in energy demand. Experimental results demonstrate that the proposed hybrid model achieves high prediction accuracy, with performance metrics ranging from 83 % to 96.4 %. These results highlight the model’s effectiveness in managing complex consumption behaviors and its potential to support advanced energy optimization strategies in academic microgrid environments.