Integration of Artificial Intelligence in Future Smart Grids: An LSTM-RNN based Approach for Optimizing Energy Efficiency in Smart Grids

Smart grids leverage data-driven methodologies to anticipate consumers' energy demand, enabled by deep learning models analyzing vast amounts of data. These modern approaches address demand forecasting challenges, facilitating efficient electricity transportation based on anticipated consumption patterns. Deep learning's trend identification in customer data empowers accurate demand estimates across diverse forecasting horizons. In this paper, we present a customized lightweight LSTM-RNN model for load prediction. By virtue of low parameter count, the proposed model is amenable to embedded applications in real-time monitoring equipment. The prediction results show marked improvement over the predecessor models, while reducing the number of trainable parameters. The overall MSE and MAE of the model, which has only 50,241 parameters, was found to be 0.000525 KW
²
and 0.022930 KW when tested over dataset.