Deep Learning for Short-Term Electricity Demand Forecasting for Consumer Perceptive

Le résumé fourni par la source

Due to changes in consumer behaviour, climate variability, and the increasing proliferation of IoT devices, electricity consumption patterns are becoming more complex. Forecasting models must be accurate and responsive to these changes. Short-term electricity demand is often nonlinear and time-dependent, so traditional statistical methods are often ineffective. LSTM neural networks are employed in this study to enhance forecasting accuracy from a consumer-centric perspective, utilising multiple layers of stacked bidirectional long-term memory cells. This model captures intricate temporal patterns using historic load profiles, temperature, and humidity data and provides day-ahead, hourly predictions of electricity demand. The learning of the model is optimised by normalising data, incorporating forward and backward passes, and updating the weight matrix. An evaluation of the proposed model using MAE, RMSE, and MAPE reveals that it outperforms conventional methods, such as ARIMA, SNN, and DSHW, significantly. Using deep learning, it is possible to minimise forecasting errors and manage energy consumption according to individual requirements, making it a highly suitable method for integrating smart grids and managing personal energy consumption.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.