Medium-Term Regional Electricity Load Forecasting through Machine Learning and Deep Learning

Due to severe climate change impact on electricity consumption, as well new trends in smart grids (such the use of renewable resources and advent prosumers energy commons), medium-term long-term load forecasting has become a crucial need. Such forecasts are necessary support plans decisions related capacity evaluation centralized decentralized power generation systems, demand response strategies, controlling operation. To address this problem, main objective study is develop compare precise district level models for predicting electrical based machine learning techniques including vector (SVM) Random Forest (RF), deep methods such non-linear auto-regressive exogenous (NARX) neural network recurrent networks (Long Short-Term Memory—LSTM). A dataset nine years historical Bruce County, Ontario, Canada, fused with climatic information (temperature wind speed) used train after completing preprocessing cleaning stages. The results show that by employing learning, model could predict more accurately than SVM RF, an R-Squared about 0.93–0.96 Mean Absolute Percentage Error (MAPE) 4–10%. can be not only municipalities utility companies distributors management expansion grids; but also households make adoption home- district-scale technologies.