An Ensemble Deep-Learning-Based Model for Hour-Ahead Load Forecasting with a Feature Selection Approach: A Comparative Study with State-of-the-Art Methods

The realization of load forecasting studies within the scope periods varies depending on application areas and estimation purposes. It is mainly carried out at three intervals: short-term, medium-term, long-term. Short-term (STLF) incorporates hour-ahead forecasting, which critical for dynamic data-driven smart power system applications. Nevertheless, based our knowledge, there are not enough academic prepared with particular emphasis this sub-topic, none related evaluate STLF methods in regard. As such, machine learning (ML) deep (DL) architectures forecasters have recently been successfully applied to STLF, state-of-the-art techniques energy area. Here, methods, majority frequently preferred high-performing up-to-date literature, were first examined different using two aggregated-level datasets observing effects these both. Case comparison conducted before, but many examples studied data from structures. Although used study each other terms time step, they also had very varied features. In addition, feature selection was both a backward-eliminated exhaustive approach performance artificial neural network (ANN) validation set proposed development models. A new DL-based ensemble after examining results obtained separate by applying working numerical illustrate that it can significantly improve compared methods.