Electric load forecasting based on deep learning and optimized by heuristic algorithm in smart grid

•A novel hybrid electric load forecasting model based on MMI-FCRBM-GWDO is developed for the decision making of a smart grid.•To overcome the problem of curse of dimensionality a novel MMI technique is proposed for feature selection.•The forecast accuracy and convergence rate of FCRBM is enhanced by GWDO.•The proposed model is tested on hourly load data of USA power grids: FE, Dayton, and EKPC. Accurate electric load forecasting is important due to its application in the decision making and operation of the power grid. However, the electric load profile is a complex signal due to the non-linear and stochastic behavior of consumers. Despite much research conducted in this area; still, accurate forecasting models are needed. In this article, a novel hybrid short-term electric load forecasting model is proposed. The proposed model is an integrated framework of data pre-processing and feature selection module, training and forecasting module, and an optimization module. The data pre-processing and feature selection module is based on modified mutual information (MMI) technique, which is an improved version of the mutual information technique, used to select abstractive features from historical data. The training and forecasting module is based on factored conditional restricted Boltzmann machine (FCRBM), which is a deep learning model, empowered via learning to forecast the future electric load. The optimization module is based on our proposed genetic wind-driven (GWDO) optimization algorithm, which is used to fine-tune the adjustable parameters of the model. The accuracy of the proposed framework is evaluated through historical hourly load data of three USA power grids, taken from publicly available PJM electricity market. The proposed model is validated by comparing it with four recent forecasting models like Bi-level, mutual information-based artificial neural network (MI-ANN), ANN-based accurate and fast converging (AFC-ANN), and long short-term memory (LSTM) in terms of accuracy and convergence rate.