Residential Electricity Power Forecasting using GRU-based CNN and SAE Ensembler

Smart grids play an important role to resolves issues related to electricity. Electrical load forecasting can be performed through smart grids to acquire knowledge about the electrical load that will be needed in the future. A combination of Stacked Autoencoding (SAE) and Gated Recurrent Unit (GRU) and a combination of Convolutional Neural Network (CNN), Grey Wolf Optimization (GWO), and GRU are proposed in this paper. Firstly, the historical load data is preprocessed and then preprocessed data is given to the feature selection and extraction methods to measure the importance of each feature. Features that have highest values are selected and input to the proposed classifiers. According to the results the proposed CNN-GRU-GWO gave the highest accuracy rate of 96.31%. This application of this article will help the utility purchase the exact amount of electricity from the grid, which will lead to grid stability.