An Innovative Method for Short Term Electrical Load Forecasting Based on Adaptive CNN - MRMR Model

The importance of precise electric demand forecasting has grown in light of the recent deregulation and privatization of power infrastructure. More accurate predictions, at all scales from end-users to the system as a whole, are required by the new energy market and smart grid paradigm. Social activities, environmental factors, and seasonal components are only a few of the many elements that make electric demand projection challenging. There is a specific order in which preprocessing, feature selection, and training the model must be executed. Using Wold's theorem, a “prototypical” day/week was generated for use in preprocessing filters. Feature extraction is a method that particle swarm optimization use. A CNN-MRMR method was used to train the model. When contrasted with the cutting-edge CNN and MRMR, this method appears to be quite novel. With an accuracy percentage of 97.20 percent, the data showed a significant improvement.