Predicting Electricity Consumption in Commercial and Residential Buildings Using A-CNN-LSTM Model

In this proposed approach an A-CNN-LSTM model that, over the medium to long term, can forecast, down to the hourly level, the patterns of electricity use in commercial and residential buildings. Institutional and residential buildings account for the vast bulk of America's overall energy usage. Thanks to developments in smart technologies and sensors, there is a need to forecast the medium- to long-term hourly electricity consumption of home and business buildings. This is to help with decision-making for operations, demand response strategies, and distributed generation system installations. To achieve the best possible results during model training, feature selection, and preprocessing, it is essential to adhere strictly to this strategy. Since some data on daily precipitation was lost during pre-processing, the Lagrange interpolation method was used to fill in the gaps. Feature selection makes use of PSO-K-means clustering technique. An A-CNN-LSTM was employed for the purpose of training the model. Traditional methods, such as LSTM and CNN, are surpassed by the suggested approach. Applying the strategy resulted in a 93.15 % improvement in accuracy.