Prediction of Electric Energy Consumption for Demand Response using Deep Learning

This paper emphasizes the capability of Deep Learning (DL) models to conquer the Demand Response (DR) inherent when predicting the Electric Energy Consumption (EEC) of an office building. The prediction of EEC plays a key role in DR programs in a smart grid environment. In this study, historical energy consumption and ambient temperature data of three different climatic days (summer, winter, and cloudy days) of an office building located in Portugal at 10 seconds intervals are taken. A DL technique-based Deep Neural Network model is proposed for the prediction of future EEC. In this paper predictability of EEC of the whole office building has been analyzed. This study describes an evince DL application for commercial energy consumption prediction at 10 seconds intervals and performed precursory success. Moreover, two conventional Machine Learning (ML) models i.e., Support Vector Regressor (SVR) and Random Forest (RF) are developed and analyzed. Furthermore, the proposed DL model is compared with SVR and RF in terms of performance evaluation parameters such as Mean Absolute Error (MAE), Mean Square Error (MSE), and Root Mean Square Error (RMSE). All the models are developed and executed on TensorFlow deep learning platform. The proposed model defeats SVR by 91.65%and RF by 87.38% on a summer day, similarly defeats SVR by 93.85% and RF by 91.68% on a winter day and defeats SVR by 95.63% and RF by 92.67% on a cloudy day in terms of MSE.