Research on building energy consumption prediction algorithm based on customized deep learning model

Abstract

Forecasting energy usage in buildings is essential for implementing energy saving measures. Precisely forecasting building energy use is difficult due to uncertainty and noise disruption.To achieve enhanced accuracy in predicting energy use in buildings, a deep learning approach is proposed. This paper proposes a customized convolutional neural network with Q-Learning (CCNN-QL) based reinforcement learning algorithm for predicting energy consumption in building.The suggested CCNN-QL model offers an auto-learning feature that predicts building energy consumption through an automated method, continually improving its predictive accuracy.To assess its performance, various building types were selected to study the factors influencing excessive energy consumption, and data were collected from multiple Chinese cities. The suggested model’s performance has been assessed using evaluation metrics, resulting in a low Mean Absolute Error (MAE) and Root Mean Square Error (RMSE), indicating superior accuracy relative to comparable studies. Experimental results indicate that the suggested technique has superior predictive performance across several scenarios of building energy usage.