Prediction of building energy consumption for public structures utilizing BIM-DB and RF-LSTM

The demand for innovative predictive models to improve building energy efficiency underscores the urgent need to address environmental issues stemming from high energy consumption, particularly in public buildings with independently installed air conditioning systems for heating and cooling. This study presents a novel approach for early prediction of electricity consumption in public buildings utilizing random forest (RF) and long short-term memory (LSTM) networks, marking a significant advancement in this domain. The methodology was applied to enhance the early energy efficiency of a university research facility in Xinyang. Initially, a building information model of the teaching structure was developed using Design Builder energy analysis software and subsequently imported. Next, relevant datasets were generated through energy consumption simulations. Building on this foundation, the RF model facilitated feature selection across various factors and dimensionality reduction prior to applying the LSTM method. Finally, the predictive outcomes from the LSTM model were compared with those from alternative models. The findings indicate that the combination of RF and LSTM demonstrates substantial advantages in predicting building energy consumption, establishing itself as the most effective predictive model among several comparative frameworks. This breakthrough allows for accurate estimation of annual energy usage during the preliminary stages of building design while identifying opportunities for enhancing energy efficiency.