Evaluating the Performance of the AI Models to Predict the Energy Performance of Selected Office Building

The largest energy consumers in the world are buildings and also these are major contributors to the greenhouse gas emissions. Hence, it is important to improve the efficiency of the building energy. Hence, predicting the performance of energy used in building helps in designing the buildings in a sustainable manner. The conventional tools for simulation such as EnergyPlus are accurate but there is need of expert knowledge, high computational efforts and detailed inputs to use them. For this machine learning provides a simple and faster alternative. As there exists many machine learning models like ANN, Support vector machine, Random Forest and Gradient Boosting. A dataset was generated by EnergyPlus for an office room which is located in Amritsar, India. Inputs for the system includes window-to-wall ratio, shading properties and surface absorption. The outputs were daylight performance, energy load and thermal comfort. The performance of the model was evaluated using MAE, RMSE and R². The results showed that the performance of the ANN was best overall, especially while doing prediction of energy load and thermal comfort, with lower errors and higher accuracy than other models.