A Hybrid Method of Cooling and Heating Consumption Prediction for Six Types of Buildings Based on Machine Learning

Sustainable development is a vital strategy that is being implemented in China. To achieve sustainable development in terms of building energy efficiency, accurately estimating the amount of energy that buildings will consume is crucial. A theoretical framework for machine learning-based building energy consumption prediction is presented in this study; six different types of building information models in five major thermal design zones of China were used for gathering information and forming a database. The suggested prediction model’s distinctive feature is that nine factors affecting building energy consumption in three aspects, including macro-view, middle-view, and micro-view aspects, are proposed, eight machine learning techniques are employed to predict building energy consumption, and the factors influencing energy consumption are identified. Two standard measures were employed to evaluate the framework’s performance: the coefficient of determination (R²) and the root mean square error (RMSE). It was found that the accuracies of all eight models were above 90%. Among them, the kNN model and GBRT have the best prediction results. Using the optimal GBRT model, the feature importance ranking was obtained. The proposed machine learning prediction model informs similar studies and can be applied to predict different buildings’ cooling and heating loads accurately.