Assessment of low energy potential of a school building using operation optimization and surrogate models

The retrofit of existing buildings has a notable potential to save energy and reduce the environmental impact. This paper presents an approach to assess the primary energy saving potential related to the retrofit of a school building. The retrofit includes wall and roof insulation, fenestration replacement, installation of heat pumps with thermal storage, and usage of photovoltaic panels. The approach relies on building simulations and operation optimization evaluate a limited number of retrofit combinations. The results are used to formulate two kinds of surrogate models based on gradient boosting: classifier that finds feasible options and regressor that estimates the primary energy consumption. The results show high precision of the predictive models. The F1 score of the classifiers exceeds 0.99 even for very small training samples. The most important feature for estimating feasibility is the heat pump capacity. The coefficient of determination of the regressors is close to 1 and the root mean square error is lower than 1 kWh/m2, even for moderate sizes of the training set. The most important feature for predicting energy consumption is related to the area and orientation of photovoltaic panels.