Building Performance Simulation To Support Tree Planting For Cooling Needs Reduction: A Machine Learning Approach

Greening the city is recognised as a main strategy to improve cities liveability, outdoor environment and buildings’ energy efficiency in summer. This work proposes a machine learning approach to predict, based on certain number of previously run simulations, the contribution of trees’ shadows to cooling needs reduction in Mediterranean climates. This procedure can allow urban planners to evaluate a specific situation in terms of some easily observed parameters (building shape, type of trees, distance from the main facade, orientation, number of facades shadowed) and to obtain a fast estimation of cooling reduction or a classification in ranges of effectiveness of the configuration examined. We used two strategies to predict cooling loads of buildings: a single threshold and a five categories evaluation. The obtained accuracy is about 95% with a single threshold value and about 70% with a five-categories classification.