A Novel Approach for Optimizing Building Energy Models Using Machine Learning Algorithms

The current practice with building energy simulation software tools requires the manual entry of a large list detailed inputs pertaining to characteristics, geographical region, schedule operation, end users, occupancy, control aspects, and more. While these allow evaluation consumption various combinations parameters, information considering number parameters related design global optimization is extremely challenging. In present paper, novel approach developed for models (BEMs) using Python EnergyPlus. A Python-based script automate data into modeling tool (EnergyPlus) numerous possible designs that cover desired ranges multiple variables are simulated. resulting datasets then used establish surrogate BEM an artificial neural network (ANN) which optimized through two different approaches, including Bayesian genetic algorithm. To demonstrate proposed approach, case study performed on campus Florida Institute Technology, located in Melbourne, FL, USA. Eight selected 200 variations them supplied EnergyPlus, produced results from simulations train ANN-based model. model achieved maximum 90% R2 hyperparameter tuning. algorithm method, were applied model, optimal annual consumptions 11.3 MWh 12.7 MWh, respectively. It was shown presented bridges between physics-based strong available Python, can achievement computationally efficient fashion.