Optimization of Night Cooling of Commercial Premises Using Genetic Algorithms and Neural Networks

This paper investigates if it is possible to optimize night cooling control setpoints and ventilation schedule regarding energy consumption and indoor climate. A retail store, located in Gothenburg, was used as a case study. The investigation was done by numerical modelling and simulations. It started with development and calibration of a building energy model for the store with data collected from the field. Afterwards, the calibrated model was used in the optimization of the night cooling. Initially, a genetic algorithm was applied to find the global minimum of the problem and further refined with a local search algorithm. The optimization speed was increased by neural networks, as they can approximate results faster than the building energy model. The study suggests that the cooling and fan energy consumption can be reduced by 16% in the studied facility, compared to the currently used trial-and-error schemes. The project concludes that the use of logged control data in combination with genetic algorithms and neural networks are an efficient way for both calibration and optimization of building energy models. The industry moves towards an increase of available logged control data. As such, it is important to be able to properly utilize the data, for improving the accuracy of building energy simulations and improving the results.