抄録
This paper presents a method to develop HVAC equipment models with BMS data using Soft Computing. Four types of models are made and predicting performance are evaluated and compared to each other. First model is multiple regression model. Second is Neural Network model. Both third and fourth models are physical models from TRNSYS, widely known dynamic simulation program, but one's parameters are estimated by genetic algorithm and the others are estimated by experiment in laboratory. Electric power consumption of air-cooled chiller is predicted. Ratio of average error to rated power consumption is 5.7%, 4.6%, 3.8% and 14.3%, respectively.
