A GA-LSSVM approach for predicting and controlling in screw chiller

Abstract

Performance of varying speed screw chiller is affected by many uncertainties. High precision prediction of its characteristics can guide the chiller to reach a better performance. This study presents an artificial intelligence model named least square support vector machine (LSSVM) with genetic algorithm (GA). Five parameters are predicted with the model, including COP, discharge pressure, suction temperature, suction pressure and cooling capacity. By comparing the simulation results with the test results, this model shows a high precision ability to predict the performance of the on-site chiller. Additionally, a newly control strategy is introduced to help the chiller with optimizing performance. Cooling capacity and superheat degree are separately used as input to train the model to control openness of EXV. The prediction of this control strategy process shows enough ability to predict openness of EXV. The results can be used to guide the chiller to reach better performances by adjusting the corresponding parameters.