Energy Performance Prediction And Optimization Research Of Water-cooled Chiller System

With the continuous improvement of living quality and desire for high quality of theindoor environment, the central air-conditioning system has been widely used in variouspublic buildings. Due to the operation of central air-conditioning at part load and the lack ofeffective control, a large amount of energy has been wasted. As the most significant energyconsumption equipment, the sensible control of chiller can bring energy saving effect to thecentral air-conditioning system.An optimal control method of chiller is studied in this paper. For the need of generatingchiller models, an evolved BP neural network improved by additional momentum andadaptive learning rate was used, trained with data collected from actual operating data of fourchillers in a shopping mall in Guangzhou. According to Pearson correlation coefficient andbilateral significant test, chilled water supply temperature, cooling water temperature, chilledwater flow, cooling water flow and part load ratio were chosen as the input of the networkwhile COP was taken as the output. Compared with the traditional BP neural network, SLmodel, BQ model, MP model, SMP model, this evolved BP neural network achieves higherpredicted accuracy.The effect of the operating parameters, including chilled water supply temperature,cooling water return temperature, chilled water flow, cooling water flow and part load ratio,and the cooling load sharing strategy to the chiller’s COP was analysed. The results show thatthe change of the parameters can cause the variation of COP. Besides, the characteristic ofrangeability of COP of each unit is not consistent. Compared with two units with samerefrigerating capacity, the COP of one is not always higher than the other. The results alsoshow that unequal cooling load sharing brings higher aggregate COP of chillers than equalcooling load sharing.According to the analysis above, the optimal control method of chiller based on geneticalgorithm is purposed. Aiming at achieving maximum aggregate COP, this method determinesthe sequencing and optimum load sharing strategy for the multiple-chiller system based on theoperating condition of on-line chillers. The energy conservation effect of this optimizationunder summer, transition season and winter is verified by simulation based on the chiller models. The results show that energy consumption reduces by7.51%using the optimalcontrol method instead of the original control method.The optimal control method was realized by developing a program on the basis of hybridprogramming using Matlab under C#programming environment. Experiments were taken toprove the energy saving effect of the optimal control. The results indicate that energyconsumption reduces by5.79%using the optimal control method instead of the originalcontrol method.