Auto Machine Learing Based Energy Consumption Prediction And Optimization Research Of Chillers

As the main energy-consuming equipment of the central air-conditioning system,the control strategy of water chillers will directly affect the overall energy consumption of the air conditioning system and even buildings.Therefore,how to make the central air-conditioning system energy-saving under the requirements of comfort has become one of the research hotspots in this field,and accurate energy consumption modeling of chillers and operation optimization are of great significance for energy saving of central air conditioning.Based on the designed collection scheme and the magnetic suspension chiller in a hotel in Nanjing,this paper collects 16 kinds of operating parameters and chiller energy consumption with sensors,and uses it to evaluate the performance of the model based on automatic machine learning energy consumption prediction.Then,the air-conditioning water system model of a simple mall building with two chillers of one large and one small rated refrigerating capacity is established through simulation,and the annual hourly energy consumption simulation operation is carried out to obtain simulated energy consumption data,which are used to verify the performance of the energy consumption prediction model and the feasibility of the chiller operation optimization strategy.The traditional prediction method is difficult to apply to engineering practice due to its complicated system,different types of operating parameters,and its serious dependence on physical principles and other disadvantages.This paper applies auto machine learning methods to the prediction of energy consumption of chillers,which can effectively solve these problems.The results show that,based on the simulated operation data of the simulated building,the automatic machine learning model obtains a good energy consumption prediction result,and its prediction error is generally less than 5%.Based on the real system data,the RMSE,CV-RMSE,and R~2 of the automatic machine learning model reached 1.82k W,12.95%,and 0.9917,respectively,which are the best results among all models.Based on the establishment of the chiller energy consumption prediction model and the energy consumption simulation data,this paper has identified the main factors affecting the chiller energy consumption.Taking the lowest energy consumption of water chillers as the optimization objective,the operation parameter optimization research and load distribution optimization research were carried out respectively.The former outputs control parameters including chilled water outlet temperature and cooling water inlet temperature,and the latter outputs the load rate of each chiller.The above optimization problem is solved by a non-dominated sorting genetic algorithm with elite strategy,and the energy consumption at each load rate before and after optimization is compared.The results show that the proposed method can achieve better energy-saving performance at different load rates.