Research On Energy Consumption Prediction And Control Optimization Of Central Air Conditioning Based On Data Mining

With the continuous increase of carbon emissions,energy issues have become an essential topic concerned by the world.As a typical air conditioner,the energy consumption of central air-conditioning system has a high ration in building energy consumption due to its large volume and long working time.Therefore,the research on efficient operation of central air conditioning system is of important implications to energy conservation and pollution reduction.The central air-conditioning system of a large market is taken as the study of object in thesis.Based on the operation data collected by sensors,the energy efficiency ratio of the system is optimized,so as to ensure that the air conditioning energy consumption is reduced on the basis of meeting the working conditions,and energy saving and emission reduction of building energy consumption are realized.The main contents of this thesis are as follows（1）The operation data are preprocessed and feature selection is carried out.According to the parameter characteristics,the stored data in the database is preprocessed,and all extracted parameters are processed by feature engineering.According to the characteristics of random forest and extreme gradient boosting,the fusion algorithm of both was selected.The gain of loss function and the frequency of decision tree branch attribute are taken as the evaluation function,and a set of feature subset with high importance order is selected.（2）Different neural networks are trained to predict energy consumption.The corresponding data of selected feature parameters are constructed into sequential data set associated with each other in the form of sliding Windows.Four network models,including back propagation neural network,long and short-term memory network,convolutional neural network and deep residual network,were respectively trained and constructed.The mean square error and mean absolute percentage error were used as evaluation functions to evaluate the regression energy efficiency ratio at each time point.Compared with other models,the overall error of the deep residual network model is lowest,which could more accurately predict the subsequent change of the energy efficiency ratio of the system.（3）Then an optimization model is established to optimize the energy efficiency ratio of the central air-conditioning system.In the input running data,the frequent pattern growth algorithm is used to obtain the potential correlation between each parameter value in thesis.And the convergence characteristic of the particle swarm optimization algorithm is improved by niche rule.The deep residual network model is set as the objective function in the optimization model,the parameters of the normal operating range and mining the association rules are set as constraint condition of optimization.The improved particle swarm optimization algorithm is applied to optimize the energy efficiency ratio of system,and the relative optimal operating point under various conditions of the system was acquired.（4）A Simulink model was established to verify the optimization effect.The empirical formula is used to build the energy consumption identification model of each device,and the corresponding Simulink model is established.Dynamic simulation was executed from the initial operating parameters to the set parameters of the optimization,which proved the actual effect of the optimization model.After that,a long period of energy consumption was optimized.Through the optimization of the data model and the verification of the parameter identification model,the feasibility and accuracy of the optimization model are demonstrated,which provides a certain reference for the energy consumption optimization of water-cooled central air conditioning.