Research On Control Method Of Air Conditioning System Based On Data Mining Technology

In order to solve the limitation of the current air conditioning control strategies,on the basis of the large amount of data of the air-conditioning system accumulated in the building energy management system,this thesis firstly use data mining technology to build the air conditioning system energy model,optimize the control parameter and forcast the load of air conditioning system,then propose a feedforward control strategy and develop a feedforward air conditioning control strategy and thus achieve energy saving of air conditioning system.Firstly,based on the quality characteristics of the monitoring data of building energy consumption,a preprocessing framework and method suitable for building energy monitoring data is proposed.Then it is applied to the data preprocessing of the actual case,for example,the kNN algorithm is used to fill in the missing value,the K-Means algorithm is used to identify and clean the abnormal data,and the PCA algorithm is used to reduce the data dimension,etc.The availability of data is realized.Secondly,on the basis of the analysis of the basic model of the chiller,the principle and method of feature engineering in machine learning are applied.The model of chiller suitable for this study is put forward,and the modeling method of the interval section of the load division is put forward.Then,according to the generalized linear regression algorithm,cross validation and error analysis,the energy consumption model of a chiller,a chilled water pump and a cooling water pump is set up based on the actual operating data,and the model error is within 5%,then the energy cunsumption model of cold source system is established.Next,according to the energy consumption model of the cold source system,the minimum energy consumption of the cold source system is as the objective function.The chiller load is the input parameter,the chilled water outlet temperature,the cooling water inlet temperature,the chilled water pump flow and the cooling water pump flow are used as the output control parameters,thus the multi-objective optimization problem of the cold source system control optimization is established.The optimization control method based on genetic algorithm and particle swarm optimization is established respectively,and the results of the algorithm are compared.The particle swarm optimization algorithm is superior to the genetic algorithm in two aspects of convergence and efficiency.Then using the particle swarm optimization algorithm,it realizes the optimization of the control parameters of the cold source system in the partial load section,and compares the energy consumption before and after the optimization of each load rate,and achieves different energy saving effects at different load rates.Then,the factors that affect the energy consumption of the air conditioning system and the availability of data are analyzed,and the input parameters are radicated as input parameters of load forecasting.In this study,the air conditioning load forecasting method are set up based on kNN algorithm,support vector machine SVM algorithm,support vector regression SVR algorithm and artificial neural network ANN algorithm respectively.On the basis of the actual data,the load forecasting results of different methods are compared.The results of SVR algorithm and artificial neural network ANN algorithm are best,the prediction accuracy can be whithin 5%,and the training time of support vector regression SVR algorithm is about 2~3 times of the artificial neural network ANN algorithm.Under the condition of sufficient data,artificial neural network ANN model is the best option for load forecasting,however when the data amount is small,the SVR algorithm is the first choice for load forecasting.At last,on the basis of control optimization and load forecasting,the feedforward control strategy of air conditioning system is proposed in this study,that is,"taking a hour as the basic unit,the upper limit load in the interval of time to hourly forecast load is used as the input parameter,and the optimized control parameters are used as the control set values of each equipment of the air conditioning cold source system".After the analysis of the typical day,it is found that the daily energy saving by 20.8% is realized through the feedforward control.According to the proposed feedforward control strategy,the feedforward unattended control system of the air conditioning system is developed,and the practical engineering application of the control system is carried out,then the energy saving effect of one month is analyzed.The system can achieve about 27.4% energy saving compared with the conventional control system.