Research On Online Modeling Method Of Absorption Solar Heat Pump System Based On Clustering Method

The development of a steady-state model of the components of an absorption solar heat pump system plays a very important role in the optimal control of the system.Due to the non-linear nature of absorption solar heat pump systems and the complexity of the operating conditions,it is very difficult to establish an accurate theoretical model for absorption solar heat pump systems,so models of the components of heat pump systems are often based on data-based empirical modeling methods or hybrid modeling methods,both of which require the collection of data over a wide range of operating conditions to ensure the accuracy of the model and its applicability.The acquisition of large amounts of experimental data will not only increase the workload of the experiments,but will also increase the computational burden of model identification.The clustering method is used to cluster the observed data of the identification model,and the data of the cluster centre reflecting the basic characteristics of each type of data is used instead of the data of that type to participate in the model identification,which not only can greatly reduce the amount of data involved in the model identification,reduce the computational burden,but also can retain the basic characteristics of each type of data.By clustering the model identification data,in addition to reducing the burden of offline model identification,it is also possible to cluster and analyze the online data,identify data with characteristics different from those of existing classes to create new classes,and use the data centre of the class to make online corrections to model parameters,improve the accuracy of the model application and expand the range of model use.Based on the above analysis,this thesis proposes a method for online modeling of absorption solar heat pump systems using an improved K-means clustering method for clustering the data identified by the model,with regard to the characteristics of absorption solar heat pump systems.Its main research contains the following aspects:1)By analyzing the literature review on modeling and clustering methods for absorption solar heat pump systems,the K-means clustering method was chosen to cluster the model identification data according to the characteristics of the steady-state data of absorption solar heat pumps,and the determination of K-values and initial cluster centers in the method was improved.The Davies-Bouldin（DB）Index is used to determine the reasonable number of classifications in the K-means clustering method;the Particle Swarm Optimization（PSO）algorithm is used to determine the initial optimal clustering centres to improve the convergence speed of the K-means clustering method;the effectiveness of the improved K-means clustering method is verified by using three arithmetic examples.2)Based on an improved K-means clustering method,a clustering screening method for online model recognition data is proposed.The method uses a modified K-means clustering method to cluster the model identification data,and needs to use the cluster centre data representing the characteristics of each class of data as the observation data to identify the model parameters,and at the same time,the online measured data will be clustered and filtered for analysis,and data with new characteristics will be found,and new classes will be re-established,and the cluster centre data of the new classes will be added to the observation data,and the model parameters will be corrected online to improve the applicability of the model.3)Based on the improved K-means clustering method and the proposed online clustering filtering method for data,an online modeling method for each component of an absorption solar heat pump system is proposed.The online modeling method follows the existing hybrid model and parameter identification method for the components of an absorption solar heat pump system（evaporator,condenser,electronic expansion valve,heat recovery heat exchanger,solution pump,etc.）,uses an improved K-means clustering method to process the model identification data,upgrades the offline modeling method to an online modeling method,and validates the model with experimental data obtained from a built absorption solar heat pump system.The experimental results show that the accuracy of the model does not decrease after using the improved clustering method to cluster and filter the model identification data,while the computational effort of the model identification is reduced and the accuracy and applicability of the model is improved in the application.4)By analysing the operating principles of the absorber and generator in an absorption solar heat pump system,the input and output relationships of the two component models are established and an online modeling method using BP Neural Networks（BPNN）to establish the absorber and generator models is proposed.Through theoretical analysis,the absorber and generator input and output structures are determined,the number of hidden layers of the BP neural network is established using empirical formulas,and the weights and thresholds of the BP neural network are determined using a parameter identification method for a hybrid model of absorption solar heat pump system components.For the characteristics of BP neural network model,the determination range of the classification number K in the K-means clustering method is adjusted,and the online modeling of the absorber and generator is realised.The experimental results show that the BP neural network model established by this method has the function of online correction,and its model accuracy improves with the process of use.