A modeling approach for the absorber of absorption heat pump systems

Abstract

The study on the data-and-knowledge-driven modeling methods is very important for the digital twinning and the optimal control of absorption heat pump systems. For the purpose of obtaining an absorber model with high accuracy, low computational effort and less time-consuming for the absorption heat pump systems, a modeling method is proposed by combining the Simple Particle Swarm Optimization (SPSO) and the Wavelet Neural Network (WNN). The input-output structure of the model is presented by analyzing the operational principles of the absorber. In view of the complexity of the heat transfer and mass transfer processes inside the absorber, a single hidden layer WNN is employed as the internal structure of the model. The number of hidden layer nodes in the network model is determined by the trial-and-error method using the root mean square error and coefficient of determination. To overcome the drawback of WNN which can easily fall into the local optima during the error backpropagation process and improve the prediction accuracy of the model, the Morlet wavelet basis function and a SPSO algorithm are exploited respectively to replace the S-function of Neural Network and determine the optimal parameter values of the neural network (e.g. the weight, scaling factor, and translation factor, etc.). Experimental simulation results show the root mean square errors of the proposed model are 1.59 % and 2.21 % for the two outputs, respectively, which are smaller than the conventional WNN model. The proposed modeling approach effectively solves the problems of the BP neural network on easy falling into the local optimum and the slow convergence, which will be widely used in the process of the optimized operation control for heat pump systems.