Optimization-oriented online modeling for generators of absorption heat pump systems

Abstract

An innovative online modeling approach for generators in absorption heat pump systems, which utilizes a combination of the K-means clustering algorithm and back propagation neural network (BPNN), is proposed. The input and output structure of the model is established by analyzing the generator's operating principle. To reduce the modeling burden, the K-means method is employed to cluster the observed data for the model parameter identification and the cluster center are used to train the model parameters. The cluster number K* is determined by the Davies-Bouldin Index, and the Particle Swarm Optimization (PSO) algorithm is employed to substitute the iterative computations of the clustering algorithm to accelerate the cluster centers' search process. The online model parameter correction function can utilize the online data to improve the accuracy and application range of the model. The experimental results show that the advantages of the model's online parameter correction function are highlighted with the various operating conditions, and the model's MRE and RMSE exhibit a steady decline, while the accuracy of the initial model continues to improve. This modeling technique will be widely used in the modeling process of heat pump systems and their components.