Online modeling method of generator based on K-means clustering algorithm and BP neural network

• Published in: IEEE Conference on Industrial Electronics and Applications (Online) pp. 1857 - 1862
• Main Authors: Ma, Haoxiang, Ding, Xudong, Sun, Hao, Yang, Dongrun, Sun, Mei, Zhao, Xingkai
• Format: Conference Proceeding
• Language: English
• Published: IEEE 18.08.2023
• Subjects: BP neural network; Clustering algorithms; Data collection; Filtering algorithms; generator; Heating systems; K-mean clustering algorithm; Neural networks; online modeling; Prediction algorithms; Predictive models
• ISSN: 2158-2297
• DOI: 10.1109/ICIEA58696.2023.10241454

This paper proposes a novel method for online modeling of generators in absorption refrigeration systems, which utilizes a combination of the K-mean clustering algorithm and BP neural network. Firstly, the method analyzes the working principle of the generator in the refrigeration system and establishes an input and output structure for the generator model. Then, the K-mean clustering algorithm and particle swarm optimization algorithm are employed to cluster and filter the experimental data, with the clustering centers being used to replace the original data. Subsequently, the BP neural network modeling method is employed to obtain the model parameters, which are substituted into the model to predict the heat exchange of the generator, and compared with the measured data. Moreover, the proposed method utilizes the parameters of an offline model as the initial parameters of the online model, and then re-identifies and clusters the model parameters to obtain the BPNN model of the generator. The experimental results demonstrate that the proposed modeling method can accurately predict the heat exchange of the generator under a small range of operating conditions. Furthermore, the online model identification method improves the accuracy of the model and reduces the scope of experimental data collection required for identifying the model parameters.