Machine learning enhanced inverse modeling method for variable speed air conditioning systems

Various faults may occur in the air conditioning systems due to improper installation and poor maintenance. Various fault detection and diagnosis methods have been developed, which need lots of data to evaluate the protocols. However, experimental data is usually not sufficient. The FDD protocols especially machine learning based can easily overfit the limited experiment data. It may be not satisfied for the real applications because of wider range of operation. The machine learning enhanced inverse modeling method is presented to generate the simulation data under various conditions of different scenarios. The clustering algorithm is used to classify the training data reasonably balancing the weights of different conditions. The particle swarm optimization (PSO) is developed to obtain the global optimal estimation of model parameters under wider operation conditions. The experimental data of both variable and constant speed systems are used to validate clustering-PSO enhanced algorithm, which shows acceptable capability and accuracy of prediction. (C) 2020 Elsevier Ltd and IIR. All rights reserved.