Variable refrigerant flow (VRF) valve fault detection and diagnosis usually face the problems of too many features and low efficiency. Therefore

a high-efficiency hybrid model based on a genetic algorithm (GA) and back propagation neural network (BPNN) was proposed. In this hybrid model

the feature subset is extracted from the original feature set of the VRF using the GA

and then the parameter-optimized neural network is used to detect and diagnose VRF valve faults. In this study

the hybrid model was used to detect and diagnose faults with electronic expansion valve sticking

leaking

and a four-way valve. The results showed that the hybrid model proposed in this paper could effectively and reliably diagnose faults. The integrated correct rate of fault diagnosis reached a peak value of 99.27%. In particular

the correct rate of electronic expansion valve sticking fault diagnosis was improved by 8%. In addition

the hybrid model obviously improved the detection and diagnosis efficiency

decreasing the operating time by 52.17%.