**6. Conclusions**

A prediction model of water content in transformer oil using multi frequency ultrasonic with a PCA-GA-BPNN was established. The topological structure of the model was 8-5-1, and the generalization ability of the model was tested with test sets. The experimental results show that the accuracy rate of this model is higher than 90%.

Different structured networks have a different prediction performance. Compared with the BPNN and GA-BPNN models, the PCA-GA-BPNN model can more accurately predict water content in transformer oil according to multi frequency ultrasonic data.

The predictive model of water content in transformer oil using multi frequency ultrasonic with PCA-GA-BPNN, which was proposed in this paper, provides a new online detection method for transformer oil for the power industry. In addition, the application of multi frequency ultrasonic testing technology to detect other parameters of transformer oil is the key point of future research.

**Author Contributions:** Conceptualization, Z.Y. and Q.Z.; methodology, Z.Y. and Q.Z.; validation, X.W., Z.Z. and C.T.; investigation, Z.Y. and X.W.; resources, C.T.; data curation, Q.Z.; writing—original draft preparation, Z.Y.; writing—review and editing, Z.Y., Q.Z. and W.C.; visualization, Z.Z.; supervision, Q.Z.; project administration, Q.Z.

**Funding:** This work has been supported in part by the National Natural Science Foundation of China (No. 51507144), Fundamental Research Funds for the Central Universities (No. XDJK2019B021), the China Postdoctoral Science Foundation funded project (Nos. 2015M580771, 2016T90832) and the Chongqing Science and Technology Commission (CSTC) (No. cstc2016jcyjA0400).

**Conflicts of Interest:** The authors declare no conflict of interest.
