| Summary: | To ensure the computational accuracy for the offshore high voltage direct current substation local structure and its large and complex electrical equipment, it is necessary to reduce the modeling complexity and improve the calculation efficiency. Thus, a simplified vibration model with constant key physical parameters was established; the converter valve was used as a test design example and a lower number of samples was used. Finally, the response surface fitting was carried out. Based on the machine learning concept, problems such as the loss of physical meaning of complex system matrices and low iteration efficiency, caused by high-dimensional matrix operations were avoided. In addition, the deterministic screening experiment design greatly reduced the number of samples needed while also ensuring the second-order effect. The fit between the response surface equation and experimental results was good after optimizing the relevant design factors based on the vibration equation. The comparison between the simplified vibration converter valve model and the original complex model has shown that the errors for various physical parameters and the vibration characteristics were below 5% and that the model simplification degree was 95%.
|
|---|