Quantitative influence of multiple factors on arc duration of high power density DC relays based on back-propagation neural network and local feature importance analysis

• Quantitative influence of multiple factors on arc duration in HVDC relays. • Local feature importance analysis method. • A novel approach for the measurement and prediction of arc duration. • A novel approach for guiding the regulation of arc duration. Arc duration is one of the key factors determining the arc-extinguishing performance and contact erosion damage of high power density direct current (DC) relays. Since the arc duration is affected by multiple factors, including power, transverse magnetic field (TMF), and gas, it is critical to clarify the quantitative influence of these factors on the arc duration for its prediction and regulation. Current research predominantly focuses on the effects of individual factors, while the effects arising from the coupling of multiple factors have received comparatively less attention. Therefore, this paper measures the arc duration under multiple factors through breaking arc experiments. It then develops a method that combines back-propagation neural network (BPNN) with local feature importance analysis to elucidate the quantitative influence and influence degree of multi-factor coupling on the arc duration. The results demonstrate that the proposed BPNN model effectively maps the quantitative relationship between the characteristic parameters of various factors and arc duration, making it applicable for arc duration prediction. The stability of the results is enhanced through K -fold cross-validation, with the average mean squared error (MSE) across all samples being 0.0077. The subsequently obtained influence degree of each characteristic parameter of multiple factors on the arc duration provides valuable guidance for regulating arc duration in high power density DC relays. Overall, the influence of TMF on arc duration is consistently negative and the most significant. In contrast, increases in voltage and current generally result in prolonged arc duration, with current having a stronger impact. The influence of gas plasma physical properties on arc duration varies across temperature ranges. In detail, density, thermal conductivity, and electrical conductivity at 5000–10000 K, as well as thermal conductivity, viscosity, and electrical conductivity at 10000–15000 K, are particularly significant. These properties exhibit an influence comparable to that of voltage and current.