Research on prediction method of flashover voltage in SF<sub>6</sub> gas under nanosecond pulse

Sun Chuyu; Chen Wei; Wang Haiyang; Ji Shengchang

doi:10.11884/HPLPB202638.250181

Article Contents

Article Navigation > High Power Laser and Particle Beams > 2026 > Accepted Manuscript

Sun Chuyu, Chen Wei, Wang Haiyang, et al. Research on prediction method of flashover voltage in SF6 gas under nanosecond pulse[J]. High Power Laser and Particle Beams. doi: 10.11884/HPLPB202638.250181

Citation:

Sun Chuyu, Chen Wei, Wang Haiyang, et al. Research on prediction method of flashover voltage in SF₆ gas under nanosecond pulse[J]. High Power Laser and Particle Beams. doi: 10.11884/HPLPB202638.250181

Citation:

PDF( 2690 KB)

Research on prediction method of flashover voltage in SF₆ gas under nanosecond pulse

doi: 10.11884/HPLPB202638.250181

Sun Chuyu^{1, 2
,},
Chen Wei²,
Wang Haiyang²,
Ji Shengchang^{1
,
,}

1.
School of Electrical Engineering, Xi’an Jiaotong University, Xi’an, 710049, China
2.
Northwest Institute of Nuclear Technology, Xi’an, 710021, China

Received Date: 2025-06-24
Accepted Date: 2025-09-02
Rev Recd Date: 2025-09-24

Available Online: 2025-11-25

Abstract

Abstract

Background
The surface flashover in SF₆ under nanosecond pulses involves complex physical processes, and accurately predicting the surface flashover voltage of insulating media in such environments constitutes a critical challenge for the design of high-voltage pulsed power equipment and the evaluation of insulation reliability. Compared with traditional AC or DC voltages, the extremely short rise time and high amplitude of nanosecond pulses lead to significant space charge effects and distinct discharge development mechanisms, thereby posing severe challenges to prediction models based on classical theories. In recent years, with the rapid improvement of computer computing power and breakthroughs in artificial intelligence algorithms, data-driven machine learning methods have demonstrated great potential in solving complex nonlinear insulation problems.
Purpose
Targeting this specific challenge under nanosecond pulses, this paper selects four algorithms, including support vector machine (SVM), multi-layer perceptron (MLP), random forest (RF), and extreme gradient boosting (XGBoost), to train and predict flashover voltage data under different experimental conditions within the multi-scale distance range of 15 mm to 500 mm.
Methods
First, external operating conditions such as electric field distribution, voltage waveform, and gas pressure were parametrically extracted and characterized. The Pearson correlation coefficient was employed to conduct a correlation analysis on the aforementioned characteristic parameters, and ultimately 22 feature quantities were screened out as the model inputs. Subsequently, the Bayes hyperparameter optimization algorithm was utilized to perform hyperparameter optimization for four types of algorithms, and the 10-fold cross-validation method was adopted to select the optimal hyperparameter combination for each algorithm. After that, the sample training set was input into the four algorithms for training, and each algorithm was validated on the test set.
Results
The four algorithms demonstrated overall good performance. Among them, Random Forest (RF) and XGBoost exhibited excellent performance on the training set but poor performance on the validation set, which is likely a manifestation of overfitting in ensemble learning and indicates weak generalization ability. Support Vector Machine (SVM) achieved relatively outstanding performance on both the training set and the validation set. Furthermore, the generalization performance of the SVM and XGBoost algorithms was validated using data outside the sample dataset. The results showed that SVM yielded better prediction outcomes on the data outside the sample dataset.
Conclusions
SVM achieved high prediction accuracy on the training set, test set, and data outside the sample dataset, making it more suitable for the insulation design of electromagnetic pulse simulation devices.
- surface flashover,
- support vector machine,
- voltage prediction,
- artificial intelligence,
- data-driven

FullText(HTML)

References(29)

References

[1]	邱孟通, 呼义翔, 吴伟, 等. 西北核技术研究所强脉冲辐射模拟装置近年发展综述[J]. 现代应用物理, 2024, 15: 030101 Qiu Mengtong, Hu Yixiang, Wu Wei, et al. Review of the development of intense pulsed radiation simulator and its technology in northwest institute of nuclear technology in the past decade[J]. Modern Applied Physics, 2024, 15: 030101
[2]	朱湘琴, 吴伟, 王海洋. 大型水平极化电磁脉冲有界波模拟器的辐射场分布特性分析[J]. 现代应用物理, 2020, 11: 040502 Zhu Xiangqin, Wu Wei, Wang Haiyang. Characteristics of radiation electric field distribution in large EMP bounded wave simulator with horizontal polarization[J]. Modern Applied Physics, 2020, 11: 040502
[3]	Peek F W. The law of corona and the dielectric strength of Air-IV the mechanism of corona formation and loss[J]. Transactions of the American Institute of Electrical Engineers, 1927, XLVI: 1009-1024.
[4]	Zaengl W S, Petcharaks K. Application of streamer breakdown criterion for inhomogeneous fields in dry air and SF₆[M]//Christophorou L G, James D R. Gaseous Dielectrics VII. New York: Springer, 1994: 153-159.
[5]	Niemeyer L, Ullrich L, Wiegart N. The mechanism of leader breakdown in electronegative gases[J]. IEEE Transactions on Electrical Insulation, 1989, 24(2): 309-324. doi: 10.1109/14.90289
[6]	Seeger M, Niemeyer L, Bujotzek M. Partial discharges and breakdown at protrusions in uniform background fields in SF₆[J]. Journal of Physics D: Applied Physics, 2008, 41: 185204. doi: 10.1088/0022-3727/41/18/185204
[7]	MartinT H, Guenther A H, Kristiansen M. J. C. Martin on pulsed power[M]. New York: Springer, 1996.
[8]	刘锡三. 高功率脉冲技术[M]. 北京: 国防工业出版社, 2005 Liu Xisan. High pulsed power technology[M]. Beijing: National Defence Industry Press, 2005
[9]	Ghosh P S, Chlakravorti S, Chatterjee N. Estimation of time-to-flashover characteristics of contaminated electrolytic surfaces using a neural netlwork[J]. IEEE Transactions on Dielectrics and Electrical Insulation, 1995, 2(6): 1064-1074.
[10]	杨庆, 司马文霞, 蒋兴良, 等. 复杂环境条件下绝缘子闪络电压预测神经网络模型的建立及应用[J]. 中国电机工程学报, 2005, 25(13): 155-159 Yang Qing, Sima Wenxia, Jiang Xingliang, et al. The building and application of a neural network model for forecasting the flashover voltage of the insulator in complex ambient conditions[J]. Proceedings of the CSEE, 2005, 25(13): 155-159
[11]	石岩, 蒋兴良, 苑吉河. 基于RBF网络的覆冰绝缘子闪络电压预测模型[J]. 高电压技术, 2009, 35(3): 591-596 Shi Yan, Jiang Xingliang, Yuan Jihe. Flashover voltage forecasting mode of iced insulator based on RBF network[J]. High Voltage Engineering, 2009, 35(3): 591-596
[12]	王永强, 王巨伟, 张斌, 等. 基于FLN的覆雪绝缘子闪络电压预测方法[J]. 电瓷避雷器, 2017(6): 241-245 Wang Yongqiang, Wang Juwei, Zhang Bin, et al. Study on prediction method of flashover voltage of snow-covered insulators based on FLN[J]. Insulators and Surge Arresters, 2017(6): 241-245
[13]	舒立春, 白困利, 胡琴, 等. 基于支持向量机的复杂环境条件下绝缘子闪络电压的预测[J]. 中国电机工程学报, 2006, 26(17): 127-131 Shu Lichun, Bai Kunli, Hu Qin, et al. Insulator flashover voltage forecasting under complex circumstance based on support vector machine[J]. Proceedings of the CSEE, 2006, 26(17): 127-131
[14]	阮江军, 徐闻婕, 邱志斌, 等. 基于支持向量机的雾中棒–板间隙击穿电压预测[J]. 高电压技术, 2018, 44(3): 711-718 Ruan Jiangjun, Xu Wenjie, Qiu Zhibin, et al. Breakdown voltage prediction of rod-plane gap in fog based on support vector machine[J]. High Voltage Engineering, 2018, 44(3): 711-718
[15]	Vapnik V N. The nature of statistical learning theory[M]. New York: Springer, 1995: 138-146.
[16]	邓乃扬, 田英杰. 支持向量机: 理论、算法和拓展[M]. 北京: 科学出版社, 2009: 43-62 Deng Naiyang, Tian Yingjie. The theory, algorithm and application of support vector machine[M]. Beijing: Science Press, 2009: 43-62
[17]	高同虎. 基于SVM的附着藻类绝缘子电场畸变与闪络电压预测研究[D]. 淄博: 山东理工大学, 2020 Gao Tonghu. Prediction of electric field distortion and flashover voltage of algae-covered insulators based on SVM[D]. Zibo: Shandong University of Technology, 2020
[18]	李云淏, 咸日常, 张海强, 等. 基于改进灰狼算法与最小二乘支持向量机耦合的电力变压器故障诊断方法[J]. 电网技术, 2023, 47(4): 1470-1477 Li Yunhao, Xian Richang, Zhang Haiqiang, et al. Fault diagnosis for power transformers based on improved grey wolf algorithm coupled with least squares support vector machine[J]. Power System Technology, 2023, 47(4): 1470-1477
[19]	郝蛟, 林宏, 李雨森, 等. 基于改进多层感知机的电网运行风险评估方法[J]. 现代电力, 2023, 40(4): 474-483 Hao Jiao, Lin Hong, Li Yusen, et al. A method to assess power grid operation risk based on improved multi-layer perceptron[J]. Modern Electric Power, 2023, 40(4): 474-483
[20]	Hayati M, Rezaei A, Seifi M. CNT-MOSFET modeling based on artificial neural network: application to simulation of nanoscale circuits[J]. Solid-State Electronics, 2010, 54(1): 52-57. doi: 10.1016/j.sse.2009.09.027
[21]	罗艳, 肖辅盛, 王庭刚, 等. 基于随机森林的电网实时运行风险评估方法[J]. 信息技术, 2020, 44(4): 23-26,31 Luo Yan, Xiao Fusheng, Wang Tinggang, et al. Real-time risk assessment method for power grid operation based on random forest[J]. Information Technology, 2020, 44(4): 23-26,31
[22]	周云浩, 杨宝杰, 刘丹, 等. 基于随机森林算法的电力工程数据预测分析建模与仿真[J]. 电子设计工程, 2024, 32(4): 103-106,111 Zhou Yunhao, Yang Baojie, Liu Dan, et al. Modeling and simulation of power engineering data prediction and analysis based on Random Forest algorithm[J]. Electronic Design Engineering, 2024, 32(4): 103-106,111
[23]	Mantas C J, Castellano J G, Moral-García S, et al. A comparison of random forest based algorithms: random credal random forest versus oblique random forest[J]. Soft Computing, 2019, 23(21): 10739-10754. doi: 10.1007/s00500-018-3628-5
[24]	Chen Jinxiang, Zhao Feng, Sun Yanguang, et al. Improved XGBoost model based on genetic algorithm[J]. International Journal of Computer Applications in Technology, 2020, 62(3): 240-245. doi: 10.1504/IJCAT.2020.106571
[25]	Su Wei, Jiang Fan, Shi Chunyan, et al. An XGBoost-based knowledge tracing model[J]. International Journal of Computational Intelligence Systems, 2023, 16(1): 13. doi: 10.1007/s44196-023-00192-y
[26]	刘琳, 李晓昂, 张锐, 等. 影响GIS支柱绝缘子闪络电压的沿面电场特征参数[J]. 高电压技术, 2019, 45(9): 2740-2747 Liu Lin, Li Xiaoang, Zhang Rui, et al. Characteristic parameters of electric field along the surface affecting the flashover voltage of GIS pillar insulators[J]. High Voltage Engineering, 2019, 45(9): 2740-2747
[27]	Qiu Zhibin, Ruan Jiangjun, Lu Wei, et al. Feature extraction of electric field distribution and its application in discharge voltage prediction of large sphere-plane air gaps[J]. IEEE Transactions on Dielectrics and Electrical Insulation, 2018, 25(3): 1030-1038. doi: 10.1109/TDEI.2018.006851
[28]	樊嵘, 孟大志, 徐大舜. 统计相关性分析方法研究进展[J]. 数学建模及其应用, 2014, 3(1): 1-12 Fan Rong, Meng Dazhi, Xu Dashun. Survey of research process on statistical correlation analysis[J]. Mathematical Modeling and its Applications, 2014, 3(1): 1-12
[29]	崔佳旭, 杨博. 贝叶斯优化方法和应用综述[J]. 软件学报, 2018, 29(10): 3068-3090 Cui Jiaxu, Yang Bo. Survey on bayesian optimization methodology and applications[J]. Journal of Software, 2018, 29(10): 3068-3090