Suzhou Electric Appliance Research Institute
期刊號: CN32-1800/TM| ISSN1007-3175

Article retrieval

文章檢索

首頁 >> 文章檢索 >> 最新索引

基于STFT與改進(jìn)ConvNeXt配電網(wǎng)故障區(qū)段定位方法研究

來源:電工電氣發(fā)布時間:2024-08-01 15:01 瀏覽次數(shù):296

基于STFT與改進(jìn)ConvNeXt配電網(wǎng)故障區(qū)段定位方法研究

鄧思敬1,2,吳浩1,2,鄧力川1,2,蔡源1,2
(1 四川輕化工大學(xué) 自動化與信息工程學(xué)院,四川 宜賓 644000;
2 人工智能四川省重點實驗室,四川 宜賓 644000)
 
    摘 要:在目前的配電線路智能故障診斷研究方法中,存在著難以充分提取故障特征、抗噪聲干擾能力弱、抗高阻能力差等問題。提出了一種基于短時傅里葉變換(STFT)并引入遷移學(xué)習(xí)的改進(jìn) ConvNeXt 配電網(wǎng)故障區(qū)段定位方法。該方法通過采集配電網(wǎng)各饋線兩端的零序電流,計算出各饋線兩端的零序電流幅值差,然后將各段的零序電流幅值差拼接成一個組合信號,用 STFT 處理組合信號,得到時頻圖,并將得到的時頻圖分為訓(xùn)練集和測試集。仿真結(jié)果表明,基于 STFT 并改進(jìn)的 ConvNeXt 配電網(wǎng)故障區(qū)段定位方法在不同的故障距離、不同的接地電阻和不同的初始故障角度下都能有效地實現(xiàn)故障區(qū)段的選擇,并且該方法具有較強的抗高阻能力以及較強的抗噪聲干擾能力,在部分?jǐn)?shù)據(jù)丟失的情況下仍能準(zhǔn)確進(jìn)行區(qū)段定位。
    關(guān)鍵詞: 配電網(wǎng);暫態(tài)零序電流;區(qū)段定位;短時傅里葉變換;ConvNeXt 模型
    中圖分類號:TM711     文獻(xiàn)標(biāo)識碼:A     文章編號:1007-3175(2024)07-0016-11
 
Research on Fault Segment Location Method of Distribution Network
Based on STFT and Improved ConvNeXt
 
DENG Si-jing1,2, WU Hao1,2, DENG Li-chuan1,2, CAI Yuan1,2
(1 School of Automation and Information Engineering, Sichuan University of Science & Engineering, Yibin 644000, China;
2 Artificial Intelligence Key Laboratory of Sichuan Province, Yibin 644000, China)
 
    Abstract: In the present research methods of intelligent fault diagnosis of distribution lines, there are some problems, such as difficulty to extract fault features fully, weak ability to resist noise interference and poor ability to resist high resistance. In this paper, an improved fault segment location method of ConvNeXt distribution network based on short-time fourier transform (STFT) and transfer learning is proposed.In this method, the amplitude difference of the zero-sequence current at both ends of each feeder is calculated by collecting the zero-sequence current at both ends of each feeder of the distribution network. Then, the amplitude difference of the zero-sequence current of each segment is spliced into a combined signal, and the combined signal is processed by STFT to obtain a time-frequency graph, and the obtained time-frequency graph is divided into a training set and a test set. The simulation results show that the improved ConvNeXt distribution network fault segment location method based on STFT can effectively realize the selection of fault segments under different fault distances, different ground resistances and different initial fault angles, and the method has strong anti-high impedance ability and strong anti-noise interference ability, and can still accurately locate the segment in the case of partial data loss.
    Key words: distribution network; transient zero-sequence current; segment location; short-time fourier transform; ConvNeXt model
 
參考文獻(xiàn)
[1] LI Y, GAO H, DU Q, et al.A review of singlephase-to-ground fault location methods in distribution networks[C]//2011 4th International Conference on Electric Utility Deregulation and Restructuring and Power Technologies(DRPT),2011 :938-943.
[2] YUAN J , WU T , HU Y , et al . Faulty feeder detection based on image recognition of voltage-current waveforms in non-effectively grounded distribution networks[J].International Journal of Electrical Power & Energy Systems,2022,143 :108434.
[3] GUAMAN A, VALENZUELA A.Distribution network reconfiguration applied to multiple faulty branches based on spanning tree and genetic algorithms[J].Energies,2021,14(20) :1-16.
[4] 李衛(wèi)國,許文文,王旭光,等. 基于 DTW 距離搜索的配電網(wǎng)故障區(qū)段定位方法[J]. 電力系統(tǒng)及其自動化學(xué)報,2020,32(6) :80-87.
[5] 賈巍,雷才嘉,葛磊蛟,等. 城市配電網(wǎng)的國內(nèi)外發(fā)展綜述及技術(shù)展望[J] . 電力電容器與無功補償,2020,41(1) :158-168.
[6] 王秋杰,金濤,梅李鵬,等. 基于模型分層的配電網(wǎng)故障診斷方法[J] . 電力自動化設(shè)備,2020,40(1) :73-79.
[7] CHEN K J, HU J, ZHANG Y, et al.Fault location in power distribution systems via deep graph convolutional networks[J].IEEE Journal on Selected Areas in Communications,2020,38(1) :119-131.
[8] 戴志輝,王旭. 基于改進(jìn)阻抗法的有源配電網(wǎng)故障測距算法[J]. 電網(wǎng)技術(shù),2017,41(6) :2027-2034.
[9] 李練兵,孫騰達(dá),曾四鳴,等. 基于多端行波時差的配電網(wǎng)故障定位方法[J] . 電力系統(tǒng)保護與控制,2022,50(3) :140-147.
[10] 劉紅文,曾祥君,柴晨超,等. 配電網(wǎng)柔性接地裝置注入非工頻小信號的接地故障檢測與區(qū)段定位方法[J] .南方電網(wǎng)技術(shù),2022,16(6) :44-53.
[11] 周改云,張國平,馬麗,等. 基于阻抗分析和行波分析的配電網(wǎng)故障定位方法[J] . 電網(wǎng)與清潔能源,2015,31(9) :21-26.
[12] 張文軒,李京,王雪菲. 基于多端行波信息的配電網(wǎng)故障定位改進(jìn)矩陣算法[J] . 水電能源科學(xué),2021,39(6) :194-197.
[13] 束洪春,劉佳露,田鑫萃. 基于故障行波沿線突變和模型匹配的輻射狀配電網(wǎng)故障定位[J]. 電力系統(tǒng)自動化,2020,44(9) :158-163.
[14] 何銳,韓濤,顧澤玉,等. 基于行波折反射特征的單相接地故障區(qū)段定位方法[J] . 智慧電力,2018,46(1) :77-82.
[15] 賈科,李論,宣振文,等. 基于擾動注入的柔性直流配電網(wǎng)主動故障定位及其仿真研究[J]. 電力系統(tǒng)保護與控制,2019,47(4) :99-106.
[16] 王玲,鄧志,馬明,等. 基于狀態(tài)估計殘差比較的配電網(wǎng)故障區(qū)段定位方法[J] . 電力系統(tǒng)保護與控制,2021,49(14) :132-139.
[17] 朱占春,潘宗俊,唐金銳,等. 基于單端暫態(tài)能量譜相似性的配電網(wǎng)故障區(qū)段定位新方法[J]. 電力科學(xué)與技術(shù)學(xué)報,2021,36(2) :180-191.
[18] 吉興全,張朔,張玉敏,等. 基于 IELM 算法的配電網(wǎng)故障區(qū)段定位[J] . 電力系統(tǒng)自動化,2021,45(22) :157-166.
[19] 葉昭暉,王薇薇,張影. 基于深度學(xué)習(xí)的圖像分類方法研究[J]. 信息網(wǎng)絡(luò)安全,2021(S1) :143-146.
[20] 張珂,馮曉晗,郭玉榮,等. 圖像分類的深度卷積神經(jīng)網(wǎng)絡(luò)模型綜述[J] . 中國圖象圖形學(xué)報,2021,26(10) :2305-2325.
[21] 季長清,高志勇,秦靜,等. 基于卷積神經(jīng)網(wǎng)絡(luò)的圖像分類算法綜述[J] . 計算機應(yīng)用,2022,42(4) :1044-1049.
[22] HOU S Z, GUO W, WANG Z Q, et al.A wavelet AlexNet network-based method for fault segment location in distribution networks[J].Electrical Measurement and Instrumentation,2022,59(3) :46-57.
[23] ZHU H Y, JIA Z T, LI Q.CNN-based fault zone location in distribution network[J].Hydropower Energy Science,2021,39(7) :188-191.
[24] YANG J, CHEN Y, FENG B, et al.Research on CNNidentification method of single-phase ground fault type in distribution network[J].Journal of Chongqing University of Technology(Natural Sciences),2022,36(8) :236-245.
[25] 唐志國,曹智,何寧輝. 卷積神經(jīng)網(wǎng)絡(luò)遷移學(xué)習(xí)在局部放電類型診斷中的應(yīng)用[J] . 高壓電器,2022,58(4) :158-164.
[26] 朱革蘭,李松奕,蘭金晨,等. 基于零序特征量的配電網(wǎng)接地故障區(qū)段定位方法[J] . 電力自動化設(shè)備,2021,41(1) :34-40.
[27] 李婷,付德義,薛揚. 基于 AE 與 STFT 的變槳軸承裂紋診斷研究[J] . 振動、測試與診斷,2021,41(2) :299-303.
[28] LIU Z, MAO H, WU C Y, et al.A convnet for the 2020s[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,2022 :11976-11986.