郭慧瑩 王毅
摘 ?要: 針對基于DGA的變壓器故障診斷方法在實際操作中存在的不足,提出兩種解決方案:基于粒子群優化支持向量機的變壓器故障診斷、基于差分進化支持向量機的變壓器故障診斷。通過分析兩種方案的算法原理建立支持向量機的變壓器故障診斷模型,從而完成參數的優化,對得到的最優參數進行驗證,獲取最優的支持向量機模型。在Matlab軟件平臺上進行仿真實驗,結果證明,采用基于粒子群優化支持向量機的變壓器故障診斷結果獲取的變壓器故障診斷率較高;基于差分進化支持向量機的變壓器故障診斷方法的誤判率較低,全局尋優能力較好,相比于粒子群優化算法,差分進化支持向量機的優化精度更高。
關鍵詞: DGA; 支持向量機; 變壓器; 故障診斷; 參數優化; SVM模型
中圖分類號: TN99?34 ? ? ? ? ? ? ? ? ? ? ? ?文獻標識碼: A ? ? ? ? ? ? ? ? ? ? ? ? ? ? 文章編號: 1004?373X(2019)19?0154?05
Abstract: In view of the shortcomings of DGA?based transformer fault diagnosis methods in practical operation, two solutions are proposed, that is, transformer fault diagnosis based on particle swarm optimization support vector machine and transformer fault diagnosis based on differential evolution support vector machine. The transformer fault diagnosis model based on support vector machine is established by analyzing the algorithm principles of the two solutions, thus completing the parameters optimization, verifying the optimal parameters and obtaining the optimal support vector machine model. The simulation experiment was carried out on Matlab software platform. The results prove that the fault diagnosis rate of transformer based on particle swarm optimization support vector machine is higher; the fault diagnosis method based on differential evolution support vector machine has lower error rate and better global optimization ability. In comparison with particle swarm optimization, the differential evolution support vector machine has better global optimization ability and higher optimization accuracy.
Keywords: DGA; support vector machine; transformer; fault diagnosis; parameter optimization; SVM model
電力變壓器是電力系統的重要組成部分,其運行狀態的好壞關系到電力系統的可靠性,一旦電力變壓器出現故障,將會造成巨大的經濟損失。變壓器出現故障后采取的診斷方式多種多樣,油中溶解氣體分析(DGA)法是檢測變壓器出現故障后最有效的手段之一,可以及時發現變壓器中存在的內部故障,同時在對變壓器進行維護的過程中還可以排除變壓器中存在的故障隱患。當變壓器出現故障后,通過產生的氣體使用不同的測試法進行故障分析診斷,特征氣體法、羅杰斯比值法以及改良三比值法是基于DGA的變壓器故障的主要診斷方法,在進行故障診斷時存在編碼盲點的問題,不能同時對多種故障進行診斷。……