






摘 要:隨著電力電子技術的深入發展,電力系統的智能化程度不斷提高。針對智能電力系統輸配電環節的變壓器故障檢測問題,提出一種基于神經網絡的智能化檢測方法。采用回聲狀態網絡作為神經網絡的基本框架,充分發揮其隱含層神經元數量大、學習能力強、非線性逼近能力強的優勢,增強檢測結果的準確性。在輸入部分,配置漏油故障、油位故障、閃絡故障、穢物故障、散熱故障、溫差故障、開關老化故障、繞組漏電故障、鐵芯絕緣故障和繞組變形故障10類故障數據,對神經網絡進行訓練。同時,完成智能算法承載軟件的客戶端和服務器端設計,并配置客戶端的數據采集系統。實驗結果顯示,在1 000個數據作為訓練集合的情況下,神經網絡達到更好的收斂。以此作為故障檢測的工具,達到98.1%的故障檢測準確率。
關鍵詞:電力電子技術;電力系統;智能化;故障檢測;神經網絡
中圖分類號:TM507 文獻標志碼:A 文章編號:2095-2945(2024)23-0032-04
Abstract: With the in-depth development of power electronics technology, the intelligent degree of power system is increasing. Aiming at the problem of transformer fault detection in the transmission and distribution link of intelligent power system, an intelligent detection method based on neural network is proposed. The echo state network is used as the basic framework of the neural network, which gives full play to the advantages of a large number of neurons in the hidden layer, strong learning ability and strong nonlinear approximation ability to enhance the accuracy of the detection results. In the input part, ten kinds of fault data such as leakage fault, oil level fault, flashover fault, foul fault, heat dissipation fault, temperature difference fault, switch aging fault, winding leakage fault, core insulation fault and winding deformation fault are configured to train the neural network. At the same time, the client-side and server-side design of the intelligent algorithm bearer software is completed, and the client-side data acquisition system is configured. The experimental results show that the neural network achieves better convergence when 1000 data are used as training sets. Using this as a tool for fault detection, 98.1% accuracy of fault detection is obtained.
Keywords: power electronics technology; power system; intelligence; fault detection; neural network
電氣技術革命的影響持續而深遠,電力系統是目前諸多領域應用的主流系統。隨著信息技術的快速發展,電力系統也進入了智能化電力系統的全新時代[1]。相比于傳統電力系統,智能化電力系統表現出無人值守、全自動化、高效率和節能等諸多優點,已經成為電力行業未來的發展趨勢[2]。在智能電力系統的實現過程中,電力電子技術起到了非常重要的作用,在諸多方面為智能電力系統建設提供技術支撐,例如智能電力系統中的輸配電技術、智能電力系統中的發電儲能技術、智能電力系統中的電網優化技術[3]。……