應用蟻拓倒傳遞神經網路發展風力發電機轉子磁極損壞檢測技術

本文利用蟻拓演算法，最佳化倒傳遞神經網路模型權重，稱為蟻拓演算法之倒傳遞神經網路模型(ABPNN)，利用本ABPNN模型在搜尋最佳權重路徑的優越性，分析風力發電機各種異常狀況。研究首先，分別以放電加工及高溫噴槍製作 (1) 單一轉子磁極穿孔、(2) 相鄰轉子磁極穿孔、(3) 轉子磁極自然斷裂，及(4) 轉子磁極燒毀等四種風力發電機故障試品，並配合一組正常發電機試品，量測在不同負載（25、50及75歐姆）及轉速（200、250及300轉），各研究試品之發電機輸出電流訊號。其次，分別以S轉換及多重解析對電流訊號進行頻譜分析，以擷取發電機訊號特徵，並以本文提出之ABPNN模型，選取最適權重，以提升神經網路對於發電機異常訊號特徵辨識能力。最後，本研究在發電機電流訊號中分別加入20dB及10dB雜訊，以測試本文所提模型抗雜能力。測試結果顯示，本研究所提模型，於雜訊20dB時辨識率分別可達88%，雜訊10dB時辨識率亦可達80%。

關鍵字

轉子磁極損壞檢測技術；蟻拓神經網路

並列摘要

This study utilizes ant colony optimization to obtain optimal weight of back propagation neural network, which is called ABPNN. ABPNN model is used to search for the superiority of the optimal weight path to analyze different abnormalities of wind generator. First, manufacture four testing type of the wind generator by electrical discharge machining (EDM) and high temperature spray gun,which are respectively (1) one broken rotor bar (2) two adjacent broken rotor bars (3) rotor bar fracture and (4) burn down, moreover assort one healthy generator testing type, then measure the fault current signals of the generator under different loads (25, 50 and 75 ohm) and speeds (200, 250 and 300 turn). Second, the S-transform（ST）and multi-resolution analysis（MRA） are used to analyze spectrum of the current signals, and then extract the features of spectrum. In addition, this study proposes a model that selects optimal weight of back propagation network by ant colony optimization. Through this model, the recognition ability of abnormal generator signals can be upgraded effectively. Finally, 20dB and 10dB signal-to-noise ratio（SNR）are mixed into the generator current signal to verify the robustness of model respectively. The results show that when 20dB noise is mixed, the recognition accuracy reaches 88%; whereas the recognition accuracy also reaches 80% when 10dB noise is mixed.

並列關鍵字

ABPNN ； Wind Turbine Rotor Pole ； Damage Detection

參考文獻

[4] F. C. Trutt, J. Sottile, and J. L. Kohler, “Detection of AC machine winding deterioration using electrically excited vibrations,” IEEE Transactions on Industry Applications, vol. 37, no. 1, pp. 10-14, 2001.

[5] G. C. Stone, C. V. Maughan, D. Nelson, and R. R. Schultz, “Impact of slot discharges and vibration sparking on stator winding life in large generators,” IEEE Electrical Insulation Magazine, vol. 24, pp. 14-21, 2008.

[6] W. L. Cameron, “Precise expression relating the Fourier transform of a continuous signal to the fast Fourier transform of signal samples,” IEEE Transactions on Signal Processing Society, vol. 43, no. 12, pp. 2811-2821, 1995.

[7] I. S. Reed, and T. K. Truong, “A new hybrid algorithm for computing a Fast Discrete Fourier transform,” IEEE Transactions on Computer Society, vol. 28, no. 7, pp. 487-492, 1979.

[8] Y. Liu, Z. Nie, and Q. H. Liu, “DIFFT: A fast and accurate algorithm for Fourier transform integrals of discontinuous functions,” IEEE Microwave and Wireless Components Letters, vol. 18, no. 11, pp. 716-718, 2009.

國際替代計量

應用蟻拓倒傳遞神經網路發展風力發電機轉子磁極損壞檢測技術

全文下載

主題瀏覽