由於太陽能光伏（PV）發電系統必須安裝在室外，惡劣的環境很容易導致故障，例如強風損壞、電弧故障等。電弧故障可能導致火災等重大安全事故。連續電弧故障引起的高溫沒有及時發現和解決。目前，由於檢測框架複雜、執行時間長，很多檢測效果好的演算法難以付諸實踐。為了解決這個問題，本技術提出基於經驗模態分解（EMD）和門迴圈單元神經網路（GRU-NN）的智慧電弧故障檢測演算法。該演算法利用EMD從電流信號中提取故障資訊，然後根據模態順序對EMD中各模態的統計指標進行排序。此外，GRU-NN用於捕捉不同模式之間的特徵和變化趨勢，實現電弧故障檢測。實驗結果表明，在所有檢測條件下，檢測準確率均在 98.7% 以上。此外，本研究還提出了一種線上更新方法，以保證所提演算法的適應性。結合這種線上更新方法，該方案可以快速修改模型並保證電弧故障識別的準確性，即使對於不同的光伏電站也是如此。線上更新能力的表現也將通過臺灣和中國光伏電站的實驗來驗證。
Because solar photovoltaic (PV) power generation system must be outdoor installed, the harsh environment is easy to lead to failure, e.g., strong wind damage, arc faults, etc. The arc fault may lead to a major safety accident such as fire if the high temperature caused by the continuous arc fault is not identified and solved in time. At present, many algorithms with good detection effect are difficult to put into practice because of complex detection framework with long execution time. In order to solve this problem, an intelligent arc-fault detection algorithm based on the empirical mode decomposition (EMD) and the gate recurrent unit neural network (GRU-NN) is investigated in this study. The proposed algorithm uses the EMD to extract the fault information from current signals, and then sequences the statistical indexes of each mode from the EMD according to modal orders. Moreover, the GRU-NN is used to capture the features and variation trends among different modes, and realize the arc-fault detection. As for experimental results, the detection accuracy is over 98.7% under all examined conditions. In addition, an online updating method is also proposed in this study to ensure the adaptability of the proposed algorithm. Combined with this online-updating method, the proposed scheme could quickly modify the model and ensure the accuracy of the arc fault identification, even for different PV stations. The performance of online updating ability will be also verified by experiments in Taiwan and China PV stations.
|
專利行銷平台
