應用同儕比較偵測風機異常之研究

Abstract

本研究以同儕比較的概念來建立風場中風機故障檢測與性能評估之流程，流程分為風場分群分析與風場風機性能評估與異常偵測兩部分進行。其中風場分群分析利用每部風機量測的風速與風向資料，透過自組織拓譜映射圖，圖像矩陣法及高斯混合模型三個方法來建立風況模型後以信心值公式量測彼此間的相似性，再利用聚合階層式分群法分群並比較三個方法的優劣。根據分群結果再分成兩類，第一類為有與其他風機形成同儕關係的風機，針對同群内風機的功率、轉子轉速、發電機轉速、葉片旋角與偏航誤差之資料，透過高斯混合模型與信心值做相似性的測量，並逐日比對群內不同風機間行為是否相異，第二類為未與其他風機形成同儕關係的風機，則針對風向較特別的風機檢查其風向計是否有異常。 本研究使用台電麥寮23部風機進行資料分析，依風況分群結果風機被分成三群，群集一為風向較偏東北的群集、群集二風向較偏西北的群集。未與其他風機建立同儕關係的風機即為獨立群。群集一中，ML15風機表現最佳，ML9風機異常天數最多，且ML9、ML10、ML12風機性能明顯比ML15風機差，應盡早做維護；群集二中ML6風機表現最佳，ML13風機異常天數最多，且ML13風機性能明顯比ML6風機差，也應盡早做維護；獨立群方面，ML3、ML8、ML11、ML21、ML22風機的風向有出現異常情況，其中以ML8風機最嚴重應盡早進行檢查轉向系統及風向計，ML19風機則是整體風向較為特殊因此被歸類到獨立群。

The peer comparison method was used to establish a process of wind turbines fault detection and performance assessment within a wind farm. This process can be divided into two parts, which are wind turbines clustering and fault detection. The wind speed and wind direction were used for the wind turbine clustering by Gaussian mixture model, Self–Organizing map and Image Matrix Method to measure the similarity of wind turbines with the confidence value. Then, Hierarchical Clustering was used to group the wind turbines which have similar wind conditions. The clustering results of the three confidence value calculation methods were compared and the advantages of these three methods were discussed. According to the clustering result, the turbines were divided into two types. The first type is the turbine that has the similar wind condition within the group. Then the power, rotor speed, generator speed, pitch angle of blade and yawing misalignment were analyzed by Gaussian mixture model day by day to find the abnormal wind turbine. The second type is the wind turbine that does not have similar wind condition with other turbines, The wind cup performance will be discussed. This SCADA data of the 23 wind turbines of the Mai-liao wind farm of the Taipower were used in this study. According to the clustering results, the wind turbines were divided into three groups by using the wind conditions. The group 1 has the wind direction from the northeast, and the group 2 has the wind direction from the northwest. The group 3 is an independent group which the wind condition of the wind turbines have less similarity with other two groups. In group 1, turbine ML15 has the best performance, turbine ML9 has the most abnormal days, and turbine ML9, ML10, ML12 performance are worse than that of turbine ML15, which should do the maintenance as soon as possible. In group 2, turbine ML6 has the best performance, turbine ML13 has the most abnormal days, and turbine ML13 performance is worse than that of the turbine ML6, which should do the maintenance as soon as possible. In independent group, turbine ML3, ML8, ML11, ML21, ML22 wind directions have contrary wind direction with group 1 and 2. The turbine ML8 has the most serious problem, which should do the maintenance as soon as possible.