MORIKAWA NARUSUE LABORATORY

風車は内部機器が故障すると長期間停止させる必要があり，大きな損害が生じる．そのため，故障予兆を検知し，適切な延命処置を施すことが必須である．本研究では，風車の健全性評価を低コストに実現するため，Supervisory Control And Data Acquisition (SCADA) と呼ばれる風車の監視制御用途で設置される標準装置のデータを用いた健全性評価手法を検討している．本手法では，内部機器を個別に解析するのではなく，風車負荷のかかる稼働モードを選別することで，健全性評価に適した稼働状態を抽出する．具体的には，SCADAが収集する主軸回転数，ナセル方向，ピッチ角度，風向，風速，外気温，ギアボックス温度などのデータに対してフィルタリング処理を行って特徴量を抽出し，正常稼働状態の分類器を構築する．そして，評価対象のSCADAデータをこの分類器に入力し，正常な稼働状態と比較することで，風車の状態悪化を検知する．本手法に関し，多サイトのデータで健全性評価の有効性を確認している．

Failures of internal components of wind turbines lead to prolonged downtime, and maintenance cost will increase. It is necessary to detect incipient faults and improve availability by reducing unscheduled maintenance. In this study, in order to realize system health monitoring of wind turbine by low cost, we are developing a method for evaluating the health of the wind turbines using the data of standard equipment installed for supervisory control of almost all wind turbines, called Supervisory Control And Data Acquisition (SCADA). The proposed method extracts the operating condition suitable for evaluating the health of the wind turbines by exploiting the operation mode data which indicates excessive loading condition on wind turbines, instead of focusing on individual internal components. Specifically, we perform the filtering based on the domain knowledge of time series data such as rotational speed of the main shaft, nacelle direction, pitch angle, wind direction, outside temperature and gearbox temperature collected by SCADA system. We then build a classifier for determining the normal operating condition of the wind turbines. Consequently, our classifier can assess the deterioration of the wind turbines by utilizing SCADA data and comparing with the trained normal condition. The effectiveness of proposed method is evaluated with many wind turbine site data.