｜多工作狀況下風機健康監控的方法

摘要：近年來，在再生能源技術的研究與需求刺激之下，風力發電場的蓬勃發展正向全球擴散。相對於投入風力發電機的設計，製造和安裝，另一項風能發電領域的關鍵議題卻尚未獲得世人關注，那就是維持風力發電機組的運轉可靠性。如何在動態環境下，建立強健的風力發電機組狀態監測系統是當今發展風能發電最主要的挑戰。

美國辛辛那提大學智慧維護系統中心（IMS）專注於預測與健康管理（Prognostics and Health Management, PHM）的工具和演算法，並應用於風力發電機及其關鍵零組件。本文介紹一風力發電機組健康評估系統，監測動態環境下風力發電機的實際運轉情形。數據由監控與數據採集系統（SCADA）採集和處理之後，同時利用工作狀況參數進行分類和識別。根據正常狀態下的資料建立特定變數於各工作狀況下的健康基線模型後，即可為後續監測資料，依據不同工作條件，進行有效的風力發電機組健康評估。

Abstract: In recent years, the spurring demand for the adoption and development of renewable energy has led to a booming global proliferation of wind farms.  Compared with the design, manufacturing and installation of wind turbines, a key aspect in wind energy generation that has not received an equitable amount of research attention is wind turbine reliability. A major challenge in the development of a robust condition-based monitoring system is the dynamic conditions that wind turbines are subjected to.

Center for Intelligent Maintenance Systems (IMS) focuses on applying Prognostics and Health Management (PHM) tools and algorithms to wind turbines and its critical components. This article presents a systematic framework for system-level health assessment of wind turbines, that addresses the issue of the dynamic operating conditions that a wind turbine experiences. Data from a wind turbine SCADA (Supervisory Control and Data Acquisition) system are acquired and processed while working condition parameters are utilized to classify and identify the working regimes. Measurements from selected variables that are indicative of the turbine status when a turbine is known in a nominal healthy condition are used to build a baseline model in each operating regime. New data from the monitored wind turbine system is then input into a regime identification algorithm that outputs the operating regime based on the working conditions. Accordingly the turbine health condition is estimated by comparing the data from the monitored system with its corresponding baseline.

關鍵詞：風力發電機、動態運轉、預測與健康管理

Keywords：Wind Turbine, Dynamic Operation, Prognostics & Health Management (PHM)

Introduction
Wind power is a growing renewable source of electrical energy globally. Within the last decade, the American wind industry has emerged as the world leader in installed energy generation capacity with an impressive and sustained annual growth rate of more than 30% [1]. According to the Global Wind Energy Council [2], the United States became the world leader in the fast-growing wind energy generation industry by the end of 2008. The Global Wind Energy Council reports [2] that the wind energy generating capacity installed in the United States at the end of 2008 was approximately 21% of the total world generation. Wind turbine reliability is essential for continuous capture and efficient generation of the maximum power output. Enhancing the availability, reliability, and lifetime of wind turbines is critical and requires the exploitation of fundamental research in condition-based monitoring of core wind turbine system components. Such research would be instrumental in determining power generation efficiency and offer support for the creation of advanced wind turbine designs.

Wind turbine systems are increasing in technical complexity, and are tasked with operating in dynamic environments （e.g. continuous changes in wind speed, direction, and other turbine operating parameters）. Thus, sustaining the reliability of such systems is becoming a more complex and challenging, although critical, task. These conditions necessitate the development of a scientific understanding of the contributing factors to the dynamic mechanisms at work in wind turbine systems, the effect of these factors on the health of the turbine components and the underlying relationship of component health on the resulting energy generation efficiency.  The three overall challenges and unmet research needs in the prognostics and health management of wind turbine systems can be summarized as follows:

（1）The main deployment of collected data from wind turbine in most applications is basic monitoring, and inference of the actual operating health condition is lacked.
（2）Drivetrain components of a wind turbine are constantly working under highly dynamic loads and more susceptible to failure.
（3）Health degradation of critical drivetrain components may impact wind turbine power generation capability across the spectrum of wind speed input.

The NSF I/UCRC for Intelligent Maintenance Systems （IMS） is focused on smart prognostics and tether-free technologies to enable products and systems to achieve near-zero downtime performance. Currently, the Center is supported by over 40 global companies including P&G, GE Aviation, Goodyear, Omron, National Instrument, etc. The Center has developed the Watchdog Agent^®, a set of computational prognostic algorithms and a software toolbox for predicting machine performance and ultimately identifying a failure point. The Watchdog Agent^® is capable of extracting multiple inputs （features） from components or systems, fuses these features and quantitatively evaluates the performance degradation. （visit Center for IMS at www.imscenter.net）

The involvement with various research project activities on rotary machinery （spindles, machine tools, alternators, bearings, etc.） has driven the IMS Center to apply its PHM tools and algorithms to wind turbines and its critical components （Figure 1）. The main challenge is to research a new paradigm that deviates from the traditional use of vibration data, to diagnose and predict faults. IMS collaborates with Industrial Technology Research Institute （ITRI）, Taiwan and its wind farm partners to apply the framework of assessing turbine performance degradation under dynamic operating conditions with SCADA data on large size onshore wind turbine. The preliminary results are presented in this article, as well as demonstrated at http://windturbinephm.com.