風力發電機監控平台之演算法研究

Study on the Algorithms of the Wind Turbine Health Monitoring Platform

高睦修(Mu-Hsiu Kao)

指導教授 ： 蔡進發

國立臺灣大學/工學院/工程科學及海洋工程學研究所/碩士(2018年)

https://doi.org/10.6342/NTU201803665

摘要

摘要
本研究使用了BIRCH演算法(Balanced Iterative Reducing and Clustering Using Hierarchies) 擅長多分群的優點改進了傳統上DBSCAN(Density-based spatial clustering of applications with noise)無法處理雜訊密度過高的問題，並可將其套用在風機健康監控平台的前處理來過濾訓練資料。分別使用高斯混合模型(Gaussian Mixture Model, GMM)和自組織映射(Self-Organizing Map ,SOM)來近似訓練資料，再根據近似出的模型或是神經元，來與測試資料計算信心值。
在計算信心值的部分，高斯混合模型針對中屯一號風機所計算出的信心值有效的反映風機的健康狀況，本文另文提出一套利用自組織映射所算出之神經元權重為依據的信心值計算方式，在面對情形較為特殊之測試資料時正確性較高斯混合模型的信心值高，並且也能表現整體風場年度信心值之變化趨勢以及對單一風力發電機進行預測。

關鍵詞：風機健康監控平台、高斯混合模型、自組織映射、信心值

關鍵字

風機健康監控平台 ； 高斯混合模型 ； 自組織映射 ； 信心值

並列摘要

Abstract
This study takes the advantage of the BIRCH algorithm (Balanced Iterative Reducing and Clustering Using Hierarchies) which is good at multi-grouping and solves the problem of traditional DBSCAN (Density-based spatial clustering of applications with noise) which can’t handle excessive noise density to filter training data of pre-processing on wind turbine health monitoring platform. This study uses Gaussian Mixture Model and Self-Organizing Map to approximate training data, then calculates the confidence value with the approximated model or neurons. The confidence value calculated by Gaussian Mixture Model of Jung-Tuen No. 1 Wind turbine effectively reflects the health condition of the wind turbine. This study proposes a calculation model of confidence value which based on the weight of the neurons calculated by the self-Organizing Map. The calculated results show that the confidence value calculated by SOM is better than the confidence value calculated by Gaussian mixture model. It also shows the change trends of confidence value of multiwindtrubine. And the prediction of single windturbine.

Key word: Wind Turbine Health Monitoring Platform, Gaussian Mixture Model, Self-Organizing Map, Confidence Value

並列關鍵字

Wind Turbine Health Monitoring Platform ； Gaussian Mixture Model ； Self-Organizing Map ； Confidence Value

參考文獻

1. REN21, Renewables 2018 Global Status Report. 2018, Paris: REN21 Secretariat.

2. Lee, J., B. Bagheri, and H.-A. Kao, A cyber-physical systems architecture for industry 4.0-based manufacturing systems. Manufacturing Letters, 2015. 3: p. 18-23.

3. Song, B. and J. Lee, Framework of designing an adaptive and multi-regime prognostics and health management for wind turbine reliability and efficiency improvement. Framework, 2013. 4(2): p. 142-149.

4. Niknam, S.A. and R. Sawhney, A framework for prognostic-based real-time life extension.

5. Ciang, C.C., J.-R. Lee, and H.-J. Bang, Structural health monitoring for a wind turbine system: a review of damage detection methods. Measurement Science and Technology, 2008. 19(12): p. 122001.