Fault Automatic Identification Of Wind Turbine Generator Based On Intelligent Video Image Analysis

With the increasing attention of clean and renewable energy all over the world, the wind power industry in China has developed rapidly in recent years. Generlly, wind turbine equipments worth expensively and failure rate is high.And wind farms are generally located in the district which is remote and harsh environments. And operation and maintenance personnel is less.So it needs the remote real-time monitoring and automatic identification system in the wind farms. Using the video surveillance technology and image recognition technology, the establishment of remote real-time monitoring identification system can be automated unattended and solve the problems of the big labor intensity.All the same time,it improve operation and maintenance management.This paper used the monitoring image of the tower entrance, cabin, gearbox bottom, cable protection cover and other parts of the wind turbine obtained from the video surveillance system to research how to obtain abnormal status information by image analysis techniques and achieve automatic alarm. The main research work and results are as follows:(1)For the detection of tower invasion, used different image method to automatically identify.That is, update background model in real time based on Surendra method, and get a motion area based on background subtraction method, then get a more accurate image feature of the movement area, finally distinguish the abnormal state of the normal state based on a threshold determination method and alarm automaticly when it is an abnormal state.(2)For the detection of cabin flame, firstly get video images from the camera, then use the color feature to segment image area having suspected a small flame, and then use the growth of flames to judge initially, and finally use the texture features to verify small flame.(3)For the detection of gearbox leakage, we used a gearbox oil leak detection method based on color characteristics.That is, extract H-S color histograms and compare a series of test pictures, then large changes in dust-color can judge the abnormal oil leakage.(4)For the crack detection in the cable protection cover, because of the low contrast, it joins a series of pre-processing to improve image quality and make full preparations for crack image detection and recognition.Then judge the crack based on geometric shapes.