Researches On Wind Turbine Health State Assessment Based On SCADA Data

With the increasing size and installed capacity of wind turbines,the operation and maintenance cost of wind turbines is also rising.In order to realize the steady and rapid development of wind turbine,reduce cost and increase efficiency,the fault prediction and health assessment technology of wind turbine has become the key technical means to solve these problems.Based on the data collected by the supervisory control and data acquisition(SCADA)system of wind turbine,combined with the operation mechanism of wind turbine,this paper carries out relevant research on fault early warning and health status assessment of wind turbine,mainly including three health status assessment methods under different monitoring levels: unit component level,whole machine level and wind farm level,The main work and innovations are as follows:First,in view of the single dimension of information fusion in the current wind turbine bearing health status assessment method,a wind turbine front bearing fault early warning model based on one-dimensional convolution neural network and long short-term memory network is proposed: in terms of state characteristic parameter selection,state characteristics are selected based on state parameter correlation and wind turbine operation mechanism,To ensure the high correlation and independence between input parameters and prediction parameters,and improve the modeling efficiency while simplifying the model structure;In the aspect of network structure construction,the one-dimensional convolutional neural network and long short-term memory network are creatively integrated to capture not only the anomaly of the coupling relationship between the observed variables on the spatial scale,but also the anomaly of their development trend on the time scale.The experimental results show that the early warning model can monitor the bearing anomaly three months in advance,which proves the effectiveness of the proposed method.Second,aiming at the problems of imperfect evaluation index and inflexible index weight in the current health status evaluation method based on the aging degree of the whole machine,a wind turbine aging evaluation model based on variable weight theory is proposed.Firstly,according to the physical structure and key performance indexes of the wind turbine,the three evaluation dimensions of unit output,engine room vibration and large component temperature are defined,and then the corresponding evaluation indexes are set under each dimension to calculate the aging degree of each evaluation index based on SCADA data,and then the evaluation results of each index are combined with the method of variable weight theory and threshold evaluation to obtain the aging degree of the whole machine.The proposed aging evaluation method is verified by using the SCADA monitoring data of 10 units of a wind farm in Shanxi,China.The experimental results show that this method can obtain reliable aging evaluation results of wind turbines.Finally,aiming at the lack of research on wind farm level outlier detection at home and abroad,a wind farm outlier detection model based on PCA and statistical threshold is proposed.Firstly,according to the key evaluation indexes of unit operation performance,the wind turbine state characteristics participating in outlier detection are selected,and then PCA is used to reduce the dimension of the state characteristic curve to make the data show the most separable form.Finally,the judgment index and threshold of unit outlier are set based on statistical method and sample variation coefficient.The operation data of 33 units with rated power of 1.5MW in a wind farm in Pakistan are used to test the proposed method.The experimental results show that the method can effectively detect the outlier units in the wind farm and has good practical value.