Research On Model-based Fault Prediction Of Pitch System For Wind Turbines

With the rapid development of wind power utilities, variable pitch wind turbine is replacing the fixed pitch wind turbine gradually, becoming the mainstream models of wind turbines. The pitch system plays a crucial role in wind turbines, which According to the size of the wind speed and power changes at any time, it can adjust the blade pitch angle to the maximize wind energy conversion and make output power a constant. At the same time, the frequent fine-tuning of variable pitch system has also exacerbated its wear and aging. Compared to other parts of the wind turbine, the possibility of the pitch system failure increases greatly. In addition, pitch system is installed in the hub, which is tens of meters from the ground and the internal space is very narrow, making pitch systems very difficult to maintain and also increasing the manpower, material and other input costs. Moreover, if you cannot discover pitch system failure and take reasonable measures to deal with it immediately, this could trigger a catastrophic incident to the whole wind turbines. So the fault prediction of pitch system is meaningful for prolonging the stable operation time and reducing maintenance costs.This paper presents a model-based fault prediction approach, which will be used to fault prediction for two types of variable pitch system of wind turbine.(1) The paper adopts the model-based fault prediction of electric pitch system for wind turbines. Firstly, the electric pitch system mathematical model is established according to control processes and structures of the electric pitch system, which can reflect the dynamic characteristics of electric pitch system. Then in order to form a complete closed-loop system model for electric pitch system failure prediction, the models of other key components of wind turbine are established and a residual is formed by the wind turbine model and the actual system running in parallel. Lastly, the residual root mean square(RMS) is used as an evaluation function to deal with residual errors and extract fault features. Taking the system delays and false alarm rate into consideration, the maximum value of the residual variation which was caused by noise, disturbances and other interference factors is used as failure prediction threshold to realize the fault prediction of electric pitch system.(2) The model-based approach combined with adaptive threshold is used to the hydraulic pitch system fault prediction of wind turbines. The hydraulic pitch system works completely different from the electric pitch system and the structure is more complex. This paper established mathematical model of the hydraulic pitch system which can reflect the actual dynamics characteristics when making it running in parallel with the fault model which was established by fault modeling theory. Then the fault prediction of hydraulic pitch system is realized by the adaptive threshold. The adaptive threshold reduces model errors and environmental factors relating to the wind turbine hydraulic system failure prediction and improves the prediction accuracy of the results.On MATLAB experimental platform, this paper completes the implementation of the above algorithms. The fault prediction of electric pitch system, pitch angle sensor, electro-hydraulic proportional directional valves and hydraulic cylinders are simulated and verified respectively. All the algorithms achieve good simulation results. The results show that the proposed methods are correct and effective.