Research On Fault Diagnosis Method Of Wind Turbine Hydraulic Variable Pitch Systems

The failures of wind turbine happen frequently. As an important part of the wind turbine, hydraulic pitch system plays a vital role in wind energy utilization, stability, and reliability. So it has a very important significance to do fault diagnosis for the wind turbine hydraulic pitch system. First of all, the paper has overviewed fault diagnosis research background and analyzed the basic idea and characteristics for various fault diagnosis methods. Due to the traditional fault diagnosis method has its own limitations, the paper has introduced power bond graph theory and fuzzy neural Petri net theory.Firstly, for the wind turbine hydraulic pitch system, a new type fault diagnosis method of fault tree based on bond graphic model is proposed. The method based on bond graph model generates time causal diagram, and establishes system fault tree to do fault diagnosis. Then compare the fault reasoning with Matlab/Simulink simulation module results to verify the correctness of this diagnostic method.Moreover, in allusion to uncertainties and fuzziness of diagnosis information, a fault diagnosis method based on fuzzy Petri net have been proposed for the wind turbine hydraulic pitch system. By analyzing immediately reachable set, reachable set and adjacent library set of faults library, reverse parallel reasoning is carried out on the basis of fuzzy membership so as to search and confirm the paths and reliabilities of faults. Then, a fault diagnosis system of wind turbine generators had been developed on the Lab VIEW virtual instrument platform. The simulation cases show that fault diagnosis system of fuzzy Petri net with its strong vivacity can express the transition transfer process of hydraulic variable pitch system clearly, locate the causes and positions of faults exactly and calculate the reliabilities of faults library promptly.Finally, in order to make the fuzzy Petri nets with self-learning and adaptive ability, the paper introduces the concept of artificial neural network and proposes a fault diagnosis method based on self-learning fuzzy neural Petri net. This method combines the advantages of fuzzy Petri nets and self-learning ability of neural network to do fault diagnosis and reasoning. It can not only clearly express the association among the faults and the fault dynamic propagation characteristics, but also has the ability of neural network self-learning to learn new expert experience. The method breaks the defects which the fuzzy Petri net weights is needed to pre-set. It has the characteristics of flexibility and intelligence in a certain degree.