Wind Turbine Pitch System Faults Diagnosis And Prognosis Based On ANFIS

In the background of energy depletion and environmental degradation, the wind power industry of China has achieved great progress, and becomes China’s third-largest power, meanwhile, China has turned into the world’s superpower in wind power.With the fast growing of wind power industry, the fast improving of manufacturing level, the power increase of single wind turbine and the increasing complexity of the operating environment, as one of the core control technologies, the variable pitch system’s faults had become the primary reason which caused the wind turbine downtime. According to the statistics from part of the wind farm, the percent of normal downtime was only 0.24% of total downtime, while the downtime caused by pitch system faults accounted for 52.57%. In conclusion, pitch system faults have caused huge economic losses.A new faults diagnosis and faults prognosis method was proposed in the thesis using Adaptive Neuro-Fuzzy Inference System(ANFIS). ANFIS is very suitable for wind turbine pitch system faults diagnosis modeling, because it combines the advantages of self-learning ability of artifical neural network and the ability to explain complex system of fuzzy system, it has stronger self-learning ability and adaptability than the artifical neural network, and overcomes the shortcomings of fuzzy system which requires strong expertise. Then, by analyzing the basic structure and working principle of wind turbine pitch system, the thesis analyzed the pitch system failure modes, which laid foundation for the faults diagnosis and prognosis.Based on the Supervisory Control and Data Acquisition(SCADA) systems’ data of 1.5MW running grid-connected wind turbines of some wind farm in Hebei Province, the ANFIS was used to construct the pitch angle faults diagnosis model, the motor speed faults diagnosis model and the power output faults diagnosis model. The accuracy and effectiveness of the models had been demonstrated using five metrics:(1) the relative root mean square error;(2) the diagnostic accuracy;(3) the error rate;(4) the troubleshooting success rate;(5) the diagnosis precision for faults. On the basis of the proposed faults diagnosis models, the pitch angle faults prognosis model, the motor speed faults prognosis model and the power output faults prognosis model were constructed by using the historical data and current time data to forecast the wind turbine state in the next 10 minutes. The results of the simulation have demonstrated that the proposed approach has strong protential for wind turbine pitch system faults prognosis.