Fault diagnosis of hydroelectric unit based on AFSA-BP hybrid algorithm

• Published in: 2014 ISFMFE - 6th International Symposium on Fluid Machinery and Fluid Engineering p. 008
• Main Authors: Qiao, Liangliang, Chen, Tao, Chen, Qijuan
• Format: Conference Proceeding
• Language: English
• Published: Stevenage, UK IET 2014; The Institution of Engineering & Technology
• Subjects: Civil and mechanical engineering computing; Mechanical engineering applications of IT; Neural computing techniques; Optimisation; Optimisation techniques; Power and plant engineering (mechanical engineering); Vibrations and shock waves (mechanical engineering); AFSA-BP hybrid algorithm; fault diagnosis; hydroelectric power stations; fault sets; frequency spectrum feature extraction; artificial fish swarm algorithm; vibration symptom; optimisation; BP neural network; mechanical engineering computing; backpropagation; fault symptoms; hydroelectric unit; neural nets; vibrations
• ISBN: 9781849199070; 1849199078
• DOI: 10.1049/cp.2014.1136

The fault of hydropower unit is affected by many factors, it is difficult to find the correspondent fault symptoms and causes through the theoretical analysis. Considering the disadvantages of BP neural network, such as slow convergence rate and getting into local extremum, the initial parameters are optimized by the improved artificial fish swarm algorithm and the model for fault diagnosis is established. The vibration symptom and fault sets of hydropower unit are formed through the extraction of frequency spectrum feature. By the method of improved artificial fish swarm algorithm and BP neural network, the fault of hydropower unit is diagnosed. The results show that this method has high diagnostic accuracy.