Research of converter transformer fault diagnosis based on improved PSO-BP algorithm

2017 2nd International Seminar on Advances in Materials Science and Engineering

Long, Qi^1 ; Guo, Shuyong^2 ; Li, Qing^3 ; Sun, Yong^3 ; Li, Yi^4 ; Fan, Youping^4

CSG EHV Power Transmission Company, Guangzhou

510663, China^1

Tianshengqiao Bureau, CSG EHV Power Transmission Company, Xingyi, 562400, China^2

Maintenance and Test Center, CSG EHV Power Transmission Company, Guangzhou

510663, China^3

School of Electrical Engineering, Wuhan University, Wuhan

430072, China^4

关键词: BP (back propagation) neural network;    BP neural networks;    Concave function;    Converter transformers;    Global search ability;    Improved PSO-BP neural networks;    Pre-mature convergences;    Time-varying acceleration coefficients;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/231/1/012015/pdf
DOI  :  10.1088/1757-899X/231/1/012015

来源: IOP

【 摘 要 】

To overcome those disadvantages that BP (Back Propagation) neural network and conventional Particle Swarm Optimization (PSO) converge at the global best particle repeatedly in early stage and is easy trapped in local optima and with low diagnosis accuracy when being applied in converter transformer fault diagnosis, we come up with the improved PSO-BP neural network to improve the accuracy rate. This algorithm improves the inertia weight Equation by using the attenuation strategy based on concave function to avoid the premature convergence of PSO algorithm and Time-Varying Acceleration Coefficient (TVAC) strategy was adopted to balance the local search and global search ability. At last the simulation results prove that the proposed approach has a better ability in optimizing BP neural network in terms of network output error, global searching performance and diagnosis accuracy.