A Study of PSO-Based Fusion Neuron Network
A PSO-based fusion neuron network model is given for the complex non-linear dynamic mathematical modeling of a water turbine-generator set. Due to BP neuron networks shortcomings such as slow training speed and easily falling into the local extremum, an optimized PSO algorithm which has a better capacity of searching the globe extremum is used to adjust the weights and thresholds of the network during training. By using the data from various heterogeneous field-mounted sensors as input training sample data of the neuron network, the method dynamically adjusts the weights and thresholds of the BP network to approach the complex non-linear model. Simulation study shows the fusion neuron network model has an accuracy of 96.5％, and may meet requirements of practical applications.
Feng Xu Lin Xu Hai Zhao
College of Information Science and Engineering Northeastern University Shenyang, P.R. China, 110004
国际会议
南宁
英文
2007-07-20（万方平台首次上网日期，不代表论文的发表时间）