摘要
提出一种利用改进粒子群算法和反向传播算法相结合的混合算法训练神经网络进行电力变压器故障诊断的方法。在改进的粒子群算法中考虑了邻居粒子中最优粒子信息,修正个体行动策略,增强粒子群的社会学习功能,保证全局搜索的有效性;引入随机粒子群机制,利用粒子群进化过程中的种群变异机制提高算法的寻优性能。变压器故障实例仿真和分析表明,该算法在收敛速度、计算精度和平均收敛性能方面都有较大改进,可有效诊断电力变压器故障。
A new power transformer fault diagnostic method was proposed, which is based on a neural network trained by the hybrid algorithm combining modified particle swarm optimization （modified PSO） algorithm with back propagation （BP） in order to enhance the fault diagnostic ability of conventional dissolved gas-in-oil analysis in power transformer. In the modified PSO algorithm, each particle is attracted towards the best previous positions visited by its neighbors in order to overcome the problem of premature convergence observed in many applications of PSO. The random particles mechanism was introduced to improve global convergence. Finally, simulation results of power transformer fault diagnosis show that convergence speed, computational accuracy and average convergence property of the hybrid algorithm are all improved to some extent. That shows the validity of this method in power transformer fault diagnosis by dissolved gas-in-oil analysis.
出处
《中国电力》
CSCD
北大核心
2009年第5期13-16,共4页
Electric Power
基金
国家自然科学基金资助项目(10774043)
华北电力大学青年教师科研基金资助项目(200811023)
关键词
粒子群优化算法
BP算法
神经网络
变压器
故障诊断
particle swarm optimization algorithm
BP algorithm
neural network
transformer
fault diagnosis
