A system reliability model based on Bayesian network(BN)is built via an evolutionary strategy called dual genetic algorithm(DGA).BN is a probabilistic approach to analyze relationships between stochastic events.In con...A system reliability model based on Bayesian network(BN)is built via an evolutionary strategy called dual genetic algorithm(DGA).BN is a probabilistic approach to analyze relationships between stochastic events.In contrast with traditional methods where BN model is built by professionals,DGA is proposed for the automatic analysis of historical data and construction of BN for the estimation of system reliability.The whole solution space of BN structures is searched by DGA and a more accurate BN model is obtained.Efficacy of the proposed method is shown by some literature examples.展开更多
油浸式变压器在运行老化过程中难免会出现各种潜伏性故障,及时正确诊断出变压器的状态至关重要,传统利用基于油中溶解气体分析法(dissolved gas analysis, DGA)数据的三比值法因存在编码不足的缺陷,限制了故障的诊断效果。为此提出了一...油浸式变压器在运行老化过程中难免会出现各种潜伏性故障,及时正确诊断出变压器的状态至关重要,传统利用基于油中溶解气体分析法(dissolved gas analysis, DGA)数据的三比值法因存在编码不足的缺陷,限制了故障的诊断效果。为此提出了一种改进的蝠鲼算法(manta ray foraging optimization, MRFO)优化反向传播(back propagation, BP)网络的故障诊断模型。首先利用逻辑映射与反向学习(opposition based learning, OBL)融合的多阶段算法为MRFO提供初始位置,加强算法全局寻优能力;同时提出利用正交实验法优化蝠鲼算法的3种觅食策略,调节蝠鲼个体的探索与开发,以加强该算法在特定问题上的寻优能力;最后将改进的蝠鲼算法寻得的最优解赋予BP网络的权值和偏置,建立变压器故障诊断系统。利用IEC TC 10故障数据进行了实验,并与其他算法进行了结果对比分析。结果表明,所提方法与BPNN、未改进的MRFO-BP、三比值法的结果相比,分别高出16%、8%、24%,是一种积极有效的方法。展开更多
基金National Natural Science Foundation of China(No.61203184)
文摘A system reliability model based on Bayesian network(BN)is built via an evolutionary strategy called dual genetic algorithm(DGA).BN is a probabilistic approach to analyze relationships between stochastic events.In contrast with traditional methods where BN model is built by professionals,DGA is proposed for the automatic analysis of historical data and construction of BN for the estimation of system reliability.The whole solution space of BN structures is searched by DGA and a more accurate BN model is obtained.Efficacy of the proposed method is shown by some literature examples.
文摘油浸式变压器在运行老化过程中难免会出现各种潜伏性故障,及时正确诊断出变压器的状态至关重要,传统利用基于油中溶解气体分析法(dissolved gas analysis, DGA)数据的三比值法因存在编码不足的缺陷,限制了故障的诊断效果。为此提出了一种改进的蝠鲼算法(manta ray foraging optimization, MRFO)优化反向传播(back propagation, BP)网络的故障诊断模型。首先利用逻辑映射与反向学习(opposition based learning, OBL)融合的多阶段算法为MRFO提供初始位置,加强算法全局寻优能力;同时提出利用正交实验法优化蝠鲼算法的3种觅食策略,调节蝠鲼个体的探索与开发,以加强该算法在特定问题上的寻优能力;最后将改进的蝠鲼算法寻得的最优解赋予BP网络的权值和偏置,建立变压器故障诊断系统。利用IEC TC 10故障数据进行了实验,并与其他算法进行了结果对比分析。结果表明,所提方法与BPNN、未改进的MRFO-BP、三比值法的结果相比,分别高出16%、8%、24%,是一种积极有效的方法。
