摘要
传统的人工智能方法处理电网故障诊断中交叉数据模式识别问题的效果不甚理想。为此,作者提出运用量子神经网络进行故障诊断的算法,借鉴量子力学的相关概念,不断更新各层神经元的连接权以及隐含层各神经元的量子间隔,以达到提高故障诊断容错性的目的。仿真结果表明,在保护动作信息不完备的情况下,该算法的故障判断准确性明显优于传统神经网络。另外,该算法对存在一定错误数据的故障信息也具有良好的识别能力。
When traditional artificial intelligence approaches are used to recognize the cross data pattern in the power grid fault diagnosis, its result is not ideal. For this reason, the authors propose a new fault diagnosis algorithm in which the conception of quantum neural network is adopted. In the proposed algorithm, the connection weights of neurons of various layers as well as the quantum intervals of neurons in hidden layers are constantly updated to attain the expected purpose of improve the fault toleration in power grid fault diagnosis. Simulation results show that under the condition of incomplete protection action information the accuracy of fault recognition by the proposed algorithm is better than those by traditional neural network methods. Otherwise, the proposed algorithm can also recognize such fault information in which certain incorrect data exists..
出处
《电网技术》
EI
CSCD
北大核心
2008年第9期56-60,共5页
Power System Technology
基金
教育部优秀新世纪人才支持计划项目(NCET-06-0799)
四川省杰出青年基金项目(06ZQ026-012)~~
关键词
量子神经网络
故障诊断
激励函数
电力系统
quantum neural network
fault diagnosis
incentive function
electric power system
