基于人工免疫网络的断路器在线自学习故障诊断

On-line Self-learning Fault Diagnosis for Circuit Breakers Based on Artificial Immune Network

现有的高压断路器故障诊断方法在跟踪机械状态行为变化的能力上存在不足。在人工免疫网络aiNet的基础上,提出了一种具有在线自学习能力的断路器机械状态分类网络C-aiNet;详细介绍了C-aiNet的训练流程和在线自学习机制,讨论了影响网络性能的参数选取原则。该方法能够在线跟踪机械状态在特征空间中的变化,形成新的分类边界,适当遗忘旧的分类边界,并且能够识别新的故障模式。应用实测断路器振动特征数据的仿真结果表明,该方法能够取得较基于神经网络的故障诊断方法更为准确的诊断结果。

The existing voltage circuit breakers lack in of mechanical state. An on-line diagnostic methods for high the ability to pursue the change self-learning classifier, C-aiNet, for identifying mechanical failures of HVCBs based on artificial immune network aiNet is presented. The training and self-learning process were introduced in details, and principles for selecting network parameters were discussed as well. This method can trace the new regions of clusters, discard old ones, and recognize new patterns. The results of simulation based on measured vibration characteristic data of HVCBs show that self-learning method can achieve more precise judgment of the mechanical state of HVCBs over neural network method.