基于量子神经网络信息融合的变压器故障诊断

Fault diagnosis for power transformer based on quantum neural network information fusion

针对电力变压器故障的多样性和故障信息的不确定性、数据量大及无规律性的特点,提出了基于量子神经网络信息融合的故障诊断方法。将多个电力变压器故障参数信息引入到各子量子神经网络进行局部融合诊断,再将各局部诊断信息引入决策融合网络进行全局融合,最终诊断出5种电力变压器故障并给出可信度评价。实验仿真结果表明:量子神经网络信息融合方法有效,诊断结果可靠,能将不确定性的数据合理地分配到各类故障模式中,故障正判率达到97.78%,远高于BP神经网络信息融合及改良IEC三比值法。

In view of the diversity of power transformer faults and the characteristics of fault information such as uncertainty, large quantity of data and randomness, a method for fault diagnosis based on quantum neural network information fusion is presented. The power transformer fault data are input into each sub quantum neural network to perform partial fusion diagnosis, and then the results of partial fusion are sent to decision-making fusion network to complete total fusion, and ultimately five kinds of power transformer faults are detected and reliability evaluation is provided. The simulation results show that quantum neural network information fusion is effective and reliable, which can assign the uncertain data to each fault mode reasonably, and the right rate of fault diagnosis reaches 97.78% which is higher than that of BP neural network information fusion and the improved IEC method.