【论文推荐】华北电力大学 李俊卿等：基于GRU-CNN的同步发电机励磁绕组匝间短路故障预警

同步发电机是电力系统的重要组成部分，对于维持电力系统正常稳定运行至关重要。为了实现同步发电机励磁绕组匝间短路故障预警，提出了一种采用改进粒子群优化（IPSO）确定结构参数的门循环单元卷积神经网络模型（GRU-CNN），该模型的输出是励磁电流和无功功率。选取励磁电流偏移距离和无功偏移距离融合而成的总偏移距离作为故障预警标准，采用反熵加权法确定励磁电流和无功功率的融合权重。为了避免误诊，设置故障预警时间和故障预警率。故障预警阈值和故障预警率根据正常总偏移距离设置，故障预警时间根据实际情况设置。所提出方法通过实验验证了其有效性。

Synchronous generators are important components of power systems and are necessary to maintain its normal and stable operation.To perform the fault diagnosis of mild inter-turn short circuit in the excitation winding of a synchronous generator, a gate recurrent unit-convolutional neural network (GRU-CNN) model whose structural parameters were determined by improved particle swarm optimization (IPSO) is proposed.The outputs of the model are the excitation current and reactive power.The total offset distance, which is the fusion of the offset distance of the excitation current and offset distance of the reactive power, was selected as the fault judgment criterion.The fusion weights of the excitation current and reactive power were determined using the anti-entropy weighting method.The fault-warning threshold and faultwarning ratio were set according to the normal total offset distance, and the fault warning time was set according to the actual situation.The fault-warning time and fault-warning ratio were used to avoid misdiagnosis.The proposed method was verified experimentally.

Synchronous generator, Inter-turn short circuit, Excitation winding, Fault warning, GRU-CNN, IPSO.