中文摘要：

针对学习向量量化（1earningvectorquantization，LVQ）神经网络在气体绝缘全封闭组合电器GIS特高频局部放电识别过程中存在初始权值敏感、竞争层未被充分利用的问题．提出了利用思维进化算法（mindevolutionaryalgorithm,MEA）优化LVQ神经网络的GIS特高频局部放电识别模型。该模型采用K交叉验证来确定LVQ网络竞争层中最佳神经元数目．并在此基础上利用思维进化算法寻找LVQ网络的最优初始权值。构建最佳的局部放电识别网络模型。对比该模型和BP网络、LVQ网络以及K交叉验证LVQ网络的放电识别准确率．结果表明：MEA优化的LVQ神经网络具有更高的识别精度。文中的研究对于提高局部放电识别准确率具有一定的价值。

英文摘要：

Because the learning vector quantization learning （LVQ） neural network is sensitive to initial weights and its competition layer is not fully utilized in GIS UHF partial discharge recognition process, a GIS UHF partial discharge recognition model is proposed by employing the mind evolutionary algorithm （MEA）to optimize LVQ neural network. This model uses K-cross validation to determine the optimal number of neurons in the LVQ network competition layer, then adopts MEA to find the optimal initial weights of the LVQ network, thus the optimal partial discharge recognition network model is constructed. Compared with the models with BP neural network, LVQ neural network, and K-cross validation LVQ neural network in GIS UHF partial discharge recognition accuracy, the MEA optimized LVQ neural network has the highest recognition accuracy.