基于SOM神经网络的煤矿智能变电站故障录波启动判据算法

Start Criteria Algorithm of Fault Recording for Coal Mine Smart Substation Based on SOM Neural Network

由于传统故障录波启动判据算法具有一定局限性,论文提出一种基于SOM神经网络的算法。以A相电流越限为例进行了算法的研究,依次完成SOM神经网络的构建,网络训练以及聚类预测,将输入向量归一化后输入到训练好的SOM网络中,输出结果会在二维平面阵列中显示出来,网络拓扑结构中的蓝色神经元代表A相越限,此时需要启动录波。为了验证模型的正确性,依次将维数不同的两组向量输入网络模型中,输出结果表明,基于SOM神经网络的故障录波启动判据算法自适应能力较强,能有效地完成录波启动,误差较小。

Aiming at the limit of criteria algorithm of traditional fault recording start, an algorithm based on SOM neural network was proposed. A fault of A - phase current out of limit was taken as example to study the algorithm. The SOM neural network was established and trained and cluster prediction was completed, then the normalized input vector was inputted to the trained SOM network, and the output was to be shown in dimensional plane array, and blue neurons in network topology represent phase A over limit, at this point, wave recording should be started. In order to verify the accuracy of the model, two input vector of different dimensionality were inputted into the network model. The outcome result shows that, adaptive ability of the proposed algorithm is strong and can effectively complete the recording start with minor error.