基于自组织神经网络的模糊聚类同调机群识别  

A Fuzzy Clustering Method to Recognize Coherent Generator Groups Based on Self-Organizing Neural Network

作　　者：刘绚[1] 文俊[2] 刘天琪[1] 

机构地区：[1]四川大学电气信息学院,四川省成都市610065 [2]重庆市电力公司沙坪坝供电局,重庆市沙坪坝区400030

出　　处：《电网技术》2010年第7期98-102,共5页Power System Technology

基　　金：国家自然科学基金资助项目(50595412)~~

摘　　要：给出了一种利用基于融合自适应共振理论和Kohonen网络基本思想的自组织神经网络(简称自组织神经网络)的模糊聚类方法识别电力系统同调机群的算法。首先对输入数据进行模糊预处理,即采用最大–最小法建立能够反映发电机组间同调程度的模糊相似矩阵;然后将其每行或每列输入自组织神经网络模型进行训练,最终竞争获胜的输出层神经元代表不同的动态类型,即不同的同调发电机组;最后在EPRI-36节点系统上分别对自组织神经法和自组织神经模糊聚类法进行了仿真计算。结果表明:自组织神经模糊聚类法的识别结果比自组织神经法更加接近基于时域仿真的结果,没有出现误判,且自组织神经法能在更大时间范围内对同调机群进行准确识别。An algorithm to recognize coherent generator groups in power grid by a fuzzy clustering method based on self-organizing neural network, in which the fundamental thought of adaptive resonance theory is merged with that of Kohonen’s self-organizing neural network, is proposed. Firstly, the fuzzy preprocessing of input data is performed, that is, by use of maximum and minimum value algorithm a fuzzy similar matrix that can reflect the coherent extent among generator groups is built; then inputting each row or each column of the fuzzy similar matrix into the self-organizing neural network to train them, and the neurons in output layer that eventually win in the competition represent different dynamic types, i.e., the different coherent generator groups; finally, the simulation of CEPRI 36-bus system in power system analysis software package （PSASP） are performed by the self-organizing neural network merging the adaptive resonance theory with Kohonen’s self-organizing neural network and the fuzzy clustering method based on the self-organizing neural network respectively. Simulation results show that the recognition result by the fuzzy clustering method based on the self-organizing neural network is more close to the result based on time-delay simulation than that by pure self-organizing neural network, and there is not misjudgment, and the proposed fuzzy clustering method based on self-organizing neural network can accurately recognize coherent generator groups in a wider time range.

关 键 词：电力系统 自适应共振 KOHONEN网络 自组织神经网络 模糊聚类:同调机群 同调识别 

分 类 号：TM71[电气工程—电力系统及自动化]