| With unstable new energy wind, solar, tidal and so on in the power system growing proportion, making the structure of power grid more complex. This brings a great challenge to the power grid fault diagnosis and other related works. While the rapid development of digital and smart power grid enabling a lot of historical operating data and real time data were acquired. Under the background of big data, machine learning techniques can get as much as possible implicit information of the system. As one of the most important machine learning algorithms, FCM algorithm can obtain more information in the grid with little prior knowledge case, and it is one of the most important power system fault diagnosis algorithms.Feature weighting algorithm has a great influence on the clustering effect. For traditional feature weighting algorithm ignores the features in the class and among the classes, an adaptive weighted FCM algorithm was proposed. The algorithm from the distribution of features after clustering, study the orderliness of features impact on the clustering results. With the feature of entropy and information gain after clustering as a criterion to adjust feature weights, by clustering iteration and feature weight updates until the feature of entropy convergence, and the optimal partitioning of the data set we obtained. Experimental results show that the adaptive weighting FCM algorithm can effectively distinguish the features of the importance of clustering, has a better division of clustering and higher clustering accuracy, and also has a better clustering effect.Power grid fault prone to chain reaction, fault diagnosis quickly and accurately is the key to the safe operation of power grid. View the fault analysis from the global power grid, power grid fault region identification algorithm is proposed based on adaptive weight FCM algorithm. Simulation results show that the algorithm can accurately identify the grid fault region, and the other grid buses are divided into serious region affected by the fault, less serious region affected by the fault, lighter region affected by the fault and the region isn’t affected by the fault. For the power grid bus fault diagnosis, grid bus failure types of off-line analysis and online diagnosis based on adaptive weighted FCM algorithm was proposed; the bus fault model obtained by offline learning the history operation data with adaptive weighted FCM algorithm. And then the fault model used in grid system, on-line diagnosis of grid real-time operation data quickly. Experiments show that the method is feasible and can fast diagnosis the power grid bus fault type. |
