Abstract. The detection and identification of the bad data of the power system plays an important role in dispatching personnel to grasp the running status of the power grid in real time. In order to overcome negative effects of random selection of clustering initial values of traditional GSA bad data identification algorithm on identification precision and computation rate, this paper propose an optimized GSA algorithm based on area density statistics method. This algorithm by computing the area density of each cluster object to select k points that are farthest from each other and are at the highest area density as the initial cluster center. The experimental results show that the optimized GSA algorithm improves the accuracy of the degree of clustering dispersion and the recognition accuracy of the bad data. At the same time, the algorithm greatly reduces the computational complexity of iterative computation, improves the computing speed and saves a lot of computing time. In the case of huge system and large amount of data, this method is a rapid and efficient algorithm, and has potential of good application.
Aiming at the limitation of traditional abnormal monitoring method of wind turbine. In this paper, we propose a multidimensional clustering method called WPMCLU. Firstly, traverse the FP-Tree storage structure to find subspace instead of APRIORI self-connection mode. Secondly, define K Gauss models and identify clusters in each subspace. At last, according to the parameter Eq specified by user to divide normal and abnormal cluster of each subspace, remove redundancy and recognize abnormal data. And the method we propose runs on the Spark platform in order to easily extend to large data set of wind power. In experiment, the method we propose in this paper has a high recognition rate compares with CLIQUE, K-Means and DBSCAN algorithms.
Abstract. In order to solve the problem that the identification method of wind turbine is not suitable for large data environment. In this paper, we propose FSIQUE. Firstly, the data space is divided into dense grid cells and sparse grid cells. Based on the FS-Tree storage structure proposed in this paper, a dense grid cell is stored and the subspace is traversed by this storage structure. Secondly, traversing the connected grid cells in the subspace to find the clusters. Finally, the maximum and minimum coverage of the cluster is calculated. According to the parameter, the clustering is divided into normal data and abnormal data to realize the anomaly recognition. The proposed method is run on the Spark platform and compared with the WPMCLU and DBSCAN methods, has the highest abnormal recognition rate and the runtime is shortest.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.