Design And Implementation Of Grid Coherent Clustering Algorithm Based On Sequence Non-parameter Clustering Strategy

In the construction project of smart grid,in order to achieve the goals of efficient output emergency control measures and active grid disconnection,the primary task of power system is to complete the automatic identification of generator cluster.With the rapid development of PMU real-time measurement system technology in WASM,the reliability of applying real-time monitoring time series data of generator attributes after faults to the coherent cluster identification task is also increasing.From the perspective of data mining,the coherent clustering algorithm based on the dynamic disturbance trajectory of the generator belongs to the clustering problem of multidimensional time series.The common problem of the above traditional clustering algorithms is that they need to complete the clustering task under the premise of manually setting external parameters.At the same time,the external parameter value has a decisive effect on the quality of the clustering result.However,in complex grid structures,there are massive disturbance failures.Each fault corresponds to a large-scale,high-dimensional time series data set.Each data set has various characteristics,strong unknown and high complexity.The external reference range is difficult to quantify.Traditional clustering algorithms cannot adaptively obtain ideal clustering results in coherent clustering of large power grids.Therefor,Therefore,in order to meet the requirements of rapid analysis and timely regulation of smart grids,it is very important to realize automatic coherent grouping of the grid.In order to solve the problem of traditional clustering algorithms that completely rely on external parameters,this paper introduces the concept of natural neighbors.Consider that the original natural neighbor definition does not have the ability to process noise samples in the data set.This paper improves the definition of natural neighbor and its search algorithm.A non-parametric hierarchical clustering algorithm NNFPCH based on natural neighbors is proposed.In this paper,based on the characteristics of coherent clustering time series data,IPET and IPEAT trend feature extraction algorithms are proposed,and this feature is used in conjunction with natural neighbor-based non-parametric clustering algorithm to perform generator coherent clustering experiments.Experimental results show that the trend features extracted in this paper are low-dimensional and efficient,and combined with the NNFPCH algorithm can quickly and accurately complete the grid automatic coherent grouping task.In order to enhance the applicability of NNFPCH on other types of data sets,this paper proposes an improved algorithm NCFPCH.This article gives the definition of the core point of nature.Based on the natural core point,a new and efficient sampleconnection rule is defined to form an adjacency matrix containing all samples.Finally,by searching all connected sub-graphs of the adjacency matrix to complete the clustering task of the data set.In order to discuss the practical applicability of the algorithm,this paper conducts experiments on many types of data sets,including classic data sets such as common handwritten numbers and face images.The experimental results show that,compared with the traditional clustering algorithm and the existing non-parametric clustering algorithm,NCFPCH has a wide range of applicable data sets,fast operation speed and excellent performance of comprehensive clustering results.