Black-start partitioning method based on semi-supervised spectral clustering

Abstract

The invention designs a black-start partitioning method based on semi-supervised spectral clustering, wherein the method comprises the steps: grouping to-be-recovered units through combining with a genetic algorithm, taking a grouping result as a paired constraint condition, improving a spectral clustering algorithm, and then formulating a black-start partitioning recovery method. The method comprises the steps: firstly, indiscriminately treating nodes and lines in a system to be subjected to black start, and abstracting a power system to be subjected to partition start into an undirected weighted graph; secondly, grouping the units before partitioning the system, taking the minimum value of the number of the black-start units and the minimum value of the load number as the grouping number, establishing a unit grouping model which maximizes the generating capacity of the units and minimizes the sum of the recovery time and susceptance value of a line used for unit recovery, and solvingthe model by selecting the genetic algorithm; and finally, taking the unit grouping information as paired constraint conditions to correct an adjacency matrix, and improving unsupervised spectral clustering into a semi-supervised spectral clustering algorithm to solve the partition model of the power system.

Description

technical field [0001] The invention relates to the field of black start of electric power system, in particular to a partition method of black start. Background technique [0002] With the continuous development and construction of the power grid, the scale of the power grid continues to expand, the types of power sources and loads continue to increase, the operating status of the power grid is becoming more and more complex and changeable, and the probability of large-scale power outages in the power system is also increasing. Once a major blackout occurs, the losses and impacts will increase rapidly over time. Therefore, in order to reduce the harm of a major blackout, in addition to taking daily measures to avoid the occurrence of accidents, it is also necessary to formulate measures when the accident does not occur. A reasonable emergency recovery plan. [0003] Black boot recovery is divided into serial recovery and parallel recovery. Parallel recovery is to divide t...

AI Technical Summary

Problems solved by technology

The principle of the genetic algorithm is easy to understand, and the implementation is relatively simple, but the result is easy to converge to the local optimal solution; the K-Means spectral clustering algorithm has a large amount of calculation and the clustering result is easily affected by the distribution of the initial clustering center. The initial clustering center selection Unreasonable, easy to lead to unsatisfactory clustering results

The clustering effect of applying the spectral clustering algorithm for partition restoration is not affected by the sample space, and the amount of calculation is not large. However, if the unit grouping is not considered, the power system must be abstracted into a graph and divided to obtain the partition result without any difference. If the number of units in each partition is unreasonable, there may be no black start power supply in a certain partition

Images