The Research And Application On Density Gradient Algorithm In The Classification And Synthesis Of Load Dynamic Characteristics

The digital simulation applied in the research of power system has became the main measure in system decision, plan, dispatch, operation. Load model is one of the most important factors affecting the precision and reliability of the simulation result. The composite load has the characteristics of complex composition, random time variance and spread over different area, the most difficulty of composite load modeling is the time-variation character of the load. Researches for near years have showed that the classification and synthesis for load characteristics is the effective way of settling the time-variation problem.This article studies classification and synthesis for dynamic load characteristics. Date mining is the method that can solve some problems in electric power system and it has a good prospect of applying, based on the systematic researches of the dynamic load characteristics classification and synthesis problem and their methods, this article presents a density gradient clustering method based on data mining is applied in classification and synthesis for dynamic load characteristics. Its main ideal is: based on the random process correlation theory, it uses the field measured response space as the character vector space of dynamic load characteristics, by calculating the density of measured response space data sample and its neighbors, searched points as the original centers of clusters; then it combined according to the distribution of border points between clusters. Clustering center appears in the classifying process, take it as equivalent sample, and by identifying the equivalent sample, the synthetic load model of this classification was obtained. The classification and synthesis are completed in the same process so it is convenient. By synthesizing the dynamic load characteristics data collected from a substation, the correctness and effectiveness of the method has been proved.After analyzing and comparing with other methods in clustering process, clustering centre and fitting errors, the result shows that the method has advantages in operation speed and fitting errors. Density gradient clustering method based on data mining technique is considered as a valid method.