| Relay protection system is the first line of defense for the safe and stable operation of power system.Whether the relay protection system can operate reliably or not,the defect management of various relay protection devices is very critical and important.The rapidity,selectivity and sensitivity of relay protection can be guaranteed by the setting calculation work,but the reliability is closely related to the defects of the protection devices,even if the action setting,time setting and equipment matching plan are set reasonably,the lack of secondary system functions caused by the defects of the device itself will make the safe and stable operation of the grid lose its guarantee,which will lead to the expansion of the fault range and the loss of a large area of load.Therefore,the reliability of the relay protection device is an important checkpoint to ensure the"intrinsic safety" of the relay protection.In recent years,with the popularization of artificial intelligence in the power industry,there is no doubt that artificial intelligence technology will be introduced into the defect diagnosis of relay protection devices.The large amount of defect data accumulated in the operation of system and advanced data mining technology have created favorable conditions for the defect diagnosis of relay protection devices.In view of this,this paper has carried out the defect diagnosis of the AC power system protection devices based on actual data.The main content of the paper includes the following four points:(1)To obtain the statistical characteristics of defect data.Firstly,the defect data structure was introduced and statistical analysis for the defect data from the perspectives of defect distribution,defect cause and defect location were performed,and then based on the characteristics of data,the defect diagnosis targets were clarified and Artificial intelligence algorithms were matched;(2)To analyze the influence of device defects on the incorrect action behavior of relay protection system,the defect data and the incorrect action data were combined,and the incorrect action behavior analysis model is constructed based on the fault tree algorithm,which is applied to analyze the responsibility departments of incorrect action behavior and propose defect management suggestions;(3)Aiming at the auxiliary decision-making of defect rating,based on structured data and decision tree,a defect rating model suitable for the same series of relay protection devices of different manufacturers is proposed.Firstly,the defect data of relay protection device is screened and piled,secondly,the structured data is mined using the decision tree ID3 algorithm to construct a defect rating model,finally rating rules of series of device defects are analyzed with examples;(4)To analyze the text log of defects,based on unstructured text data and natural language processing algorithms,a method for constructing a professional dictionary in the field of relay protection is proposed and a defect rating model is constructed.Firstly,the applicability of text mining technology is analyzed,and then a professional dictionary construction method suitable for secondary system text mining is proposed and applied.Furthermore,a defect text classification model is constructed based on support vector machine(SVM)technology,which can be used to assist defect rating. |
