Research On Fault Diagnosis Of Distribution Network Based On Clustering And Rough Sets

Under artificial intelligence and big data environment,energy internet products and smart home have been went into people's life.The data generated during the operation of the distribution network shows an exponential growth.On the one hand,with a large amount of information having been going into management center,when the distribution network fails,accurately determine the location of the fault is very difficult.On the other hand,the generated data has extremely high value,such as better analyzing the electricity user's electricity consumption behavior,designing the distribution network fault diagnosis system and so on.As the last step to complete the power supply to users,the reliability of distribution network operation directly determines whether the user can get sustained power supply.Therefore,this paper has some practical significance for the research of distribution network fault diagnosis.In this paper,the Rough Set(RS)theory and the classical k-means clustering algorithm are orderly researched.It is found that the RS can not directly deal with the continuous attributes.The classical k-means algorithm is sensitive to the initial center value and the clustering number 6)is difficult to choose.In view of these problems,thispaper respectively presents the solution;designs a set of fault diagnosis scheme of distribution network based on clustering and RS,and gives the algorithm description.Finally,through an example analysis,draws the following conclusion:(1)As a mathematical tool to deal with uncertain and inaccurate problems,RS possesses powerful abstraction ability and it is suitable for distribution network fault diagnosis.It can extract the importance of different attributes in the information system,remove the redundant attributes and mine the hidden decision rules.(2)RS mines information with the premise of the inextricable relationship of information system.The original decision-making table can still identify the system described by the original decision table with minimal information.(3)Sometimes a single RS fault diagnosis can only diagnose a fault area.The introduction of continuous signals such as voltage and current improves the accuracy of fault diagnosis in a distribution network.The innovation points created in the paper is listed as the following sentence:(1)Employing RS theory to process the discrete data generated by the protectors and circuit breakers,it can be concluded that the distribution network fault area or the specific fault element.In the case of failing to obtain a specific faulty component,a continuous signal is introduced to improve the diagnostic accuracy,which overcomes the shortcomings of using a rough set to obtain a faulty area.(2)The advantages and disadvantages of the classical k-means clustering algorithm are researched.For the number of clusters 6),the number of the classification of the fault areas obtained by the rough set is used as the 6)value;sensitive to its initial value,the bisection k-means clustering is used to overcome The classic k-means due to the sensitivity of the initial value into the local shortcomings of the best.(3)Because RS can not process continuous data directly and only discrete data can be processed,a clustering discretization algorithm based on bisection k-means is designed and used as the basis of the algorithm of rough set.