Font Size: a A A

Process Monitoring Of Ring Main Unit Based On NPE Algorithm

Posted on:2018-03-01Degree:MasterType:Thesis
Country:ChinaCandidate:X ZouFull Text:PDF
GTID:2382330518458908Subject:Control Engineering
Abstract/Summary:PDF Full Text Request
As the key link of Ring network power dispatching,a large amount of process data will be generated when the Ring Main Unit works.The process data can reflex its operational aspect.So monitor the whole process of can master fault condition of Ring Main Unit.Traditional fault detection method of Ring Main Unit is based on single variable.The method of single variable threshold fault detection ignores the correlation between the variables,loses data information,can only detect the direct overrun fault.However,variables of Ring Main Unit data are interrelated.Based on the univariate detection method is difficult to adapt to the Ring Main Unit data complexity and variability,the detection effect is often not meet the expected requirements.In view of the above problems,this paper puts forward the fault detection model of Ring Main Unit based on NPE algorithm.This paper uses the actual data of Ring Main Unit to model,getting the model space and residual space,compares the experimental results of the constructed model with the univariate model and the PCA algorithm model qualitative,and discusses the effect of the algorithm on fault detection.In addition,this paper also proposes three angles to improve the fault prediction model and forecasting model of Ring Main Unit which are based on NPE algorithm.There are future public weather data,cable aging life expectancy,previous user usage habits.The statistical limits of the two model statistics,T2 and SPE,are obtained after the model training which is based on a real data provided by a power supply bureau in Yunnan Province.Finally,for the fault detection model,uses the real-time normal data and historical single variables which are not exceeded the fault data to test the model.The result is the model has a low false alarm rate for normal data.Compared with the PCA algorithm,the NPE algorithm has a better detection effect on the fault data.In addition,use only a single variable overrun fault data to test,this paper finds that the model can also detect the fault.The validity of fault detection based on NPE algorithm is proved.For the fault prediction model,the model is verified by the data before the fault.The result is the prediction model can give a warning warn before the fault occurs.This proves the validity of the prediction model for the failure prediction of Ring Main Unit.
Keywords/Search Tags:Ring Main Unit, Data driven, Neighborhood Preserving Embedding algorithm, Fault detection, Fault prediction, T~2 and SPE statistics
PDF Full Text Request
Related items