Neural networks for fault location in substations

Silva, A.P. Alves da; Insfran, A.H.F.; Silveira, Paulo M.; Lambert‐Torres, Germano

doi:10.1109/61.484021

Cited by 38 publications

(12 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Hence, after fully weighing and evaluating fault and symptom masses information, we adopt SPA-based transformer fault diagnosis method in order to expect a better diagnosis result. SPA-based transformer fault diagnosis refers to that transformer fault symptoms are used to act as influence factors, and fault sources serve for the indices system of SPA, from the known fuzzy symptoms information, fuzzy subjection degrees of fault symptoms are calculated according to their fuzzy membership function established early, then in light of Table I, to search for the positive, negative and indeterminate influence factors to each fault, meanwhile, according to the connection intensity between each fault source and fault symptom, each contact number u i is also worked out according to (2). With respect to transformer fault diagnosis, we still must consider the prior probability p i of each fault occurrence, and p i is used to revise u i .…”

Section: A Spa Based Transformer Fault Diagnosismentioning

confidence: 99%

“…To deal with those incomplete and indeterminate as well as ill fault symptoms information during the transformers insulation fault diagnosis, some novel techniques, say, neural networks in [1], [2], fuzzy systems in [3], [4], rough sets in [5], [6], expert systems in [7], [8], Petri net in [9], [10] and many emergent methods in [11], [12], [13], have already been reported broad applications in diverse fields. However, all these methods expose some problems during applications, more or less.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Set Pair Analysis Applied for Identifying Power Transformer Faults

Gen-suo

2006

2006 International Conference on Machine Learning and Cybernetics

View full text Add to dashboard Cite

In order to be able to more perfectly and roundly deal with incomplete and indeterminate as well as ill fault symptom information in process of transformer fault diagnosis, Set Pair Analysis (SPA) is applied to design emulation model and implement fault diagnosis in this paper. In this method, the consistency and discrepancy and conflict of fuzzy fault symptom information to same fault sources are fully considered based on statistical data simultaneously, then, according to contact number, the likelihood of each fault occurrence is separately worked out. In addition, the prior probability of each fault occurrence is also fully considered. Thus a more intelligent fuzzy inference system is formed. By the use of it in transformer fault diagnosis, both simulation and trial show that the proposed method possesses excellent intelligence and robustness, and is an ideal pattern classifier and transformer fault diagnosis method.

show abstract

Section: A Spa Based Transformer Fault Diagnosismentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Set Pair Analysis Applied for Identifying Power Transformer Faults

Gen-suo

2006

2006 International Conference on Machine Learning and Cybernetics

View full text Add to dashboard Cite

show abstract

“…12, the Hamming distance between different classes is calculated. The distance between classes E X and E Y representing emergencies X and Y, respectively, is determined as [30]: L0 R302,CB77,R46,R32,CB82,R47,R56,CB102 CB134 E 3 L0 R302,CB134,R27,R19,CB42,R49,CB88,R13,CB28,R40,CB66,R43,CB74 CB77 E 4 L0 R27,CB77,R46,R32,CB82,R47,R56,CB102 R302,CB134 E 5 L0 R27,R19,CB42,R49,CB88,R13,CB28,R28,R40,CB66,R43,CB74,R46,CB134 R302,CB77 E 6 L0 R302,R27,R19,CB42,R49,CB88,R13,CB28,R28,R40,CB66,R43,CB74,R46,R47,R56, The entire set of 5072 corrupted patterns is randomly divided into training, test and evaluation sets. The training set includes 2536 training samples and is used for training NN-0.…”

Section: Structure Of the Neural Network Identification Systemmentioning

confidence: 99%

Computational intelligence applications to crisis management in power systems

Negnevitsky¹

2008

IJAAC

View full text Add to dashboard Cite

Abstract-In emergency and abnormal conditions, a power system operator has to deal with a large amount of data and apply most appropriate remedial actions. However, due to emotional and psychological stress, an operator may not be able to adequately respond to critical conditions and make correct decisions. Mistakes can damage very expensive power system equipment or worse lead to major emergencies and catastrophic situations. Intelligent systems can play an advisory role suggesting the necessary actions, which should be taken to deal with a given emergency or abnormal condition as well as identifying failures of protection systems and circuit breakers. This paper outlines some experience obtained at the School of Engineering of the University of Tasmania in developing intelligent systems for power systems security. An expert system for clearing overloads applies the network sensitivity factors to determine appropriate actions, which include generation rescheduling, network reconfiguration and load shedding. An expert system for voltage control is developed and used for detecting voltage violations and providing a set of effective control actions to solve voltage problems in real-time. An artificial neural network is used to identify multiple failures of protection relays and circuit breakers.This system uses information received from protection systems in the form of alarms and is able to deal with incomplete and distorted data.

show abstract

“…Neural Networks (NNs) have been extensively applied to power systems, from load profiling to security assessment [10,11]. As NN models can cope with uncertainties on power system parameters, they have been successful in dealing with fault diagnosis [12,13]. Pre and post-fault phasors at one line terminal have been used in [14] as inputs to NNs for fault location and impedance estimation.…”

Section: Introductionmentioning

confidence: 99%