Fault Diagnosis For Transmission Network Based On Timing Bayesian Suspected Degree

With the continuous expansion of the modern power system scale and the increasing complexity of the system structure, the potential impact of the power failure on the security and stability of the whole power system is increased. When faults occur, correctly dealing with failures and reducing power loss has significant economic benefits. Additionally, it is the inevitable requirement of modern power system reliability. In view of the identification of fault components and the fault diagnosis is the pilot work to deal with the fault in the power automation system, the accuracy and rapidity of the power system fault diagnosis is crucial.The power network dispatching center is mainly used to monitor and control the power system under normal condition. When fault occurs, the alarm information of each monitoring end is uploaded to provide support for the fault diagnosis of power dispatch personnel. However, redundancy and conflicts exist in the multi-source and heterogeneous alarm information, at the same time the relay protections refuse to move or maloperate sometimes. The dispatching automation system does not have the ability to analyse and process the alarm information quickly and reliably when faults occur, which can not meet the needs of the modern large-scale power grid. Therefore, the research related to fault diagnosis came into being.Essentially, the power network fault diagnosis is to use the reverse reasoning to find out the primary fault component based on the intelligent algorithm based on under protection action and breaker tripping information. At present, the key points of fault diagnosis are in three aspects, which are information processing, topology analysis and intelligent algorithm. In order to improve the fault tolerance and accuracy of fault diagnosis, this paper presents a new algorithm for fault diagnosis of transmission network based on temporal Bayesian Suspected Degree. The main work is as follows:1. A deep analysis of the fault information received from the dispatch center is carried out. With the sequence of event (SOE) information obtained by the end of dispatch center combined with the scheculing relay characteristics, the circuit breaker tripping signals are divided into class I and class II according their action time. Clustering the tripping signals according to their importance, Class â…  and class â…¡ circuit breaker information can be assigned different weights in the algorithm index. The information of the missing I type circuit breakers are checked and made up by using the clustering information of the tripping circuit breaker based on set theory.2. The system topology is expressed by the element topological connection matrix and the breaker-bus matrix, in which transmission lines can be regarded as a generalized node. Compared to the traditional topology modeling based on the circuit breakers, both topological mapping complex degree and the algorithm complexity are low; additionally, matrix expressed topology is easy for probability assignment, which can effectively reflect the uncertainty of fault information and action elements in power system.3. Based on the probability weighted bipartite graph, the Bayesian suspected degree algorithm establishes the relationship between the fault and the symptoms intuitively and effectively. By the ratio of the actual occurrence sympotoms to the sympotoms of the suspicious fault set, Bayesian suspected degree index measures the importance of the actual symptom information. In this paper, timing characteristic of protection action and breaker tripping is quantized in the Bayesian suspected degree index for different weights. The increase of alarm information dimension and improvement of the accuracy of diagnosis can be achieved.4. The abduction method is introduced into the uncertainty reasoning of the fault information to diagnose and analyze the problems of switch refused, switch misoperation, lack of information, time scale errors and so on. When determining the switch tripping, switch misoperation and missing information, set theory and abductive inference is used. With the alarm information processed in time sequence, the diagnosis result is reasonable and correct, and the fault is analyzed completely.