Comprehensive Detection and Isolation of Fault in Complicated Electrical Engineering

In this paper, real time dynamic information of complicated electrical engineering provided by phasor measurement units to analyze and locate the fault is utilized. Based on fuzzy cluster analysis theory and the principle of minimum expected cost of misclassification, we have explored a comprehensive detection and isolation system of fault in complicated electrical engineering. It has been proved by a large number of simulation experiments that the comprehensive fault detection and isolation system can accurately and reliably detect and isolate faults and provide guarantee for the safety and stable operation of electrical power and energy system.


I. INTRODUCTION
The safe and stable operation of electric power system is of great importance to the national economy and people's livelihood, and it is also the goal pursued by electrical scientists.Relay protection, one of the three defensive lines of electric power system, plays an indispensable role in monitoring the abnormal situations of power grid, clearing fault components and safeguarding system security.In power system, the main protection based on two-terminal electrical measurements, represented by fiber differential, has got to be mature and perfect [1].However, it fails to be the effective backup of the adjacent electric element while ensuring the absolute selectivity.In the practice of electrical engineering, it is possible for the main protection to be invalid due to the disappearance of DC power supply.Therefore, it is not advisable to completely eliminate the traditional backup protection represented by zone 3 impedance relay which can be the remote backup of adjacent element.
With the increasing complication of power network, the variety of operation modes, and the higher demand of real-time and super real-time, however, many problems lie in the traditional backup protection due to its inherent setting principle, configuration scheme and the choice of information source [2]- [4]:  It only reflects local information of relay equipment of power grid.To ensure the reliability, the setting has to be made under the most severe and inclement operation conditions, which makes it rather conservative to realize the protection function.Meanwhile, to ensure the selectivity, the rapidity and sensitivity have to be sacrificed, which leads to the extended action time of protection. The cooperative relationship of backup protection is quite complex.The configuration and setting are of great difficulty.The emphasis lies in the ineffective adaption to the changes of the system operating condition, which, moreover, can cause unhealthy conditions of protection mismatching and lack of sensitivity. The backup protection cannot effectively differentiate the power flow transferring caused by internal faults and external faults.In many blackouts, the cascading trip of backup protection caused by heavy load even accelerates the extension of the range and scope of accidents.It is the fundamental approach to the current problems of backup protection to get rid of the restraint of partial or local information and to make full use of the multi-point and multivariate wide area information [5]- [7].
The researches in this paper are mainly serving wide area protection system.We will explore a comprehensive detection and isolation scheme of fault in complicated electrical engineering.This paper is organized as follows.In Section II, the theoretical foundation for fault detection and isolation is introduced.Integrated the fuzzy cluster analysis and minimum expected cost of misclassification, a comprehensive fault detection and isolation system will be put forward.In Section III, for general fault modes, the fault detection and isolation in complicated electrical engineering is discussed carefully.Finally, the paper is concluded in Section IV.

A. Fuzzy cluster analysis
In the general form of cluster analysis, each object in dataset will be assigned ultimately to a certain category.In

Comprehensive Detection and Isolation of Fault in Complicated Electrical Engineering
other words, it is impossible that an individual exists between categories.But fuzzy cluster analysis use a rather different classification ideas, every observation has partial member relationship in multiple categories, rather than has perfect member relationship in a single category.
The fuzzy cluster analysis is applying the concept of fuzzy set to traditional cluster analysis, and the subjection relation can be determined by membership function.The membership degree that each object belongs to every category is just a value in continuous interval [0, 1].In fact, an object can belong to multiple categories with different levels of membership degree.One of the major advantages of fuzzy cluster analysis is that it can adapt to those data and categories with poor separation property, which allows the fuzziness of data properties and can offer detailed information for the description of data structures [8], [9].
The membership function is the basis of fuzzy theory.In fuzzy classification, the object i a in object collection A is subordinate to a class with a certain membership degree, and all of the objects are subordinate to a certain class with different membership degrees.So, each class can be considered as a fuzzy subset on the object collection A , and the corresponding category of every classification result is just a fuzzy matrix S , namely Suppose the collection of classification objects is In the course of fuzzy cluster analysis, the appropriate fuzzy classification matrix S and clustering centre matrix B will be solved, which will make the objective function and the minimum expected cost of misclassification rule is just this kind of discrimination rule that can minimize ECM, namely: As can be seen from the former relationship, for a new sample, we only need to know the corresponding ratios, its attribution will be determined provided that the corresponding ratios are known.
Integrating the above research methods, we put forward a comprehensive fault detection and isolation system, see Fig. 1.According to different kinds of short circuit faults, we have carried through large numbers of simulation experiments, and the results have demonstrated that the comprehensive fault detection and isolation system in this paper is successful.Let's take a symmetrical short circuit fault in IEEE 39-bus system as an example, the electric diagram of IEEE 39-bus system can refer to [5].In the system structure, Bus18 is the actual fault position.By BPA simulation and program calculation with MATLAB, firstly, we study the partition of fault section and non-fault section based on fuzzy cluster analysis theory.By node positive sequence voltage, the category coefficients can be calculated.Here let's specify the number of categories as three: fault category, associated category and non-fault category.The category coefficient reflects the possibility that corresponding measurement value will be assigned into an appropriate category.The concrete category coefficients and results of node positive sequence voltage have been listed in Table I.And we have also obtained classification diagram.
By the principle of MECM, the classification function coefficients have been calculated and the linear discriminant model can be expressed as: Thereupon, the ultimate posterior probabilities and MECM classification results based on node positive sequence voltages are listed in Table II.
According to the above MECM classification results, the misjudgement ratio is zero, and the accuracy of classification has reached 100 %.

TABLE I .
THE CATEGORY COEFFICIENTS AND RESULTS OF NODE POSITIVE SEQUENCE VOLTAGES.