CHAPTER ONE
INTRODUCTION
1.1.  BACKGROUND
The reliable operation of large power systems with small stability margin is highly dependent on systems and protection devices. The application of microprocessor based (numeric) relaying protection has improved performance over time but, this has not led to a major impact on speed, sensitivity, and selectivity of primary protective relays. However decision making based on elements of artificial intelligence (AI) can, in my view, lead to a major impact in the aforementioned especially as quick fault removal in lines maintenance is concerned here.
1.2.  POWER SYSTEM PROTECTION
Designing a fail-free power system, is neither economically justifiable nor technically feasible (Nagrath et al, 1994). Failure of apparatus due to surges or other causes leads to faults on a power system. Strictly speaking, a fault is any abnormal state of the system. In general faults consist of ‘short circuit’ (Stevenson, 1982). Open circuit faults pose potential hazards to personnel, but, they are less server that short circuit faults. Hence they (short circuit faults) must be removal from the system as fast as possible. In modern power systems, this short circuit fault removal process is done automatically, and the equipment that does it is called ‘protective system’ (Stevenson, 1982). This is a combination of transducers, relays and circuit breakers. Although this thesis does not concern the isolation of any part of the system in event of fault, the above premises is inevitable since fault location is a measure for ensuring speedy fault clearing, and fault clearing can only take place if the system is adequately protected.  
1.3.  ARTIFICIAL INTELLIGENCE (AI) TECHNIQUES IN POWER SYSTEM PROTECTION
In the last two decades, much of the efforts in power system analysis (control and protection has moved from the methodology of formal Mathematical modeling to the less rigorous techniques of artificial intelligence (AI) (Agggarwal et al, 1995). Today the main AI techniques found in power system applications are those utilizing the logic and knowledge representation of expert system (ES), fuzzy systems (FS), Genetic Algorithm (GA), Artificial Neural Networks (ANN) and recently, evolutionary computing (EC).
The application involves developing a programmable logic (PROLOG) for the manipulation of symbolic information in a manner that emulates human reasoning. It concerns, constraints satisfaction, and nevertheless, provides a rich in-depth look at the use of PROLOG to develope a problem description (Schalkoff, 1990). The overall goal of the application concerns the development of a reasoning  system involves developing a strategy to operate (Open) appropriate circuit breakers in the event of fault. Sample application of the above approach to power system fault clearing include fault location.
1.4.  COMPARISON OF AI TECHNIQUES
The goal of artificial intelligence (AI) is to produce intelligent machines which simulate or emulate human being’s intelligence. Artificial Neural Networks, Expert Systems and Fuzzy Systems all attempt to meet these objectives. The difference in them lies essentially in the way knowledge is represented in the system, and how it is obtained.
·        ARTIFICIAL NEURAL NETWORKS (ANN)
ANNs and fuzzy systems are similar in many ways; first they both store knowledge and use it to make decisions on new inputs. Both can generalize and produce correct responses despite minor variations in input vector. However, there are some fundamental differences. ANN has a major advantage of acquiring knowledge through training. Often the training set can be composed of actual observations of the physical world, rather than being formed from the human opinions used for fuzzy (or expert) systems. In other words, the neural network lets the data speak for itself; they cannot directly handle fuzzy information (Song et al, 1997).    
·        FUZZY SYSTEMS (FS): Like expert systems, fuzzy systems rely on If-Then- rules. These rules, while superficially similar, allow the input to be fuzzy, i.e. more like the natural way that humans express knowledge. For instance, we may say the system is ‘somewhat secure’. This linguistic input can be expressed directly by a fuzzy system. Therefore the natural format greatly eases the interface between the knowledge engineer and the domain expert, thus, a fuzzy system can represent knowledge in which an expert system may have difficulty (or needs a large set of rules). Fuzzy and expert system differ in one critical respect (Song et al, 1997). Fuzzy systems allow the representation of imprecise human knowledge in a natural, logical way, rather than forcing the use systems. Fuzzy systems allow the approximate terms that are nearly always employed by humans to express their judgments, thereby permitting more accurate knowledge representations. Thus fuzzy systems are more robust, more compact and simpler.