Please use this identifier to cite or link to this item:
Authors: Gadani, J. P.
Issue Date: 1980
Abstract: In recent years, considerable attention has been paid to reliability of electrical power systems. Continuously available power supply is an acknowledged prime requirement of all consumers. Industrial consumers place reliability of power supply very high on the list of their priorities,as loss of power supply means loss of production and their income. Therefore, a mere qualitative definition is unsatisfactory. It is well recognised that to evaluate the quality of a system or its ability to perform a required function, it is necessary to quantify the reliability of that system. For main tained systems such as power system, the measures of reliability that are usually sought are the probability and freauency of failure. Generally, the reliability models of power systems, communication systems, data processing systems etc.: are complex in nature. Nonseries-parallel configurationswith a large number of elements are the implied features of the reliability logic diagram of such systems. The problem of reliability evaluation of such a large and complex network has been the subject of intense research in the recent past. The number of papers published in the literature during the past decade is itself an indication of the importance of this problem. The consideration of maintenance in the relia bility studies requiresalot of effort and involvement. The compu tational time and memory requirements increase exponentially with -IVthe size of a network, and the minimization of the time and memory requirements is a challenging feature of the problem. Reliability evaluation techniques are constantly in the process of evolution. Some of the various approaches that have been commonly used for reliability evaluation are i 1. State enumeration 2. Graph-theoretic 3. Network decomposition k-. Tie-sets and cut-sets enumeration 5. Network reduction and transformation 6. Monte Carlo simulation Amongst the above mentioned techniques, the tie-sets and cut-sets are widely used. A number of reliability evaluation algorithms, based on these, have been published in the literature. Two major limitations of these approaches are - large memory require ments and a high CPU time,which is particularly true for large complex systems. Although, for complex networks, delta-star and star-delta transformations have been suggested in the literature, no effort has 'over been made to make this attractive technique amenable to practical was,therefore,felt necessary to exploit the useful features of this approach and moke efforts to improve its practical utility. The research work reported in the present thesis is an humble effort of the author in this direction. The main thrust of the present research \TOrk has been in the following -vareas: - Incorporation of additional conditions of equivalence of delta and star subnetworks in order to improve the accuracy of the delta-star and star-delta transformations. - Modifications of expressions for the delta-star and star-delta transformations so as to allow the consideration of unreliable nodes in the subnetworks. - Derivation of expressions of failure rate and failure frequency for delta-star and star-delta transformations in case of main tained systems. - Derivation of expression of probability of failure, in open mode and in short mode failures of 3-state device nctxrorks, for delta-star and star-delta transformations. - Solution of nonlinear algebraic equations for star-delta transformation, In the past, the major handicap in the use of the trans formation technique has been the lack of a general algorithm for reliability evaluation. Although the expressions for transforma tions were available in the literature, not a single paper on transformation approach proposed a scheme of handling a large network in selecting subnetworks for transformations. A computer algorithm has been developed which requires reliability network information as the basic data. To achieve an equivalent single element graph, the algorithm seeks to perform irious operations. These include series-parallel operations, removal or addition of a branch or of a node, selecting an appropriate subnetwork for transformation and transforming the selected subnetwork, updating the network information at each stage of network reduction, simplification etc. The algorithm has been successfully used to evaluate reliability of a number of complex networks and is found to require less CPU time as compared to any other method based on tie-sets and cut-sets approaches. Matrix representation is widely used for storing reliabi lity network information. From the point of view of computer memory requirement,this method is inefficient for large sparse reliability networks. To overcome this difficulty, a novel network information storing scheme has been proposed which necessitates minimum computer memory. Many a times?for maintained systems, it is necessary to evaluate system frequency of failure, up-time, down-time etc. in addition to system probability of failure. The network information storage scheme and the reliability evaluation algorithm have been modified to take care of the maintained elements. For large well-connected reliability networks, it has been observed that the delta-star and star-delta transformations are sometimes insufficient to tackle the complexity of the net work. Quadrilateral-star transformation is being suggested for the first time in the literature through this work in order to overcome some of the limitations of the transformation approach. Six different sets of transformation equations have been obtained and compared from the point of view of accuracy. Additionally, certain subnetwork configurations have been suggested which ensure further simplification of the reliability network after the -viitransformation has been affected. Boolean algebra has been used extensively in the deriva tion of reliability expressions. It is, therefore, logical to consider the feasibility of extending its use for the evaluation of system frequency of failure. The available expressions of frequency of failure derived using tie-sets and cut-sets approaches are modified so that Boolean algebra could be used to obtain dis joint product terms, in algorithm based on Abraham's algorithm and using tie-sets approach is developed and some simple changes for cut-sets approach have also been suggested. Generally, system performance indexes are evaluated using nominal values of failure-rate and repair-time of its elements. The uncertainty of these reliability data can be expressed by specifying either confidence level or standard deviation. It is obvious that any uncertainty in the reliability data would also introduce uncertainty in the system performance indexes. Therefore, the expressions of variances of probability and frequency of failure have been derived. Using these expressions and the developed algorithms, lower and upper bounds of system performance indexes have been obtained. A K-out-of-m:G system with maintained elements has also been considered in this thesis. The elements have been assumed to be non-identical and ^-independent. Recursive equations for availability, probability and frequency of failure have been derived and a computer amenable algorithm is developed.Another aspect of power system reliability studied by the author is the Reliability Index of Transient Stability. Several basic factors, such as type and location of fault, loading conditions of the system prior to a fault, operating time of the fault clearing devices etc., which affect the transient stability of a power system, are probabilistic in nature. The deterministic models used in transient stability studies ignore these considera tions. It is, therefore, more logical to evaluate reliability index of transient stability rather than merely determining transient stability of a power system under the worst possible condition whose probability of occurrence is very low. Reliability Index of Transient Stability is obtained for a single machine connected to an infinite-bus,using joint probability density function of initial load prior to a fault and location of the fault. The influence of various system parameters, such as operating time of fault clearing devices, machine inertia constant, reactance of machine etc. on HITS has been examined. Further,an indication of the possible extension of the research work reported in this thesis has been provided besides presenting upto date references on the subject.
Other Identifiers: Ph.D
Appears in Collections:DOCTORAL THESES (Electrical Engg)

Files in This Item:
File Description SizeFormat 

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.