This dissertation introduces advanced artificial intelligence based algorithm for detecting and classifying faults on the power system transmission line. The proposed algorithm is aimed at substituting classical relays susceptible to possible performance deterioration during variable power system operating and fault conditions. The new concept relies on a principle of pattern recognition and detects the existence of the fault, identifies fault type, and estimates the transmission line faulted section. The approach utilizes self-organized, Adaptive Resonance Theory (ART) neural network, combined with fuzzy decision rule for interpretation of neural network outputs. Neural network learns the mapping between inputs and desired outputs through processing a set of example cases. Training of the neural network is based on the combined use of unsupervised and supervised learning methods. During training, a set of input events is transformed into a set of prototypes of typical input events. During application, real events are classified based on the interpretation of their matching to the prototypes through fuzzy decision rule. This study introduces several enhancements to the original version of the ART algorithm: suitable preprocessing of neural network inputs, improvement in the concept of supervised learning, fuzzyfication of neural network outputs, and utilization of on-line learning. A selected model of an actual power network is used to simulate extensive sets of scenarios covering a variety of power system operating conditions as well as fault and disturbance events. Simulation results show improved recognition capabilities compared to a previous version of ART neural network algorithm, Multilayer Perceptron (MLP) neural network algorithm, and impedance based distance relay. Simulation results also show exceptional robustness of the novel ART algorithm for all operating conditions and events studied, as well as superior classification capabilities compared to the other solutions. Consequently, it is demonstrated that the proposed ART solution may be used for accurate, high-speed distinction among faulted and unfaulted events, and estimation of fault type and fault section.

There has been significant development in the area of neural network based power system protection in the previous decade. Neural network technology has been applied for various protective relaying functions including distance protection. The reliability and efficiency of ANN based distance relays is improving with the developing digital technologies. There are, however, some inherent deficiencies that still exist in the way these relays are designed. This research addresses some of these issues and proposes an improved protective relaying scheme.<p> The traditional ANN distance relay designs use parameter estimation algorithms to determine the phasors of currents and voltages. These phasors are used as inputs to determine the distance of a fault from relay location. The relays are trained and tested on this criterion; however, no specific relay characteristic has been defined. There is a need for development of a new methodology that will enable designing of an ANN that works as a generic distance relay with clearly defined operating boundary.<p> This research work presents a modified distance relaying algorithm that has been combined with a neural network approach to eliminate the use of phasors. The neural network is trained to recognize faults on basis of a specific relay characteristic. The algorithm is flexible and has been extended for the design of other relays. The neural network has been trained using pure sinusoidal values and has been tested on a 17-bus power system simulated in PSCAD. The training and testing of the neural network on different systems ensures that the relay is generic in nature. The proposed relay can be used on any transmission line without re-training the neural network.<p> The design has been tested for different fault conditions including different fault resistances and fault inception angles. The test results show that the relay is able to detect faults in lesser time as compared to conventional relay algorithms while maintaining the integrity of relay boundaries.

There has been significant development in the area of neural network based power system protection in the previous decade. Neural network technology has been applied for various protective relaying functions including distance protection. The reliability and efficiency of ANN based distance relays is improving with the developing digital technologies. There are, however, some inherent deficiencies that still exist in the way these relays are designed. This research addresses some of these issues and proposes an improved protective relaying scheme.<p> The traditional ANN distance relay designs use parameter estimation algorithms to determine the phasors of currents and voltages. These phasors are used as inputs to determine the distance of a fault from relay location. The relays are trained and tested on this criterion; however, no specific relay characteristic has been defined. There is a need for development of a new methodology that will enable designing of an ANN that works as a generic distance relay with clearly defined operating boundary.<p> This research work presents a modified distance relaying algorithm that has been combined with a neural network approach to eliminate the use of phasors. The neural network is trained to recognize faults on basis of a specific relay characteristic. The algorithm is flexible and has been extended for the design of other relays. The neural network has been trained using pure sinusoidal values and has been tested on a 17-bus power system simulated in PSCAD. The training and testing of the neural network on different systems ensures that the relay is generic in nature. The proposed relay can be used on any transmission line without re-training the neural network.<p> The design has been tested for different fault conditions including different fault resistances and fault inception angles. The test results show that the relay is able to detect faults in lesser time as compared to conventional relay algorithms while maintaining the integrity of relay boundaries.

This dissertation introduces advanced artificial intelligence based algorithm for detecting and classifying faults on the power system transmission line. The proposed algorithm is aimed at substituting classical relays susceptible to possible performance deterioration during variable power system operating and fault conditions. The new concept relies on a principle of pattern recognition and detects the existence of the fault, identifies fault type, and estimates the transmission line faulted section. The approach utilizes self-organized, Adaptive Resonance Theory (ART) neural network, combined with fuzzy decision rule for interpretation of neural network outputs. Neural network learns the mapping between inputs and desired outputs through processing a set of example cases. Training of the neural network is based on the combined use of unsupervised and supervised learning methods. During training, a set of input events is transformed into a set of prototypes of typical input events. During application, real events are classified based on the interpretation of their matching to the prototypes through fuzzy decision rule. This study introduces several enhancements to the original version of the ART algorithm: suitable preprocessing of neural network inputs, improvement in the concept of supervised learning, fuzzyfication of neural network outputs, and utilization of on-line learning. A selected model of an actual power network is used to simulate extensive sets of scenarios covering a variety of power system operating conditions as well as fault and disturbance events. Simulation results show improved recognition capabilities compared to a previous version of ART neural network algorithm, Multilayer Perceptron (MLP) neural network algorithm, and impedance based distance relay. Simulation results also show exceptional robustness of the novel ART algorithm for all operating conditions and events studied, as well as superior classification capabilities compared to the other solutions. Consequently, it is demonstrated that the proposed ART solution may be used for accurate, high-speed distinction among faulted and unfaulted events, and estimation of fault type and fault section.

In recent years, the world has seen a significant growth in energy requirements. To meet this requirement and also driven by environmental issues with conventional power plants, engineers and consumers have started a growing trend in the deployment of distributed renewable power plants such as photovoltaic (PV) power plants and wind turbines. The introduction of distributed generation pose some serious issues for power system protection and control engineers. One of the major challenges are power system protection. Conventional distribution power systems take on a radial topology, with current flowing from the substation to the loads, yielded unidirectional power flow. With the addition of distributed generation, power flow and fault current are becoming bi-directional. This causes loss of coordination between conventional overcurrent protection devices. Adding power sources downstream of protection devices might also cause the upstream protection device to be blinded from faults. Conventional overcurrent protection is mainly based on the fault levels at specific points along the network. By adding renewable sources, the fault levels increase and become dynamic, based on weather conditions. In this dissertation, power system faults are modelled with sequence components and simulated with Digsilent PowerFactory power system software. The modeling of several faults under varying power system parameters are combined with different photovoltaic penetration levels to establish a framework under which protection challenges can be better defined and understood. Understanding the effects of distributed generation on three phase power systems are simplified by modeling power systems with sequence networks. The models for asymmetrical faults shows the limited affect which distributed generation has on power system protection. The ability of inverter based distributed generators to provide active control of phase current, irrespective of unbalanced voltage occurring in the network limits their influence during asymmetrical faults. Based on this unique ability of inverter based distributed generators (of which PV energy sources are the main type), solutions are proposed to mitigate or prevent the occurrence of loss of protection under increasing penetration levels of distributed generation. The solutions include using zero and negative sequence overcurrent protection, and adapting the undervoltage disconnection time of distributed generators based on the unique network parameters where it is used. Repeating the simulations after integrating the proposed solutions show improved results and better protection coordination under high penetration levels of PV based distributed generation. / Dissertation (MEng)--University of Pretoria, 2017. / Electrical, Electronic and Computer Engineering / MEng / Unrestricted

Ce travail est lié au système de protection des réseaux de distribution. Les réseaux radiaux dedistribution peuvent être protégés simplement par une protection placée en tête du départ. Maisl'exploitation future des réseaux de distribution, qui se transforment en réseaux intelligents,flexibles et adaptatifs, va sûrement nécessiter une protégeabilité plus complexe. Parconséquent, un nouveau plan de protection pourrait être nécessaire afin d’augmenter la fiabilitédu réseau de distribution et le taux de productions décentralisées. Il pourrait inclure plusieursprotections déployées sur un départ. Le but principal de ce travail est d'étudier comment lesprotections pourraient agir (sur quel genre de grandeurs les protections reposeront, quellecoordination faut-il choisir) et d'analyser les limites de ces nouvelles protections. En déployantplusieurs protections qui divisent le départ en des zones plus petites, le plan de protectionproposé, reposant sur une formulation modifié et optimisée, proche de celle des protections dedistance classiques, déconnectera ainsi moins de consommateurs et de producteurs lors del’apparition de défauts. Cela devrait réduire le temps de coupures brèves et de diminuerl'énergie non fournie. Ce manuscrit présente une solution pour les réseaux HTA radiaux faceaux défauts monophasés. / This work is related to the protection system of distribution networks. Radial MV distributiongrids can be protected by a single protection relay located at the beginning of the feeder. Butthe future operation mode of distribution grids turning into Smart Grids should impose morecomplex operations. Therefore a more advanced protection scheme could be needed, in orderto improve the reliability of the distribution network and to enhance the DG interconnection. Itcould include several protection relays in series on a same MV feeder. The main purpose of thiswork is to investigate how the protection relays could work (on which measurements should theprotection decisions be based, how to coordinate the relays without communication) andanalyze the limits of such new protection schemes. Since the goal is to insert severalprotections that divide the grid into smaller sections, the proposed protection system, based onan adapted and optimized formula, inspired by distance relays algorithm, would thereforedisconnect fewer loads and producers when faults occur in the MV network. This should reducethe clearing operation time and Energy Not Supplied criteria. This paper presents a solution fora radial MV grid facing single-phased faults.

The reliability of protection systems has emerged as an important topic because protection failures have critical influence on the reliability of power systems. The goal of this research is to develop novel approaches for modeling and analysis of the impact of protection system failures on power system reliability. It is shown that repairable and non-repairable assumptions make a remarkable difference in reliability modeling. A typical all-digital protection system architecture is modeled and numerically analyzed. If an all-digital protection system is indeed repairable but is modeled in a non-repairable manner for analysis, the calculated values of reliability indices could be grossly pessimistic. The smart grid is emerging with the penetration of information-age technologies and the development of the Special Protection System (SPS) will be greatly influenced. A conceptual all-digital SPS architecture is proposed for the future smart grid. Calculation of important reliability indices by the network reduction method and the Markov modeling method is illustrated in detail. Two different Markov models are proposed for reliability evaluation of the 2-out-of-3 voting gates structure in a generation rejection scheme. If the model with consideration of both detectable and undetectable logic gate failures is used as a benchmark, the simple model which only considers detectable failures will significantly overestimate the reliability of the 2-out-of-3 voting gates structure. The two types of protection failures, undesired-tripping mode and fail-to-operate mode are discussed. A complete Markov model for current-carrying components is established and its simplified form is then derived. The simplified model can appropriately describe the overall reliability situation of individual components under the circumstances of complex interactions between components due to protection failures. New concepts of the self-down state and the induced-down state are introduced and utilized to build up the composite unit model. Finally, a two-layer Markov model for power systems with protection failures is proposed. It can quantify the impact of protection failures on power system reliability. Using the developed methodology, we can see that the assumption of perfectly reliable protection can introduce errors in reliability evaluation of power systems.