Prospective and Efficient Techniques for Model Reduction in Reliability Calculations

Abstract

A reliable electric power supply is essential for modern society. Recently, severe blackouts worldwide have attracted attention to reliability studies in power system planning and operation. The relevance of the traditional N-1 criterion has been discussed, and much focus has been directed towards developing satisfactory probability based reliability tools. Goodtech Project & Services has developed a methodology for calculation of online power delivery reliability for use in power system operation and planning. The method, based on Markov models, analyzes the entire network for a large number of fault combinations, a useful approach for relatively small networks. However, the computation time increases polynomial with increasing system size. Because the impact of an outage has a limited geographical extent, it is desirable to reduce the system to be simulated, so that it only includes the affected area. The objective of this project is to develop methods for identifying the components that can be considered influenced by a fault. The focus of the pre-study was to evaluate which post-processing method best suited for developing a reduced network system. In the pre-study, power flow results from the standard DC load flow were used. The main focus of this thesis has been on developing and implementing fast methods for obtaining the necessary power flow data needed in order to use the post-processing methods. Three approaches have been investigated and tested, namely AC load flow based on the fast decoupled load flow with compensation techniques for obtaining the post- contingency power flows, DC load flow with compensation techniques and the efficient bounding method. The key principle of the compensation methods is that the effect of outages can be calculated by introducing simple compensation terms, thus avoiding the need to rebuild and factorize the system matrices for every contingency case. The bounding method is based on the principle of sensitivity factors and the fact that given knowledge about changes inside a boundary certain conclusions can be made regarding the changes outside it, thus eliminating the need for studying the entire system. The method based on the fast decoupled load flow is the only method that gives the possibility to provide an exact solution, and is also the only method that includes reactive power and voltage magnitudes. It is therefore recommended used in cases where a high degree of accuracy is important, or if reactive power and voltages are of interest. Tests show that the DC load flow is fastest. The accuracy is assumed sufficient for most intended purposes, and should therefore the preferred choice in most cases. Bounding methods are especially useful in highly meshed grids, and if only the largest changes are of interest.