Long term voltage stability analysis for small disturbances

This dissertation attempts to establish an analytical and comprehensive framework to deal with two critical challenges associated with voltage stability analysis:

1. To study the new competitive environment appropriately and give more incentive for reactive power supports, one has to evaluate the impacts of distributed market forces on voltage stability, which complicates the voltage stability analysis.
2. Accurately estimating voltage stability margin online is always the goal of the industry. Industry used to apply static analysis for its computation speed at the cost of losing accuracy. On the other hand, dynamic analysis can result in more accurate estimation, but generally has a huge computation cost. So a challenge is to estimate the voltage stability margin accurately and efficiently at a reasonable cost, especially for large system. Considering the first challenge, this dissertation applied eigenvalue based bifurcation analysis to allocate the contribution of voltage stability. We investigate how parameters of the system influence the bifurcations. Three bifurcations (singularity induced bifurcation, saddle-node and Hopf bifurcation) and their relationship to several commonly used controllers are analyzed. Their parameters? impact on these bifurcations have been investigated, from which we found a way to allocate the contribution by analyzing the relative positions of the bifurcations. For the second challenge, a new fast numerical scheme is developed to estimate voltage stability margin by intelligently adjusting the load increase ratio. A criterion, named EMD (Equilibrium Manifold Deviation) criterion, is proposed to gauge the accuracy of the estimation. And based on this criterion, a new computation scheme is proposed. The validity of our new approach is proven based on the well-known Runge-Kutta-Fehlberg method, and can be extended to other explicit single-step methods easily. Numerical tests demonstrate that the new approach is very practical and has great potential for industrial applications. This dissertation extends our new numerical scheme to stiff systems. When a system is ill-conditioned, the implicit method would be applied to achieve numerical stability. We further demonstrate the validity to combine the intelligent load adjustment technique with the implicit method to save the computation cost without loss of accuracy. This dissertation also delves into the auto detection of stiffness of the power system, and extends our new numerical scheme to general sytems.