Power system stability assessment and enhancement using computational intelligence
Duisburg, Essen (2011), XII, 175 S.
Dissertation / Fach: Elektrotechnik
Fakultät für Ingenieurwissenschaften » Elektrotechnik und Informationstechnik
Erlich, István (Doktorvater, Betreuerin)
Strunz, Kai (GutachterIn)
The main objective of the dissertation is to develop a fast and robust tool for assessment of power system stability and design a framework for enhancing system stability. The proposed framework is - based on the investigation of the dynamic behavior of the system - a market based rescheduling strategy that increases the stability margin. The dissertation specifically puts emphasis on the following approached: Power System Stability Evaluation: System stability is investigated by simulating a set of critical contingencies to determine whether the disturbances will result in any unsafe operating conditions and extract the necessary information to classify system states. The classification is based on the computation of the critical fault clearing time (CCT) for transient stability assessment (TSA) and the minimum damping of oscillation (MDO) for power system oscillatory stability assessment (OSA). The customary method of power system transient stability analysis including time-domain simulation (TDS) is used to compute the CCT at each critical contingency and Prony analysis as an efficient identification technique to estimate the mode parameters from the actual time response. The use of Prony analysis is to account for the effects of the change in location of the small disturbances as well as the increase in system nonlinearity on oscillating modes. Fast Power System Stability Assessment Tool: An artificial neural network (ANN) is designed to serve as accurate and fast tool for dynamic stability assessment (DSA). Fast response of ANN allows system operators to take suitable control actions to enhance the system stability and to forestall any possible impending breakup of the system. Two offline trained ANN are designed to map the dynamic behavior by relating the selected input features and the calculated CCT (as indicator for transient stability) and MDO (as indicator for oscillatory stability). Input features of ANN are selected to characterize the following: Changes in system topology and power distributions due to outage of major equipment such as transmission line, generation unit or large load Change in fault location and the severity of the fault Variation in loading levels and load allocation among market participants The features are generated for a wide range of loading at each expected system topology. Initial feature sets are pre-selected by engineering judgment based on experience in power system operation. In order to improve the accuracy of ANN to map the power system dynamic behavior, final selection is performed in the following three steps. In the first step, the generators terminal voltage drops immediately after fault are selected features to characterize the severity of the contingency with respect to the generators and to detect the fault location. In the second step, new features based on the inertia constant and the generated power in each area are calculated to characterize the changes in system topology and power flow pattern during normal and abnormal operation. In the third step, a systematic clustering feature selection technique is used to select the most important features that characterize the load levels and the power flow through lines from the mathematical viewpoint. The results prove the suitability of ANN in DSA with a reasonable degree of accuracy. Dynamic Stability Enhancement: To achieve online dynamic stability enhancement an online market based rescheduling strategy is proposed in the deregulated power systems. In case of power system operation by a centralized pool in vertically integrated electric utilities, generation rescheduling based sensitivity analysis is proposed. In the proposed market for deregulated power systems, the transactions among suppliers and consumers participating in the market are reallocated based on optional power bids to enhance system stability in case the available control actions are insufficient to enhance system stability. All participants are allowed to submit voluntary power bids to increase or decrease their scheduled level with equal chance. These bids represent the offered power quantity and the corresponding price. The goal of the framework is to enhance system stability with minimum additional and opportunity costs arising from the rescheduling. In case of vertically integrated electric utility, generation rescheduling based sensitivity analysis is used to enhance the system stability. The sensitivity analysis is based on the generators response following the most probable contingency. The generators are split into critical machines with positive sensitivity and non-critical machines with negative sensitivity. The change of the generation level among critical and non-critical machines provides the trajectories for stabilization procedure. The re-allocation of power among generators in each group is calculated based on the generator capacities and inertia constant, which simplifies the optimization procedure and speeds up the iterative to find a feasible solution. The objective is to minimize the increase in the cost due to rescheduling process. Particle swarm optimization is used as an optimization tool to search for the optimal solution to enhance the system stability with a minimum cost. The handling of all system constraints including stability constraints is achieved using a self-adaptive penalty function. Comparison strategy for selecting the best individuals during the optimization process is proposed where the feasible solutions are ever preferable during selection of local and global best particles.