Oscillatory stability assessment of power systems using computational intelligence
The main focus of this PhD Dissertation was the development of new assessment tools for Oscillatory Stability Assessment (OSA) of large power systems. Since the complete power system model is usually not available for analytical stability studies, the tools use Computational Intelligence (CI) applications such as neural networks, neuro-fuzzy, and decision trees. These CI methods are based only on a small set of data, which must represent the entire power system information. Therefore, feature selection is necessary. Within this thesis, three methods for effective feature selection were developed. These selection methods apply the k-means cluster algorithm, the principal component analysis, decision trees, and genetic algorithms. Changes in the network topology, the load situation, the operating point, and the influence of bad and missing data affect the quality and reliability of the proposed methods for OSA. The issues of robustness are discussed and methods for both bad data detection and restoration are developed. The bad data detection methods use similarity analysis, the principal component residuals, and time series analysis. The data restoration involves a neural network and non-linear sequential quadratic optimization. Finally, in the case of a critical situation, the transmission system operator also would like to know how to prevent the system from collapse and which action will lead to a better-damped power system. Therefore, a new tool for counter measure computation is proposed. This tool implements the CI based OSA methods and provides the transmission system operator with information, which operator action will have a stability improving impact on the power system
Dieser Eintrag ist freigegeben.