Dynamic stability assessment of power system using a new supervised clustering algorithm

This paper proposes a new approach to dynamic stability assessments of power systems. This approach applies the supervised concept to a clustering neural network, and directly uses the voltage magnitudes or frequencies of buses. In this method, the threshold of clustering is adapted to acquire desired categories, and a noise-tolerant parameter is added to reduce the influence of noise patterns. Therefore, this method of dynamic stability assessment has the advantages of real-time assessment, parallel computing, high correct rate, less time and memory consumption, and reduced noise pattern influence. To demonstrate the efficient processing of the algorithm, a dynamic stability assessment has been simulated to a simplified Seattle power system. Results show that the simulation provides a fast and appropriate assessment of power systems.