Sensitivity characteristic of neural network as a tool for analyzing and improving voltage stability

This paper proposes a new approach for analyzing and improving voltage stability of power systems based on application of neural network technique and its sensitivity characteristic. The stability considered in this paper, is static voltage stability and the stability index, which is used, is system loadability limit associated with the point of voltage stability limit. In order to have a fast evaluation of power system voltage stability limit and also be able to perform sensitivity analysis of this limit with respect to system variables, a neural network is adopted and trained to construct a functional relationship between system variables and loadability limit. If a neural network evaluates voltage stability limit with a low security margin then it becomes necessary to improve voltage stability. For this purpose, output/input sensitivity characteristic of the neural network is used to evaluate the sensitivity of voltage stability limit with respect to system variables like reactive power generation. Then, this sensitivity is used as a guideline to modify reactive power generation to increase voltage stability limit. The proposed approach is tested on an IEEE 14-bus system with promising results.