Online voltage stability assessment of load centers by using neural networks

This paper presents a neural network based method for evaluating online voltage stability conditions for a selected load center of an electric power system. Starting with a dynamic model of the system, a suitable index is defined to evaluate the proximity of the power network to voltage collapse. Then, a three-layer feedforward neural network is trained to give, as output to a prespecified set of inputs, the expected value of the voltage stability index. For this purpose, two different neural network architectures are proposed. The error back-propagation algorithm is used in this paper to train the chosen neural network structure. Moreover, it is shown that a good estimate of the real power margin of the selected load center can also be obtained using the value of the output of the designed neural network. To demonstrate the effectiveness of the proposed neural network based approach for voltage stability monitoring, a sample power system is considered. Test results show that neural networks can yield, in real time, an accurate assessment of voltage stability conditions.