Steady state security assessment of power system using neural networks

An adaptive pattern recognition approach based on an artificial neural network for calculation of steady state security is outlined. A Rumelhart feedforward net with error back-propagation learning scheme is described and implemented on a CIGRE 225 kV, 10 bus 13 lines system. The steady state security index obtained for a test case shows the capability and adaptability of this network. By backward propagation it is possible to obtain the combination of inputs (features) which give rise to the particular class membership. This information can be of great use in determining the necessary preventive and emergency actions to be taken when the system is in an insecure state. Though the results are very encouraging further efforts are required to make it an efficient and viable tool for steady state security analysis