Static security assessment of a power system using query-based learning approaches with genetic enhancement

A new approach of using query-based learning in neural networks to solve static security assessment problems in a power system is proposed. This learning method is intrinsically different from the learning performed by randomly generated data. Query-based learning is a methodology that requires asking a partially trained neural network to respond to the questions. The response of the query is then taken to the oracle. An oracle makes judicious decisions that help improve the quality of training data, thereby guaranteeing the assessment results. Moreover, to further improve the learning performance, the method is enhanced by the aid of genetic algorithms. Therefore the neural network is intelligently guided to a near-optimal initialisation. The probability of learning stagnation can be thus decreased. This method was tested on the Taiwan Power System through the utility data. Test results demonstrated the feasibility and effectiveness of the approach for the applications considered.