Neural networks and pseudo-measurements for real-time monitoring of distribution systems

Abstract

A state estimation scheme for power distribution systems, based on artificial neural networks (ANNs), is proposed. Despite the influence of measurement uncertainties, it allows quantities describing the distribution system operation to be identified on-line, thereby constituting neural "pseudo-instruments". Details of the design and optimization of such a neural scheme are discussed, underlining the importance of ANN tuning to achieve greater levels of accuracy. The performance obtained in a study case, for different types of operating conditions, was analyzed and confirmed the feasibility and the robustness of the proposed approach. This neural estimation scheme proves to be preferable to traditional mathematical approaches whenever there are online requirements, due to the typically high operating speed of ANNs.