Improving Distribution System Observability Using Neural Networks

The increasing energy generation in electrical distribution grids creates the need to monitor and accurately control distribution grids. The requirement for increased observability of distribution networks is diametrical to the classical operation of electrical networks, which is based on power generation at the transmission level and power consumption at the distribution level. In this work, a datadriven method for state estimation in the distribution network is presented, whereby the requirement for increased observability of distribution grids can be fulfilled. The method is based on the application of artificial neural networks and therefore differs significantly from classical state estimation. This method is characterized by low and flexible use of measuring instruments and short calculation time. Besides, no detailed network data in the form of the admittance matrix is required. The validation is carried out on the basis of a low-voltage grid, whereby different network configurations are also considered. The results show that the method is suitable for increasing the observability of distribution networks.