Usage of Artificial Neural Networks for pseudo measurement modeling in low voltage distribution systems

The transition from a passive conventional grid to a dynamic smart grid system will require an accurate and reliable state estimation of the low voltage network. For this to be achieved it will be necessary to have high resolution chronological and topological information of the network. The measurement data necessary for this state estimation can be obtained either from the distribution system measurement infrastructure or from smart meters installed at the customers connection points. The latter option will provide real time measurements of high accuracy; however it will also require a significant investment in communication infrastructure. The infrastructure requirements could be reduced by utilizing power flow pseudo measurements. This would reduce estimation errors whilst restraining costs. In consideration of the fact that the consumption and production of energy in the low voltage grid is highly volatile, Artificial Neural Networks (ANN) are used to generate these pseudo measurements. This paper demonstrates how ANNs can be used to create more accurate and dynamic pseudo measurements and to analyse their effect on the State Estimation of the low voltage network.