ANN-based correlation of measurements in micro-grid state estimation

This paper examines the influence of correlated pseudo measurements on the three-phase (sequence component-based) micro-grid state estimation. Pseudo measurements are used as the external inputs to replace the unavailable real-time measurements on distributed generation (DG) units and loads to provide the minimum bus observability. Output powers of unmonitored DG (photovoltaic and wind-based) units and loads are evaluated using the weather (either measured or forecasted) data, historically recorded state estimation patterns and available real-time measurements. The historical data are classified into clusters by the Self-Organization Map Artificial Neural Network (SOM ANN). The correlation coefficients between dependent pseudo measurements are calculated from clustered weather data and corresponding powers from DG units or loads, where the Feed Forward Artificial Neural Networks (FF ANNs) with backpropagation are used for approximating the output active power of unmonitored elements. The results and practical aspects of the proposed three-phase state estimation methodology with correlated measurements are demonstrated on two (benchmark and real-world) micro-grids.