An Improved Statistical Algorithm for Topology Identification and Parameter Estimation of Low-voltage Distribution Grids

Unplanned and arbitrary access to power supply and mistaken records of meter location pose challenges on topology identification and parameter estimation of low-voltage distribution grids for power utilities. With the advent of AMI and development of information communication technology, many statistical approaches such as correlation analysis and regression models have been implemented to unearth underlying features of distribution grids. In this paper, it is explained and then verified that merely the correlation analysis is incapable of parameter estimation of complicated grid structure and use of series-circuit regression model gives rise to risks of estimation failure. Hence, an improved statistical algorithm is proposed combining a unified parallel-circuit regression model and correlation analysis technology, along with a solution of virtual nodes generated in pairing process. With these improvements, connections between nodes are reflected more close to reality and we avoid the emergence of virtual nodes or unreal branches in the final output. The algorithm is presented in detail and tested in a low-voltage distribution grid with 85 meters, manifesting its good accuracy in topology identification and parameter estimation as well as its speeding of computation.