Parameter Estimation in Electrical Distribution Systems with limited Measurements using Regression Methods

This paper presents novel methods for parameter identification in electrical grids with small numbers of spatially distributed measuring devices, which is an issue for distribution system operators managing aged and not properly mapped underground Low Voltage (LV) grids, especially in Germany. For this purpose, the total impedance of individual branches of the overall system is estimated by measuring currents and voltages at a subset of all system nodes over time. It is shown that, under common assumptions for electrical distsribution systems, an estimate of the total impedance can be made using readily computable proxies. Different regression methods are then used and compared to estimate the total impedance of the respective branches, with varying weights of the input data. The results on realistic LV feeders with different branch lengths and number of unmeasured segments are discussed and multiple influencing factors are investigated through simulations. It is shown that estimates of the total impedances can be obtained with acceptable quality under realistic assumptions.