Sensor placement for optimal estimation in power distribution grids

Abstract

The problem of placing sensors in distribution grids for optimal state estimation is studied. An improved formulation is proposed where the estimation error of any resulting electrical quantity is optimised, as opposed to traditional methods optimising the estimation quality in the state variables only. As a result, the solution is more robust, since the result is not specific to the arbitrary choice of the state variable, and more flexible, since any desired weight can be assigned to the estimation error of all possible electrical quantities in the network. A greedy algorithm with convergence guarantees is also introduced based on general results on submodular functions.