2023/04/13 by Samperio, Álvaro · 1 citation
#05C50 (Secondary) #94C15 (Primary) 05C85 #FOS: Mathematics #G.2.2 #Optimization and Control (math.OC)
paper · doi:10.48550/arxiv.2304.06676
The problem of recovering the topology and cable parameters of an electrical network from power and voltage data at all nodes is a problem of fitting both an algebraic variety and a graph. The problem is often ill-posed. In case there are multiple electrical networks which fit the data up to a given tolerance, we seek a solution in which the graph and therefore the algebraic equations associated with the electrical network are sparse, i.e. with few edges and terms. We propose an algorithm for recovering simultaneously a sparse topology and the cable parameters of any network, combining in an iterative procedure the resolution of non-negative linear regression problems, which are convex, and techniques of spectral graph sparsification. The algorithm is tested on several electrical networks.