Solar farm cable layout optimization as a graph problem

Abstract

We introduce the Solar Farm Cable Layout Problem (SoFaCLaP), a novel graph-theoretic optimization problem. SoFaCLaP formalizes the task of finding a cost-optimal cable layout in a solar farm where PV string positions are already determined but the positions of other components such as transformers can be picked from a set of candidate positions. The problem statement incorporates a network flow model in which the flow value of a connection represents the number of strings that are (indirectly) connected to a transformer via this connection. A mixed-integer linear program (MILP) formulation is proposed that uses binary variables to indicate which of several available cable types is chosen for each connection. We propose a framework to randomly generate benchmark instances to evaluate any algorithmic approach to SoFaCLaP. In particular, we generate a set of instances based on real-world solar farm characteristics. With an extensive evaluation of the MILP formulation on those instances we establish mixed-integer linear programming as a baseline for future algorithmic approaches to finding solar farm cable layouts.