Automatic planning tools for power system design

Abstract

The purpose of this thesis is to investigate different techniques for designing electrical power networks and demonstrate how algorithms can be used and combined in order to solve two power network design subproblems.

In the first part, we focus on the problem of connecting a new energy consumer to an existing electrical grid based on its distance to the possible connection points. The terrain is divided into convex or concave polygons, each having a cost for laying cable within its boundary. To solve this problem we use an approach based on Dijkstra’s algorithm in order to produce close to optimal solutions. For this particular problem we conducted computational performance experiments on two identical implementations written in C++ and PowerFactory. The results support the view that C++ is a much more efficient programming language than the Digsilent Programming Language (DPL) used in PowerFactory.

In the second part of the thesis, we analyze the problem of routing electricity from producers to consumers in a simplified electrical network, given that connections are characterized by capacity and cost. The goal is to find a solution which maximizes the amount of energy sent through the network (i.e. satisfy as much demand as the network can handle). If there are multiple such solutions, we are interested in the one with the minimum cost. The cost of sending energy is determined by the cost of acquiring connections and the cost of sending energy through those connections. We propose a branch-and-bound approach which obtains an exact solution, a simulated annealing approach and a heuristic which varies the cost of the connections in order to explore the solution space.