Generating random power grids for the verification of new load flow solvers

The purpose of this thesis is to expand the rigor of the development of new power flow solvers through graph generation. The use of the IEEE standard test cases as benchmarks is commonplace in literature, where they are used to demonstrate the effectiveness of new algorithms. This results in the use of as little as two to five grids with only tens or hundreds of nodes each. The sample size for these tests is very small and cannot fully represent the behavior of the algorithms being tested. Since this problem stems from the lack of real, publicly available grids, a solution is to generate power grids with the necessary components. This thesis is the first to compare the performance of numerical methods in this setting. Two popular numerical methods are considered: the Newton-Raphson (NR) and Fast Decoupled Load Flow (FDLF) methods. It is found that with a modern direct matrix solver, NR is more efficient and robust than the FDLF when tested over several different topological factors. The results and methodology presented herein are used to test the speed and robustness of algorithms that solve similar power system problems such as Optimal Power Flow