| This thesis analyzes and tests some new solution techniques for the optimal power flow problem. This new methodology exploits a parametric technique, called the continuation method, which is applied to different tasks in the solution procedure. In a first application, the continuation method solves the quadratic subproblems generated sequentially by the optimal power flow's nonlinear program. It first creates a simple subproblem, which is easy to solve, and then links it to the subproblem we wish to solve. Starting at the solution of the simple problem, it generates optimal solution trajectories for the intermediate problems, leading to the desired optimal solution. In a second application, the algorithm tracks optimal solutions trajectories of the nonlinear problem when the load is slowly varied. This constitutes an example of "incremental loading", a technique already used for real power dispatch, but in this case a complete network model is used. The flexibility of the algorithm at various levels allows for some excellent computation times in this load-tracking mode: we have observed reductions in computation times for new solutions of the order of 70%, compared to the computation time of the initial load.;Numerical simulations of the proposed optimal power flow algorithm using the minimum fuel cost task were performed on four test systems, with sizes ranging from 6 to 118 buses. The results are documented in detail, and results for the 30 bus test are compared to those reported by other authors. All in all, our results demonstrate quite well the potential of this technique. (Abstract shortened with permission of author.).;This thesis first presents an analysis of the various structures used in optimal power flow algorithms. Then, having chosen and presented the structure of our algorithm, we analyze the quadratic subproblems generated by this algorithm for some of its more important tasks: minimum cost, minimum losses and load shedding. New rules are proposed to link the solutions of successive subproblems to ensure the convergence of the nonlinear problem. Then, as a final contribution to the theory, some extensions are suggested for the subproblems: among them are ramp constraints, bus incremental costs, and provisions for redispatching. |
