Accounting for “mission” during co-optimization of system designs

A simulation-based approach to optimal design of systems requires collective solution to three related problems: selection of the best equipment, selection of the best system configuration, and optimal operation of the system in a mission-oriented sense. We solve these three problems by applying an improved Particle Swarm Optimization (PSO) algorithm on top of a toolbox that rapidly instantiates system models from high level specifications and system templates. The improved PSO more efficiently handles the involvement of discrete binary variables in the objective functions with universal constraints, more effectively avoids premature convergence, and yields a more accurate search for the global optimum solution. Our simulation-component-based co-optimization approach to system designs is illustrated by applying it to the design of an electric ship power system while accounting for the unit commitment problem during a set of missions. First, we prove the efficacy of the new PSO algorithm on one candidate system, and then we apply the new PSO method to evaluate and compare three candidate designs. The improved PSO shows that the system should actually consume up to 52% less fuel than is predicted by a simpler evaluation that invokes a proportionally-distributed power alignment; this is a much better measure of the true system performance. As compared with another validated hybrid PSO-GA algorithm, the improved PSO consistently finds an optimum operating point that yields up to 2.4% better fuel efficiency, and it improves the reliability of solutions by up to 32%. Next, three candidate designs are evaluated under optimal operating conditions, for a given mission profile, and compared in terms of their best performance. The simulation results successfully determine the best genset combination from the competing designs.