Investigation of optimization methods used for reconfiguration of the naval electric ship power system

The future generation of United States naval electric ships will be completely powered by electricity and will have the ability to reconfigure the power system to adapt to the dynamic needs of the ship. This thesis explores four different numerical methods used to find the optimal switch state for the system. Each optimization routine was applied to three different case studies, each with different problem constraints. The methods were evaluated based on computational speed, accuracy in finding the best solution, and amount of setup time needed The method used depends heavily on the degree that the problem is constrained and the solution requirements. The direct search method is completely accurate and requires little setup time, but its slow speed motivates the search for quicker methods. The genetic algorithm performed quickly and found an accurate solution, but it is subject to multiple parameters whose roles are difficult to quantify. The binary integer procedure works very quickly and is easy to set up, but its solution accuracy is limited. The constrained simulated annealing method works well across a variety of problems. Although it is affected by many parameters, their role in shaping the method’s effectiveness is easier to empirically and conceptually understand compared to the genetic algorithm. If the percentage error must be less than two percent, the genetic algorithm is typically the fastest method. Otherwise, constrained simulated annealing is recommended due to its computational speed and setup time