The design of power electronic and drive systems can be optimized to achieve specific requirements. Typical design goals are the maximization of efficiency (at rated or reduced power), minimization of volume/weight, and/or minimization of cost. In most cases, there does not exist a single solution that optimizes each of the design goals and compromises need to be made. Thus, the goal of an optimization is to find the solutions where a single feature cannot be improved without worsening others. These solutions are said to be Pareto optimal and form the Pareto front being equally good from an optimality standpoint.
An interesting and automated approach to optimize the design of power electronic and drive systems is the multi-objective optimization. The goals are written as a fitness function and a constraint set is defined to make sure the proposed solutions are feasible. Then, stochastic optimization techniques, e.g. particle swarm or genetic algorithms, are used to find optima. The optimizer generates design proposals, which are evaluated using system models. The type models depend on the level of detail and/or application. For example, it can be an interpolation of the volume of a capacitor series or a FEM model of an electrical machine.
In many cases, the evaluation of the fitness function takes a significant amount of time such that an optimization with high dimension is prohibitive. Thus, a design engineer has to assess the degrees of freedom that impact the result. Then, a decision maker can choose among the Pareto optimal solutions.