Automated Synthesis Tool for Design Optimization of Power Electronic Converters

Designers of power electronic converters usually face the challenge of having multiple performance indices that must be simultaneously optimized, such as maximizing efficiency while minimizing mass or maximizing reliability while minimizing cost. The experienced engineer applies his or her judgment to reduce the number of possible designs to a manageable number of feasible designs for which to prototype and test; thus, the optimality of this design-space reduction is directly dependent upon the experience, and expertise and biases of the designer. The practitioner is familiar with tradeoff analysis; however, simple tradeoff studies can become difficult or even intractable if multiple metrics are considered. Hence a scientific and systematic approach is needed. In this dissertation, a multi-objective optimization framework is presented as a design tool.

Optimization of power electronic converters is certainly not a new subject. However, when limited to off-the-shelf components, the resulting system is really optimized only over the set of commercially available components, which may represent only a subset of the design space; the reachable space limited by available components and technologies. While this approach is suited to cost-reduce an existing design, it offers little insight into design possibilities for greenfield projects.

Instead, this work uses the Technology Characterization Methods (TCM) to broaden the reachable design space by considering fundamental component attributes. The result is the specification for the components that create the optimal design rather than an evaluation of an apriori selected set of candidate components. A unique outcome of this approach is that new technology development vectors may emerge to develop optimized components for the optimized power converter. The approach presented in this work uses a mathematical descriptive language to abstract the characteristics and attributes of the components used in a power electronic converter in a way suitable for multi-objective and constrained optimization methods.

This dissertation will use Technology Characterization Methods (TCM) to bridge the gap between high-level performance attributes and low-level design attributes where direct relationship between these two does not currently exist. The loss and size models for inductors, capacitors, IGBTs, MOSFETs and heat sinks will be used to form objective functions for the multi-objective optimization problem. A single phase IGBT-based inverter is optimized for efficiency and volume based on the component models derived using TCM. Comparing the obtained designs to a design, which can be made from commercial off-the-shelf components, shows that converter design can be optimized beyond what is possible from using only off-the-shelf components. A module-integrated photovoltaic inverter is also optimized for efficiency, volume and reliability. An actual converter is constructed using commercial off-the-shelf components. The converter design is chosen as close as possible to a point obtained by optimization. Experimental results show that the converter modeling is accurate. A new approach for evaluation of efficiency in photovoltaic converter is also proposed and the front-end portion of a photovoltaic converter is optimized for this efficiency, as well as reliability and volume.