Date of Award


Document Type

Open Access Dissertation


Electrical Engineering

First Advisor

Roger A Dougal


When trying to select an appropriate power generation plant for a micro-grid power distribution system like an electric ship, designers must consider both the physical characteristics (e.g., weight, volume, power ratings) and performance characteristics (e.g., fuel consumption, quality of service) of all the design alternatives. Comparing the design alternatives in terms of the physical characteristics is relatively straightforward, but in terms of performance characteristics each design alternative has to be evaluated within its own optimal performance points to make a fair comparison. However, at present no effective method or software tools exist to enable this evaluation at the earliest design stage.

To address this problem, we develop a concept evaluation method to determine the optimized power system concept of operations (CONOPS). Incorporating this method into the power generation plant development allows the design alternatives with undesirable performance to be removed from consideration, and ensures a high level of confidence that no quasi-optimal alternative is eliminated. The CONOPS in this dissertation takes into account the operating setpoints of the generating units on the primary power distribution buses. The optimality of a CONOPS is assessed with respect to its yielded system performance metrics, namely, fuel consumption and the quality of service (QOS). These two are paramount to the operating economy and mission success of micro-grid power systems. As an example, we apply our approach to the set-based

design (SBD) of a shipboard power system to demonstrate its effectiveness. Research is performed using a three step process.

First, we identify the full set of design variables that is applicable to generic power generation and distribution architectures, and use it to formulate the optimization problems of the CONOPS. The optimization problems fit both ac and dc distribution architecture and include the parameters that we identify as essential to describe the architecture. Also, we develop two QOS metrics to investigate the different aspects of system reliability: failure probability, and failure magnitude and duration.

Second, we develop and improve a single-objective particle swarm optimization (SOPSO) and multi-objective particle swarm optimization (MOPSO) to solve the optimization problems of the CONOPS. Both are able to provide enhanced capability and reliability of searching for the global optimum as compared with the previously reported PSOs. For a given system concept and mission (i.e. a description of loading conditions), the results derived by the SOPSO can rapidly reveal the performance tradeoffs of the CONOPS and investigate how the definitions of the performance metrics affect the optimal design of CONOPS. The results derived by the MOPSO, in contrast, help designers identify the quasi-optimal set of design alternatives during the SBD with a very high confidence level.

Third, in order to generalize the formulation process of the optimization problems for generic primary generation and distribution architectures and different expressions of the performance metrics, we develop an optimization structure based on the concept of control architecture. We define five broad categories of data to describe the essential parameters and design variables of the optimization problems common to a generic micro-grid power system application. We also identify the coupling relationship of these categories of data to standardize the co-optimization algorithm of the optimization problems. Therefore, we only need to develop one coding infrastructure that can be applicable to a wide range of design scenarios. In addition, we develop a hierarchical data structure to address the software implementation of this concept evaluation method during the SBD. This data structure contains two data exchange/flow block diagrams. One block diagram defines the data sharing method between the early stage models in S3D with an optimization simulation model in MATLAB. The other block diagram defines the data implementation process of resolving the co-optimization problem of the CONOPS in MATLAB. This data structure provides an effective guidance for software engineers to implement the concept evaluation method automatically by means of the two software environments.