Decision support algorithms for power system and power electronic design
MetadataShow full item record
The thesis introduces an approach for obtaining higher level decision support information using electromagnetic transient (EMT) simulation programs. In this approach, a suite of higher level driver programs (decision support tools) control the simulator to gain important information about the system being simulated. These tools conduct a sequence of simulation runs, in each of which the study parameters are carefully selected based on the observations of the earlier runs in the sequence. In this research two such tools have been developed in conjunction with the PSCAD/EMTDC electromagnetic transient simulation program. The first tool is an improved optimization algorithm, which is used for automatic optimization of the system parameters to achieve a desired performance. This algorithm improves the capabilities of the previously reported method of optimization-enabled electromagnetic transient simulation by using an enhanced gradient-based optimization algorithm with constraint handling techniques. In addition to allow handling of design problems with more than one objective the thesis proposes to augment the optimization tool with the technique of Pareto optimality. A sequence of optimization runs are conducted to obtain the Pareto frontier, which quantifies the tradeoffs between the design objectives. The frontier can be used by the designer for decision making process. The second tool developed in this research helps the designer to study the effects of uncertainties in a design. By using a similar multiple-run approach this sensitivity analysis tool provides surrogate models of the system, which are simple mathematical functions that represent different aspects of the system performance. These models allow the designer to analyze the effects of uncertainties on system performance without having to conduct any further time-consuming EMT simulations. In this research it has been also proposed to add probabilistic analysis capabilities to the developed sensitivity analysis tool. Since probabilistic analysis of a system using conventional techniques (e.g. Monte-Carlo simulations) normally requires a large number of EMT simulation runs, using surrogate models instead of the actual simulation runs yields significant savings in terms of shortened simulation time. A number of examples have been used throughout the thesis to demonstrate the application and usefulness of the proposed tools.