Yunzhi Cheng

Graduation Semester and Year




Document Type


Degree Name

Doctor of Philosophy in Electrical Engineering


Electrical Engineering

First Advisor

Wei-Jen Lee


Following the power market deregulation, power systems have become more complex and are found to be consistently operating closer to their stability limits. Power system dynamic modeling and studies which provide significant insight into the dynamic characteristics of the system are bound to play increasingly critical roles. The dynamic simulation results are highly dependent on certain key parameters that govern the dynamics of the power system such as governors and/or exciters in the case of generation facilities. However, the dynamic parameters in the system database maintained by the Independent System Operator (ISO) are not accurate due to numerous reasons ranging from data submissions not corresponding to "as-built facilities" to data not being updated to reflect changes at the facility. Such inconsistencies in the dynamic models utilized to represent actual system facilities have led to tremendous research in the field of dynamic parameter estimation. Numerous algorithms have been proposed for dynamic parameter estimation. The conventional gradient-based optimization approach suffers from an obvious and inherent dependency on the initial conditions and is found to have convergence problems when starting with a poor initial guess. On the other hand, some inherently initial-value independent intelligent methods suffer from tremendous computation burden. This dissertation proposes a hybrid two-step method to achieve the accurate dynamic parameters in a balanced manner by making an optimal trade-off between convergence and computation speed. The concept of Particle Swarm Optimization (PSO) is employed to find an approximate solution at the first step, followed by a sensitivity analysis is run to achieve an accurate solution starting with the approximate solution obtained in the first step. This dissertation describes how various categories are set up for the dynamic parameters and identifies the key parameters for parameter estimation to decrease the complexity of the problem and computation burden. While the approach documented in this dissertation is generic in terms of applicability to dynamic parameter estimation, the generator dynamic parameters have been utilized to illustrate the efficiency of the approach. All exciter and governor models in the Electrical Reliability Council of Texas (ERCOT) system are pre-scanned to identify the key parameters using the PSS/E response test. The proposed hybrid method shows the validity and distinct advantages in the assumed test case. The exciter and governor parameters are successfully estimated using the proposed hybrid method. Reasonably accurate values can be achieved under some level of noise according to uncertainty analysis. Multi-core computation is utilized to dramatically decrease the computation burden. The proposed hybrid method also successfully tunes the dynamic parameters of exciter and power system stabilizer (PSS) in a power plant to drive the trend of simulation results to match the recording information on file following a generator trip in ERCOT system.


Electrical and Computer Engineering | Engineering


Degree granted by The University of Texas at Arlington