Anticipatory model predictive control with application to the coordination of electric power generation
Abstract
An interconnected electric power system, such as the Eastern interconnection in the United States, must generate and inject energy to the interconnection at the same rate which it is dissipated in loads and losses. If a difference exists in the rate at which energy is produced and consumed, energy will be borrowed from or stored as, the kinetic energy in the rotating machines on the interconnection, resulting in a proportional change in system frequency. Many consumers of electrical power require it to be delivered with the expected frequency (60 Hz in the United States). Maintaining constant frequency is the system-wide motivation for controlling generation such that it tracks demand. Generation control is performed in a distributed manor. Generation control areas are defined geographically, to supply power to the loads within their boundaries. The local motivation for generation control is to keep the inadvertent flow of power across a generation control area's boundary to a minimum. This local goal is congruent to the system-wide goal of frequency control. This thesis addresses the coordination and control of generation within a control area which contains large industrial customers. The demand for electrical power from highly varying industrial loads may change much faster than the control area generation's ability to track it. Anticipatory model predictive control has been applied to the automatic generation control problem. The control law previews a prediction of the power demand from the highly varying loads. This allows the generation to anticipate changes in demand, reducing inadvertent inter-area power flow, while satisfying generation unit rate constraints. The economic dispatch algorithm provides Model Predictive Control with the most efficient steady state unit generation levels. They are the target for its dynamic optimization. A robust controller is developed for generation unit set point control. A least squares estimator replaces remote real time demand measurement. A fuzzy predictor is developed for highly varying load power demand signals. Simulations illustrate the effectiveness of anticipatory automatic generation control system.
Degree
Ph.D.
Advisors
Shoureshi, Purdue University.
Subject Area
Mechanical engineering|Energy
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.