Optimized four-parameter PID controller for AVR systems with respect to robustness

A procedure for the optimization of the proportional-integral-derivative (PID) controller under constraints to robustness of model uncertainties and sensitivity to measurement noise for a synchronous generator (SG) automatic voltage regulator (AVR) is presented in this paper. Performance indices related to the load disturbance response such as integral error (IE) and integral absolute error (IAE) are used as key criteria for the design method. They are minimized for a given maximum modulus of sensitivity function Ms. By maximizing loop gain regarding given sensitivities Ms and Mn we optimized load disturbance response. Contrary to majority AVR PID controller procedures, along with proportional, integral and derivative gain, a noise filter time constant is included in a proposed optimization algorithm to satisfy given sensitivity to the measurement noise Mn. A simple and effective solution of the optimization problem is obtained by binding a ratio between the integral and the derivative time constant Ti/Td. The optimal Ti/Td ratio is determined subsequently. Besides this, a step response to voltage reference change is modified with a reference filter to avoid large overshoots. This is achieved by limiting the maximum modulus of the resulting close-loop transfer function sensitivity Msp. The fact that zeros and poles of the second order reference filter have the same damping factor simplifies the procedure for resolving two out of six parameters. Also, two additional conditions result in the fact that three design parameters {Ms, Mn, Msp} are sufficient to determine the required six parameters of the two-degree-of-freedom (2DOF) PID controller. Such controller outperforms recently presented AVR controllers especially regarding the load disturbance rejection. Experiment tests with no-load SG and SG connected on the grid are carried out to prove the claims stated in this work concerning robustness performance and sensitivity to measurement noise. Also, we showed practical use of the proposed controller parameter tuning procedure by experimental runs performed on a lab generator.