This work details a novel method for assessing the performance of a PI (proportional-integral) feedback controller when the process displays non-self-regulating dynamic behavior. By applying an intuitive process control-based pattern recognition method to the autocorrelation function of the process measurement signal, the controller's disturbance rejection performance can automatically be categorized. Stochastic data collected over days or weeks is analyzed to compute an index descriptive of current controller performance. If the control response has drifted from a user-defined target value, the analysis further provides a guide for tuning adjustments to restore desired performance. Significant aspects of this approach are that no plant disruption or process knowledge is required for evaluation. Classic examples of non-self-regulating behavior include certain liquid level control loops and pressure control loops which are prevalent in cogeneration power plants. In this work, we detail how the performance assessment method was used to improve performance of such controllers in the University of Connecticut's power plant. (C) 2009 Elsevier Ltd. All rights reserved.