A Novel PID Autotuning Approach: How to Correct Bad Tuning by Closed-Loop Performance Assessment

Abstract

This paper proposes an autotuning approach for PID controllers based on the minimization of the nominal error, a benchmark used to determine if a model is discrepant with the plant and consequently linked to bad controller tuning. First, the closed-loop is assessed to ensure that the cause of poor performance is a discrepant model. Then, the proposed autotuning procedure is triggered to find new tuning parameters that fit in the new operation point that the process has shifted to. The main contribution of the method is that, rather than just working as a standard control performance assessment to identify if there are model mismatches, it also performs a controller tuning to eliminate them. In other words, it can automatically detect if the tuning parameters being used lost their applicability to the current operation point, and, if so, find optimum tuning parameters to configure the new closed-loop. In addition, the required process data may contain a period with setpoint changes or just the presence of unmeasured disturbances, thus being suited for both servo and regulatory operating modes.