Gradient-Based Optimization for Anti-Windup PID Controls.

The controller tuning process can be challenging and time consuming for complex and constrained systems. Over time, systems may degrade or operating conditions might change and, as a result, controllers may require an extensive re-tuning process. This work will address these tuning challenges by formulating a constrained optimization problem that considers both input and output constraints when identifying a set of optimal control gains for a PID controller. The optimization problem is solved efficiently using gradient-based methods, so it can be applied to a generic class of systems of varying complexity. When applied to a thermal management system, an optimized set of PID controllers decreases the required system flows rates by up to similar to 20%. A sensitivity analysis of the optimization results show that the choice of control gains is insensitive to thermal management systems with large heat sink heat transfer coefficients.