Data Driven Tuning of Lower Level Controllers for Disturbance Rejections.
Systematic methods and tools for managing the complexity
Process Control (T4-8)
Keywords: PID tuning, Iterative Feedback Tuning
Optimal operation in chemical industry requires a (plant wide) model based, multivariable, control strategy such as Model Predictive Control (MPC). This optimal multivariable controller handles the interaction between the lower level loops. These loops are single input single output (SISO) control loops and their set points are set by the MPC according to the production objectives. For this multilevel control scheme to be effective, it is crucial that the lower level SISO controllers are working faster that the higher level MPC and that they are properly tuned for disturbance rejection.
Iterative Feedback Tuning (IFT) is a data driven method for optimization of the performance of simple controllers given a criterion expressed as a classic quadratic cost function. The method was first publish in [1] and have since been extended and tested for a number of applications [2]. IFT iteratively minimizes the cost function using a gradient based search method. The key feature of IFT is how to achieve the unbiased estimate of the gradient of the cost function with respect to the controller parameters. It an advantage that all data are collected from closed loop operation since it is the closed loop performance that is subject to optimization. The IFT method does not rely on external perturbation, due to a special design of the experiments which provide the gradient estimate. Unfortunately the IFT method only converges slowly when a controller is tuned for disturbance rejection problems. Therefore this contribution proposes to combine IFT with classic system identification and employ external perturbation in order to increase the information content in data. That leads to a faster convergence which renders the method amenable for a number of practical problems where the convergence of the standard formulation is too slow, which results in a too many of plant experiments.
The tuning of PID control loops of a four tank system [3] is considered as an example of an underlying control level. The liquid flow out of each tank has a nonlinear dependence on the level in tank, and tank system can be configured to exhibit different types of dynamics. The controllers are manipulating a pump in order to control the liquid level in the tanks. All variables are measured online, but the measurements are affected by noise. IFT tuning of the controllers for the four tank system with external perturbation are investigated and the disturbance rejection properties of the system are tested both in simulation and in practice on the pilot plant.
References
[1] H. Hjalmarsson, S. Gunnarsson & M. Gevers. A Convergent Iterative Restricted Complexity Control Design Scheme Proceedings of the 33rd IEEE Conference on Decision and Control, 1994, 2, 1735--1740
[2] H. Hjalmarsson Iterative feedback tuning - an overview International journal of adaptive control and signal processing, 2002, 16, 373--395
[3] K. H. Johanssen. The Quadruple-Tank Process: A Multivariable Laboratory Process with an Adjustable Zero IEEE Transactions on Control Systems Technology, 2000, 8, 456--465
