Data-Driven PID Controller and Its Application to Pulp Neutralization Process

Yajun Zhang,Yao Jia,Tianyou Chai,Dianhui Wang,Wei Dai,Jun Fu
DOI: https://doi.org/10.1109/tcst.2017.2695981
IF: 4.8
2018-01-01
IEEE Transactions on Control Systems Technology
Abstract:The pulp neutralization process is a complex industrial process with strong nonlinearity and time-varying dynamics. In general, this kind of process is hardly controlled by using traditional control techniques such as the proportional-integral-derivative (PID) controller. To resolve this problem, a novel PID control scheme is proposed by integrating a data-driven compensation of unmodeled dynamics and a multistep ahead of optimal control strategy. Essential difference from the classical PID controller lies in the use of unmodeled dynamics, the measured data from the pulp neutralization process is used directly for building the nonlinear PID controller, which plays an important role in control system design. Based on multistep ahead prediction optimal control theory, the controller parameters can be obtained through optimizing a cost function. Some theoretical results on the stability and convergence of the closed-loop system are established. To demonstrate the effectiveness of the proposed control techniques, a simulation model and a cobalt nickel hydroxide pulp neutralization process are employed. Simulations were carried out using data extracted from an actual pulp neutralization process, followed by some industrial experiments. Results from both simulations and experiments indicate that our proposed controller can effectively control the pH value of the pulp neutralization process with promising control performance.
What problem does this paper attempt to address?