No abstract

In this paper, we show the implementation of deep neural networks applied in process control. In our approach, we based the training of the neural network on model predictive control. Model predictive control is popular for its ability to be tuned by the weighting matrices and by the fact that it respects the constraints.We present the neural network that can approximate the behavior of the MPC in the way of mimicking the control input trajectory while the constraints on states and control input remain unimpaired of the value of the weighting matrices.This approach is demonstrated in a simulation case study involving a continuous stirred tank reactor, where multi-component chemical reaction takes place.

No abstract

No abstract

Rapid growth of the human population has led to various problems, such as massive overload of wastewater treatment plants. Therefore, optimal control of these plants is a relevant subject. This contribution analyses control of a cascade of ten biochemical reactors using simulation results with the aim to design optimal and predictive control strategies and to compare the achieved control performance. The plant represents a complicated process with many variables involved in the model structure, reduced to the single-input and single-output system. The first implemented approach is linear offset-free model predictive control which provides the optimal input trajectory minimising a quadratic cost function. The second control strategy is robust model predictive control with similar features as model predictive control but including the uncertainty of the process. The final approach is generalised predictive control, mostly used in the industry because of its simple structure and sufficiently good control performance. All considered predictive controllers provide satisfactory control performance and remove the steady-state control error despite the constrained control inputs.