Novel Formulation of Adaptive MPC as EKF Using ANN Model: Multiproduct Semibatch Polymerization Reactor Case Study.

In this paper, a novel formulation for nonlinear model predictive control (MPC) has been proposed incorporating the extended Kalman filter (EKF) control concept using a purely data-driven artificial neural network (ANN) model based on measurements for supervisory control. The proposed scheme consists of two modules focusing on online parameter estimation based on past measurements and control estimation over control horizon based on minimizing the deviation of model output predictions from set points along the prediction horizon. An industrial case study for temperature control of a multiproduct semibatch polymerization reactor posed as a challenge problem has been considered as a test bed to apply the proposed ANN-EKFMPC strategy at supervisory level as a cascade control configuration along with proportional integral controller [ANN-EKFMPC with PI (ANN-EKFMPC-PI)]. The proposed approach is formulated incorporating all aspects of MPC including move suppression factor for control effort minimization and constraint-handling capability including terminal constraints. The nominal stability analysis and offset-free tracking capabilities of the proposed controller are proved. Its performance is evaluated by comparison with a standard MPC-based cascade control approach using the same adaptive ANN model. The ANN-EKFMPC-PI control configuration has shown better controller performance in terms of temperature tracking, smoother input profiles, as well as constraint-handling ability compared with the ANN-MPC with PI approach for two products in summer and winter. The proposed scheme is found to be versatile although it is based on a purely data-driven model with online parameter estimation.

Modelling of pretreatment and saccharification with different feedstocks and kinetic modeling of sorghum saccharification.

Experiments have been performed for pretreatment of sorghum, wheat straw and bamboo through high temperature alkali pretreatment with different alkaline loading and temperatures, and the data on extent of delignification in terms of the final compositions of cellulose, hemicellulose and lignin have been generated. Further, enzymatic saccharification has been carried out in all the cases to find the extent of conversion possible after 72h. The effect of different operating parameters on the extent of delignification and cellulose conversion are evaluated. This data is employed to develop a generalized multi-feedstock and individual feedstock based models which can be used to determine the extent of delignification and cellulose conversion for any and specific biomass respectively with alkaline pretreatment and similar enzyme conditions as considered in the present study. Also, a kinetic model is developed and validated for sorghum for cellulosic conversion.

Parameterized data-driven fuzzy model based optimal control of a semi-batch reactor.

A parameterized data-driven fuzzy (PDDF) model structure is proposed for semi-batch processes, and its application for optimal control is illustrated. The orthonormally parameterized input trajectories, initial states and process parameters are the inputs to the model, which predicts the output trajectories in terms of Fourier coefficients. Fuzzy rules are formulated based on the signs of a linear data-driven model, while the defuzzification step incorporates a linear regression model to shift the domain from input to output domain. The fuzzy model is employed to formulate an optimal control problem for single rate as well as multi-rate systems. Simulation study on a multivariable semi-batch reactor system reveals that the proposed PDDF modeling approach is capable of capturing the nonlinear and time-varying behavior inherent in the semi-batch system fairly accurately, and the results of operating trajectory optimization using the proposed model are found to be comparable to the results obtained using the exact first principles model, and are also found to be comparable to or better than parameterized data-driven artificial neural network model based optimization results.