On-line set-point optimisation and predictive control using neural Hammerstein models

Neural network adaptive real-time optimizing control of industrial processes

“…Figure 10 shows the movement of set points toward the optimum set point values verses the number of iterations. As the iteration number progresses, the two set-points 22 12 and c c whichare moving slowly toward the optimum values and lastly it stopped when the tolerance error for the optimum set points has been achieved …”

“…9 Some of the recent studies carried out on the application of ANN models with the RTO strategies have demonstrated the potential of these strategies in determine optimal operating conditions for large-scale processes. 10,11,12 The objective of this study is to explore the effect of the neural network process model application on the modified two step (MTS) technique 13 for overcoming the problem of plant-model mismatch. The MTS is a control algorithm that employs the adaptive modifier scheme for determining the optimal control set-points.…”

A steady-state target calculation method based on “point” model for integrating processes

“…1, which is composed of plant-wide optimization (a plant-wide optimizer determines optimal steady-state settings for each unit in the plant [6]), Local Steady-State Optimization (LSSO, which calculates on-line economically optimal set-points for the supervisory control layer in such a way that the production profit is maximized and the constraints are satisfied [16]), MPC and basic feedback control. The control action is performed by a number of regulators working at different time scales in hierarchical multilayer systems [6,16,17,18]. The MPC portion can be further divided into a SSTC and a Dynamic Optimization (DO, which is the rolling optimization in the MPC which has been applied extensively) [6].…”

Neural network based nonlinear model predictive control for an intensified continuous reactor

“…The nonlinear model is linearized at each sample instant and a linear model with state-space representation is obtained and applied for MPC. Lawrynczuk proposes a MPC algorithm based on linearization locally for neural network models, and applies it for polymerisation reactor [20], pH neutralization process [21] and yeast fermentation biochemical reactor [22].…”

“…Such as developing a neural network model, it is easy to make a large number of simulations to feed the neural network whereas it is more difficult to make the experiments that will also lack the same reproducibility. The empirical approaches include polynomial approximations (e.g., autoregressive moving average model with exogenous inputs, ARMAX) [13][14][15], artificial neural networks [16][17][18][19][20][21][22][23], piecewise linear models [24], Volterra series [25], Wiener and Hammerstein models [19][20][21], etc. In the above empirical models, neural network models have found wide applicability because of their inherent capability of handling complex and nonlinear problems and reducing the engineering effort required in controller model development.…”