Control of polystyrene batch reactors using neural network based model predictive control (NNMPC): An experimental investigation

Controlling batch polymerization reactors imposes great operational difficulties due to the complex reaction kinetics, inherent process nonlinearities and the continuous demand for running these reactors at varying operating conditions needed to produce different polymer grades. Model predictive control (MPC) has become the leading technology of advanced nonlinear control adopted for such chemical process industries. The usual practice for operating polymerization reactors is to optimize the reactor temperature profile since the end use properties of the product polymer depend highly on temperature. This is because the end use properties of the product polymer depend highly on temperature. The reactor is then run to track the optimized temperature set-point profile. In this work, a neural network-model predictive control (NN-MPC) algorithm was implemented to control the temperature of a polystyrene (PS) batch reactors and the controller set-point tracking and load rejection performance was investigated. In this approach, a neural network model is trained to predict the future process response over the specified horizon. The predictions are passed to a numerical optimization routine which attempts to minimize a specified cost function to calculate a suitable control signal at each sample instant. The performance results of the NN-MPC were compared with a conventional PID controller. Based on the experimental results, it is concluded that the NN-MPC performance is superior to the conventional PID controller especially during process startup. The NN-MPC resulted in smoother controller moves and less variability.     

A distributed model predictive control (MPC) fault reconfiguration strategy for formation flying satellites

In this paper, an active distributed (also referred to as semi-decentralised) fault recovery control scheme is proposed that employs inaccurate and unreliable fault information into a model-predictive-control-based design. The objective is to compensate for the identified actuator faults that are subject to uncertainties and detection time delays, in the attitude control subsystems of formation flying satellites. The proposed distributed fault recovery scheme is developed through a two-level hierarchical framework. In the first level, or the agent level, the fault is recovered locally to maintain as much as possible the design specifications, feasibility, and tracking performance of all the agents. In the second level, or the formation level, the recovery is carried out by enhancing the entire team performance. The fault recovery performance of our proposed distributed (semi-decentralised) scheme is compared with two other alternative schemes, namely the centralised and the decentralised fault recovery schemes. It is shown that the distributed (semi-decentralised) fault recovery scheme satisfies the recovery design specifications and also imposes lower fault compensation control effort cost and communication bandwidth requirements as compared to the centralised scheme. Our proposed distributed (semi-decentralised) scheme also outperforms the achievable performance capabilities of the decentralised scheme. Simulation results corresponding to a network of four precision formation flight satellites are also provided to demonstrate and illustrate the advantages of our proposed distributed (semi-decentralised) fault recovery strategy.     

Cooperative vehicle path generation during merging using model predictive control with real-time optimization

This paper proposed a cooperative merging path generation method for vehicles to merge smoothly on the motorway using a Model Predictive Control (MPC) scheme which optimizes the motions of the relevant vehicles simultaneously. The cooperative merging is a merging in where the most relevant vehicle in the main lane would accelerate or decelerate slightly to let the merging vehicle merge in easily. The proposed path generation algorithm can generate the merging path ensuring the merging vehicle can access the whole acceleration area, and do not exceed it. We have introduced a state variable to the optimization problem by which the merging point for the merging vehicle is optimized. The simulation results showed that the cooperative merging path can be successfully generated under some typical traffic situations without re-adjustment of the optimization parameters.     

Optimal control of grinding mill circuit using model predictive static programming: A new nonlinear MPC paradigm

The recently developed reference-command tracking version of model predictive static programming (MPSP) is successfully applied to a single-stage closed grinding mill circuit. MPSP is an innovative optimal control technique that combines the philosophies of model predictive control (MPC) and approximate dynamic programming. The performance of the proposed MPSP control technique, which can be viewed as a ‘new paradigm’ under the nonlinear MPC philosophy, is compared to the performance of a standard nonlinear MPC technique applied to the same plant for the same conditions. Results show that the MPSP control technique is more than capable of tracking the desired set-point in the presence of model-plant mismatch, disturbances and measurement noise. The performance of MPSP and nonlinear MPC compare very well, with definite advantages offered by MPSP. The computational speed of MPSP is increased through a sequence of innovations such as the conversion of the dynamic optimization problem to a low-dimensional static optimization problem, the recursive computation of sensitivity matrices and using a closed form expression to update the control. To alleviate the burden on the optimization procedure in standard MPC, the control horizon is normally restricted. However, in the MPSP technique the control horizon is extended to the prediction horizon with a minor increase in the computational time. Furthermore, the MPSP technique generally takes only a couple of iterations to converge, even when input constraints are applied. Therefore, MPSP can be regarded as a potential candidate for online applications of the nonlinear MPC philosophy to real-world industrial process plants.     

Robust model predictive control for continuous uncertain systems with state delay

This paper is concerned with robust model predictive control for linear continuous uncertain systems with state delay and control constraints, A piecewise constant control sequence is calculated by minimizing the upper-bound of the infinite horizon quadratic cost function, At each sampling time, the sufficient conditions for the existence of the model predictive control are derived, and expressed as a set of linear matrix inequalities. The robust stability of the closed-loop svstems is guaranteed bv the proposed design method. A numerical example is given to illustrate the main results.     

Robust model predictive control for continuous uncertain systems with state delay

This paper is concerned with robust model predictive control for linear continuous uncertain systems with state delay and control constraints. A piecewise constant control sequence is calculated by minimizing the upper-bound of the infinite horizon quadratic cost function. At each sampling time, the sufficient conditions for the existence of the model predictive control are derived, and expressed as a set of linear matrix inequalities. The robust stability of the closed-loop systems is guaranteed by the proposed design method. A numerical example is given to illustrate the main results.     

Constraint-softening in model predictive control with off-line-optimized admissible sets for systems with additive and multiplicative disturbances

An approach to the softening of constraints is explored for a class of MPC algorithms that employ off-line-computed constraint-admissible sets for simplified on-line computations. The proposed approach relies on the use of exact penalty functions to ensure that the solution coincides with the actual optimal solution if the original MPC problem is feasible and that there are constraint violations at minimum possible levels if the original problem is infeasible. The approach is implemented for a class of linear systems with additive and multiplicative disturbances using a dynamic-policy-based MPC algorithm. Results specific to the cases of non-stochastic and stochastic disturbances are explored and assessed with simulation examples.     

电流型变流器的改进模型预测控制

传统的模型预测控制(MPC)采用脉冲响应的非参数模型作为系统的预测控制模型,计算量大,很难直接应用于实时控制系统.本文提出将传统的MPC进行改进,应用到电流型变流器(CSC)功率因数校正系统中.改进的MPC根据CSC控制量与被控制量的传递函数得出CSC的一阶差分方程作为预测控制模型,同时保留传统MPC反馈校正、动态优化等优点.将此改进的MPC与传统的PID控制进行对比研究,结果表明:改进的MPC用于电流型变流器,比传统的PID控制具有更好的鲁棒性和更快速的动态响应特性.     

Event-triggered robust model predictive control of continuous-time nonlinear systems

The event-triggered control is of compelling features in efficiently exploiting system resources, and thus has found many applications in sensor networks, networked control systems, multi-agent systems and so on. In this paper, we study the event-triggered model predictive control (MPC) problem for continuous-time nonlinear systems subject to bounded disturbances. An event-triggered mechanism is first designed by measuring the error between the system state and its optimal prediction; the event-triggered MPC algorithm that is built upon the triggering mechanism and the dual-mode approach is then designed. The rigorous analysis of the feasibility and stability is conducted, and the sufficient conditions for ensuring the feasibility and stability are developed. We show that the feasibility of the event-triggered MPC algorithm can be guaranteed if, the prediction horizon is designed properly and the disturbances are small enough. Furthermore, it is shown that the stability is related to the prediction horizon, the disturbance bound and the triggering level, and that the state trajectory converges to a robust invariant set under the proposed conditions. Finally, a case study is provided to verify the theoretical results.     

Intensive insulin therapy for critically ill subjects based on direct data-driven model predictive control

Patients in the intensive care units (ICU) can suffer from stress-induced hyperglycemia, which can result in negative outcomes and even death. Recent studies show that, regulation of blood glucose (BG) brings in improved outcomes. In this study, a novel direct data-driven model predictive control (MPC) strategy is developed to tightly regulate BG concentration in the ICU. The effectiveness of the proposed direct data-driven MPC strategy is validated on 30 virtual ICU patients, and the in silico results demonstrate the proposed method's excellent robustness with respect to intersubject variability and measurement noises. In addition, the mean percentage values in A-zone of the control variability grid analysis (CVGA) plots are 14% under the Yale protocol, 67% under the combination of particle swarm optimization (PSO) and MPC method (for short, termed as PSO–MPC method), and 90% under the proposed method. In summary, as a good candidate for full closed-loop glycemic control algorithm, the proposed method has superior performance to the nurse-driven Yale protocol and the closed-loop PSO–MPC method.     

Zone model predictive control for pressure management of water distribution network

In the water distribution network (WDN), although the water demand of the node is changing constantly, the water quantity and water pressure of the node need to be met at each moment. To realize energy saving and consumption reduction, it is proposed to control the nodal water head of WDN in an appropriate narrow range. The frequent large fluctuation of the water demand, which may lead the water pressure exceed the expected range, increases the difficulty of the zone control. To realize optimized WND control, a novel zone predictive control is proposed, where two switching cases are considered. The switching condition is whether there are feasible solutions to keep the pressure within the expected region over the prediction horizon. When the condition is satisfied, the controller minimizes the variation of inputs with constraints of pressure range for ensuring the tank level staying within the expected zone and obtaining optimal economic cost. When the current pressure is out of the expected region or the condition is not satisfied due to the large variation of water demand, a reference trajectory of outputs is introduced, which is combined with the inputs as an optimization variable, and the constraints of expected zone are moved from the output to the introduced reference trajectory. Through minimizing the distance between reference trajectory and output, the controller will keep the tank level from deviating too far from the expected zone and will drive the tank level rapidly into the expected zone once the tank level exceeds the expected range. An application of the proposed zone MPC to WDN in Shinan District of Shanghai is given to illustrate its effectiveness.     

A design method for indirect iterative learning control based on two-dimensional generalized predictive control algorithm

Indirect iterative learning control (ILC) facilitates the application of learning-type control strategies to the repetitive/batch/periodic processes with local feedback control already. Based on the two-dimensional generalized predictive control (2D-GPC) algorithm, a new design method is proposed in this paper for an indirect ILC system which consists of a model predictive control (MPC) in the inner loop and a simple ILC in the outer loop. The major advantage of the proposed design method is realizing an integrated optimization for the parameters of existing feedback controller and design of a simple iterative learning controller, and then ensuring the optimal control performance of the whole system in sense of 2D-GPC. From the analysis of the control law, it is found that the proposed indirect ILC law can be directly obtained from a standard GPC law and the stability and convergence of the closed-loop control system can be analyzed by a simple criterion. It is an applicable and effective solution for the application of ILC scheme to the industry processes, which can be seen clearly from the numerical simulations as well as the comparisons with the other solutions.     

On the computation of linear model predictive control laws

Finite-time optimal control problems with quadratic performance index for linear systems with linear constraints can be transformed into Quadratic Programs (QPs). Model Predictive Control requires the on-line solution of such QPs. This can be obtained by using a QP solver or evaluating the associated explicit solution. The objective of this note is twofold. First, we shed some light on the computational complexity and storage demand of the two approaches when an active set QP solver is used. Second, we show the existence of alternative algorithms with a different tradeoff between memory and computational time. In particular, we present an algorithm which, for a certain class of systems, outperforms standard explicit solvers both in terms of memory and worst case computational time.     

Distributed nonlinear model predictive control based on contraction theory

A novel distributed model predictive control algorithm for continuous‐time nonlinear systems is proposed in this paper. Contraction theory is used to estimate the prediction error in the algorithm, leading to new feasibility and stability conditions. Compared to existing analysis based on Lipschitz continuity, the proposed approach gives a distributed model predictive control algorithm under less conservative conditions, allowing stronger couplings between subsystems and a larger sampling interval when the subsystems satisfy the specified contraction conditions. A numerical example is given to illustrate the effectiveness and advantage of the proposed approach.     

Synthesizing scheduled robust model predictive control with target tracking

Systems with large operating regions and non-zero state target tracking have limited the industrial application of robust model predictive control (RMPC) with synthetic action. To overcome the problem, this paper presents a novel formulation of synthesizing scheduled RMPC for linear time varying (LTV) systems. Off-line, we compute the matrix that transforms target output into steady state first. Then a set of stabilizing state feedback laws which are corresponding to a set of estimated regions of stability covering the desired operating region are provided. On-line, these control laws are implemented as a single scheduled state feedback model predictive control (MPC) which switches between the set of local controllers and achieve the desired target at last. Finally, the algorithm is illustrated with an example.     

Distributed model predictive control of constrained weakly coupled nonlinear systems

This paper proposes a distributed model predictive control (MPC) strategy for a large-scale system that consists of several dynamically coupled nonlinear systems with decoupled control constraints and disturbances. In the proposed strategy, all subsystems compute their control signals by solving local optimizations constrained by their nominal decoupled dynamics. The dynamic couplings and the disturbances are accommodated through new robustness constraints in the local optimizations. The paper derives relationships among, and designs procedures for, the parameters involved in the proposed distributed MPC strategy based on the analysis of the recursive feasibility and the robust stability of the overall system. The paper shows that, for a given bound on the disturbances, the recursive feasibility is guaranteed if the sampling interval is properly chosen. Moreover, it establishes sufficient conditions for the overall system state to converge to a robust positively invariant set. The paper illustrates the effectiveness of the proposed distributed MPC strategy by applying it to three coupled cart-(nonlinear) spring–damper subsystems.     

Decentralized tube‐based model predictive control of uncertain nonlinear multiagent systems

This paper addresses the problem of decentralized tube‐based nonlinear model predictive control (NMPC) for a general class of uncertain nonlinear continuous‐time multiagent systems with additive and bounded disturbance. In particular, the problem of robust navigation of a multiagent system to predefined states of the workspace while using only local information is addressed under certain distance and control input constraints. We propose a decentralized feedback control protocol that consists of two terms: a nominal control input, which is computed online and is the outcome of a decentralized finite horizon optimal control problem that each agent solves at every sampling time, for its nominal system dynamics; and an additive state‐feedback law which is computed offline and guarantees that the real trajectories of each agent will belong to a hypertube centered along the nominal trajectory, for all times. The volume of the hypertube depends on the upper bound of the disturbances as well as the bounds of the derivatives of the dynamics. In addition, by introducing certain distance constraints, the proposed scheme guarantees that the initially connected agents remain connected for all times. Under standard assumptions that arise in nominal NMPC schemes, controllability assumptions, communication capabilities between the agents, it is guaranteed that the multiagent system is input‐to‐state stable with respect to the disturbances, for all initial conditions satisfying the state constraints. Simulation results verify the correctness of the proposed framework.     

基于线性二次型高斯基准的多变量预测控制技术经济性能评估

由于受控过程参数的漂移及缺乏维护,令采用的控制器性能逐渐降低,需要做经济性能评估,以确保其最佳运行状态.因为目前最小方差评估算法没有考虑控制器的约束条件,对此我们采用线性二次型高斯(linearquadratic Gaussian,LQG)基准的模型预测控制(model predictive control,MPC)双层优化控制结构,将控制和输出的加权值引入上层经济性能指标,通过求解LQG问题获取控制与输出方差关系的离散点集,进一步拟合Pareto最优曲面方程,建立优化命题并求解最优经济指标及设定值.对延迟焦化加热炉的多变量MPC控制进行了性能评估及分析,证明该方法可以改进控制器设计,提高经济效益.     

Nash-optimization enhanced distributed model predictive control applied to the Shell benchmark problem

This paper presents an efficient distributed model predictive control scheme based on Nash optimality, in which the on-line optimization of the whole system is decomposed into that of several small co-operative agents in distributed structures, thus it can significantly reduce computational complexity in model predictive control of large-scale systems. The relevant nominal stability and the performance on single-step horizon under the communication failure are investigated. The Shell heavy oil fractionator benchmark control problem is illustrated to verify the effectiveness of the proposed control algorithm.