Multi-model Predictive Control Research Articles

Launch ascent guidance by discrete multi-model predictive control

This paper studies the application of discrete multi-model predictive control as a trajectory tracking guidance law for a space launcher. Two different algorithms are developed, each one based on a different representation of launcher translation dynamics. These representations are based on an interpolation of the linear approximation of nonlinear pseudo-five degrees of freedom equations of translation around an elliptical Earth. The interpolation gives a linear-time-varying representation and a linear-fractional representation. They are used as the predictive model of multi-model predictive controllers. The controlled variables are the orbital parameters, and constraints on a terminal region for the minimal accepted precision are also included. Use of orbital parameters as the controlled variables allows for a partial definition of the trajectory. Constraints can also be included in multi-model predictive control to reduce the number of unknowns of the problem by defining input shaping constraints. The guidance algorithms are tested in nominal conditions and off-nominal conditions with uncertainties on the thrust. The results are compared to those of a similar formulation with a nonlinear model predictive controller and to a guidance method based on the resolution of a simplified version of the two-point boundary value problem. In nominal conditions, the model predictive controllers are more precise and produce a more optimal trajectory but are longer to compute than the two-point boundary solution. Moreover, in presence of uncertainties, developed algorithms exhibit poor robustness properties. The multi-model predictive control algorithms do not reach the desired orbit while the nonlinear model predictive control algorithm still converges but produces larger maneuvers than the other method.

Multi-Model Predictive Control with Kalman Filter on DC/DC Converter

In this paper, a new kind of Multi-model Predictive Controller (MMPC) is presented. Multiple models switch MMPC with Kalman filter based on neural network optimization was implemented in a DC/DC converter. Multiple models direct switch was used according to the two modes of the Boost-Buck circuit and different loads condition. Kalman filter was successfully added to the MMPC to transfer the optimal state value to MMPC from the output of the DC/DC circuit. Parameters optimization method is also discussed in the paper when the controller was applied to real model simulation. Simulation result demonstrates that the good performance of MMPC compared with single model predictive control and PID control. With the help of high speed FPGA, it is possible to apply this method to hardware simulation.

LMI-Based Multi-model Predictive Control of an Industrial C3/C4 Splitter

In this paper, the robust Model Predictive Control (MPC) of systems with model uncertainty is addressed. The robust approach usually involves the inclusion of nonlinear constraints to the optimization problem upon which the controller is based. At each time step the sequence of control actions is then calculated through the resolution of a NonLinear Programming problem, which can be too computer demanding for high dimension systems. Here, the conventional Multi-model Predictive Control (MMPC) problem is re-casted as an LMI-based problem that can be solved with a lower computational cost. The conventional and LMI-based robust controllers’ performances and computational costs are compared through simulations of the control of an industrial C3/C4 splitter.

Robust Nonlinear Method for Steam Generator Level Control

In this paper, a new, robust control method based on a multimodel predictive control scheme is developed for steam generator level (SGL) control in nuclear power plants. For a multiramp power increase from low to full power, the proposed controller is capable of keeping the SGL within the admissible range by minimizing the level transients and improving the stability of the control loop. Simulation results and a general framework for systematically studying the SGL are presented to demonstrate the effectiveness of the proposed control method by comparing the performance of the designed controller with that of a properly tuned conventional three-element proportional-integral-derivative (PID) controller. Moreover, it has been demonstrated that the proposed controller is more robust than a conventional PID controller to steam flow disturbances caused by load variations.

Multi-model Switching Predictive Functional Control of Boiler Main Steam Temperature

Multi-model Switching Predictive Functional Control of Boiler Main Steam Temperature

Blood glucose concentration regulation in Type 1 diabetics using multi model multi parametric model predictive control: An empirical approach

A Glucose- Insulin steady state static map is obtained from the Hovorka's 8th order virtual patient model. Three First Order Plus Time Delay (FOPTD) models are derived for the three piecewise linear regions in it. Through polyhedral vector space partitioning based on constraint violation, critical regions in state vector space are identified. A state feedback gain based controller is designed for each critical region. The controller design prevents constraint violations and ensures convexity while regulating the state vector to origin. The solution is also globally minimal. The state vector space of each empirical model is subjected to such analysis, resulting in three different set of critical regions and corresponding controllers. Gain Scheduling (GS) based on the Blood Glucose Concentration (BGC) measurement ensured proper profile selection. Through delay time compensation techniques, the multi model multi-parametric Model Predictive Control (mp-MPC) is designed for pure dynamics of each linear region. It is observed that the gain scheduled controller regulates the BGC within the acceptable range (80mg/dL to 160 mg/dL) during multiple meal disturbances. The explicit state feedback gain nature of the controller implies ease of deployment on memory constrained embedded devices.

Multilinear-Model Predictive Control of a Tubular Solid Oxide Fuel Cell System

As solid oxide fuel cells (SOFCs) are highly nonlinear systems, a single linear controller cannot perform satisfactorily over a wide range of operating conditions of the processes. This work studies multilinear-model predictive control of a tubular SOFC. The objective is to control the fuel cell outlet voltage over a wide range of operating conditions by manipulating inlet fuel pressure (flow rate). A first-principles model of an ammonia fed-tubular solid oxide fuel cell is used for the controller design. The model accounts for diffusion, inherent impedance, transport (momentum, heat and mass transfer), electrochemical reactions, activation and concentration polarizations, and the ammonia decomposition reaction. The servo and regulatory performances of the multimodel predictive controller (MMPC) are compared with those of a single-model predictive controller (SMPC) and a proportional-integral (PI) controller. For small load changes, the MMPC, SMPC, and PI controller all provide zero offset, and the MMPC yields the best closed-loop performance. However, for large load changes, the SMPC and PI controller fail to provide zero offset; under these two controllers the closed-loop system with the large load changes is unstable.

Multiple Model Predictive Flood Control in Regulated River Systems with Uncertain Inflows

This paper presents a novel approach to real time automatic flood control in a managed river network that is subject to uncertain inflows. The proposed approach uses multiple models to represent inflows ranging from low to high flow. Optimal model selection is achieved in a minimum mean square error sense using a bank of Kalman filters to identify the most likely inflow characteristic. There are no a-priori probabilities assigned to the individual models. Model Predictive Control is used for water level controller design. Our Adaptive Multi Model Predictive Control (AMMPC) method is proposed as an alternative to existing techniques that also use multiple inflow models but with a-priori inflow model probabilities, either weighted or equally likely. The performance of the approach is demonstrated using a simulated river-reservoir model as well as using data collected at the Wivenhoe Dam during the 2011 floods in Queensland, Australia.

The Research of Predictive Control Algorithm for the Three-Tank Water Level System Based on the Multi-Model

Three-tank water level system has the characteristics of nonlinear, large time delay, time-varying parameters, and the system is broadly representative in the research of non-linear and inertial process control. For the controlled object’s strong-nonlinear characteristic when the liquid level given value changes, a multi-model predictive PID cascade control strategy was presented to improve response performance of the control system. On the basis of analyzing the dynamic characteristics of the object, multiple sub-model of the object was established, and a global approximate model of the complex object was obtained as the predictive model by the weighted sub-model. Owing to the rapid response characteristic of PID control, we built the predictive control PID cascade control system by combining PID algorithm with predictive control. The experiment results show that the control algorithm can achieve good tracking control, and greatly improve dynamic and static performance of the system.

A robust multi-model predictive controller for distributed parameter systems

In this work a robust nonlinear model predictive controller for nonlinear convection–diffusion-reaction systems is presented. The controller makes use of a collection of reduced order approximations of the plant (models) reconstructed on-line by projection methods on proper orthogonal decomposition (POD) basis functions. The model selection and model update step is based on a sufficient condition that determines the maximum allowable process-model mismatch to guarantee stable control performance despite process uncertainty and disturbances. Proofs on the existence of a sequence of feasible approximations and control stability are given. Since plant approximations are built on-line based on actual measurements, the proposed controller can be interpreted as a multi-model nonlinear predictive control (MMPC). The performance of the MMPC strategy is illustrated by simulation experiments on a problem that involves reactant concentration control of a tubular reactor with recycle.

Multi-Hierarchical Model Predictive Control Based on K-Means Clustering Algorithms

Multi-model predictive control has become an effective method for nonlinear system. But the traditional multi-model has large tracking error compared with desired output when it is used to solve the operating condition with large scale transition. To solve this problem, this paper presents a new structure of multi-model called multi-hierarchical model. The new structure consists of many layers that each layer is constituted of different number of multiple models. In each layer, multi-model is obtained by partial least squares method after k-means clustering algorithm divides the global working spaces into desired parts. Because of this special structure, the models chose from different layers can deal with the operating condition changed with large scale. At the end of this paper, experiments are carried on the pH neutralization process which is a MIMO nonlinear system and the simulation results demonstrate that the multi-hierarchical model is superior to single-hierarchical model with smaller model tracking error faster convergence speed and better stability.

Multi-model Predictive Control for Pressurized Water Reactor

Multi-model Predictive Control for Pressurized Water Reactor

Combination of multi-model predictive control and the wave theory for the control of simulated moving bed plants

In this work a new approach to the control of simulated moving bed (SMB) chromatographic separation processes is presented. This approach is based on the combination of the wave theory and Multi-Model Predictive Control (MMPC). The wave theory provides the theoretical framework in which the control law is formulated whereas receding-horizon MPC is used for determining the appropriate controller parameters. As SMB plants are distributed parameter systems (DPS) with nonlinearity arising from the expression of the adsorption isotherms, classical numerical methods for the solution of DPS are computationally demanding for MPC purposes. Reduced-order models are therefore derived using the proper orthogonal decomposition (POD) technique. To ensure stability, the POD model is updated on-line, resulting in MMPC.

Multi-model predictive control based on LMI: from the adaptation of the state-space model to the analytic description of the control law

This article proposes an explicit robust predictive control solution based on linear matrix inequalities (LMIs). The considered predictive control strategy uses different local descriptions of the system dynamics and uncertainties and thus allows the handling of less conservative input constraints. The computed control law guarantees constraint satisfaction and asymptotic stability. The technique is effective for a class of nonlinear systems embedded into polytopic models. A detailed discussion of the procedures which adapt the partition of the state space is presented. For the practical implementation the construction of suitable (explicit) descriptions of the control law are described upon concrete algorithms.

ON MODEL PREDICTIVE CONTROL OF QUASI-RESONANT CONVERTERS

In this paper, different model predictive control synthesis frameworks are examined for DC–DC quasi-resonant converters in order to achieve stability and desired performance. The performances of model predictive control strategies which make use of different forms of linearized models are compared. These linear models are ranging from a simple fixed model, linearized about a reference steady state to a weighted sum of different local models called multi model predictive control. A more complicated choice is represented by the extended dynamic matrix control in which the control input is determined based on the local linear model approximation of the system that is updated during each sampling interval, by making use of a nonlinear model. In this paper, by using and comparing these methods, a new control scheme for quasi-resonant converters is described. The proposed control strategy is applied to a typical half-wave zero-current switching QRC. Simulation results show an excellent transient response and a good tracking for a wide operating range and uncertainties in modeling.

Support vector machine-based multi-model predictive control

In this paper, a support vector machine-based multi-model predictive control is proposed, in which SVM classification combines well with SVM regression. At first, each working environment is modeled by SVM regression and the support vector machine network-based model predictive control (SVMN-MPC) algorithm corresponding to each environment is developed, and then a multi-class SVM model is established to recognize multiple operating conditions. As for control, the current environment is identified by the multi-class SVM model and then the corresponding SVMN-MPC controller is activated at each sampling instant. The proposed modeling, switching and controller design is demonstrated in simulation results.

Multi-model predictive control method for nuclear steam generator water level

The dynamics of a nuclear steam generator (SG) is very different according to the power levels and changes as time goes on. Therefore, it is an intractable as well as challenging task to improve the water level control system of the SG. In this paper, a robust model predictive control (RMPC) method is developed for the level control problem. Based on a multi-model framework, a combination of a local nominal model with a polytopic uncertain linear parameter varying (LPV) model is built to approximate the system’s non-linear behavior. The optimization problem solved here is based on a receding horizon scheme involving the linear matrix inequality (LMI) technique. Closed loop stability and constraints satisfaction in the entire operating range are guaranteed by the feasibility of the optimization problem. Finally, simulation results show the effectiveness and the good performance of the proposed method.

An investigation on the application of predictive control for controlling screw position and velocity on an injection molding machine

AbstractInjection velocity is one of the key parameters in the injection molding process that has significant effect on part quality, influencing common problems, such as flashing and short shots. Different products require specific molding conditions, including mold and melt temperatures, velocity profiles, and polymers, making the injection velocity dynamics vary significantly and difficult for high precision velocity control. Two predictive controllers were developed to control the screw injection velocity. The first approach uses a position sensor as feedback and a simplified predictive controller to track an injection velocity setpoint profile. The controller was developed and implemented utilizing single‐step change and multistep change open loop tests. The second strategy uses a multimodel dynamic matrix predictive controller to overcome the nonlinear characteristics of the injection velocity dynamics as the mold is being filled. Velocity feedback is provided by high speed processing of the position analog signal. This approach utilizes several open loop injection velocity profiles to generate corresponding dynamic matrices for the controller. As a result, this controller is modified or retuned automatically when setpoint changes in injection velocity occur, as well as when using different polymers and molds. The close loop results for both simulations and real time control have demonstrated that the two predictive controllers provided good setpoint tracking performance for wide ranging position and velocity profiles. POLYM. ENG. SCI., 47:390–399, 2007. © 2007 Society of Plastics Engineers.

Multi-Model Predictive Control for Nonlinear Systems Based on Mixed Logic

Big prediction errors are brought into being as the local linear model is used to predict the future output in the model prediction process for the existent multi-model predictive control algorithms. To solve this problem, this paper introduces causality relationship between multi-model of nonlinear process and output prediction into model predictive control framework in the term of constraint conditions, so that the nonlinear process can be described by a mixed-logic dynamic model. This paper also introduces switch rules into the multi-model predictive controller as a kind of pre-experiential knowledge. This new mixed logic dynamic model can characterize the nonlinear process entirely, thus solving the problem of model prediction and model switch for multi-model constrained nonlinear predictive control.

Model Predictive Control for Transparent Teleoperation Under Communication Time Delay

Prior efforts in bilateral teleoperation under communication delay have mainly yielded control algorithms that sacrifice performance in order to guarantee robust stability. In contrast, this paper proposes a multimodel predictive controller that can enhance the teleoperation transparency in the presence of a known constant delay. Separate controllers are designed for free motion/soft contact and contact with rigid environments, with switching between these mode-based control laws occurring according to the identified contact mode. Performance objectives such as position tracking and tool impedance shaping for free motion/soft contact, as well as position and force tracking for contact with rigid environments, are incorporated into a multi-input/multi-output state-space representation of the system dynamics. New Artstein-type state and measurement transformations are proposed to generate delay-free dynamics suitable for output-feedback control, based on the original dynamics with delays in various input and output channels. The application of the continuous-time linear quadratic Gaussian control synthesis to the resulting mode-based delay-free dynamics yields control laws that guarantee closed-loop stability and enhanced performance in each phase of teleoperation. The robustness of the mode-based controllers with respect to parametric uncertainty is analyzed. Experimental results with a single-axis teleoperation setup demonstrate the effectiveness of the proposed approach