Performance Of Model Predictive Control Research Articles

An Impact of Switching Frequency and Model Accuracy on Model Predictive Current Control Performance for Reluctance Synchronous Motors

Abstract The present paper investigates the feasibility of utilizing the simplified prediction model for finite control set model predictive current control (FCS-MPCC) applied to reluctance synchronous motors (RSMs). The FCS-MPCC exhibits torque and current ripples, and a crucial consideration is the reduction of these ripples by increasing the switching frequency. The algorithm’s computational complexity is tied to the accuracy of the adopted model. Two approaches with varying levels of accuracy in predicting current dependencies concerning changes in the inductance of the RSM are investigated. The findings highlight the potential of employing simplified fixed inductance values for efficient control in drive systems, particularly those amenable to high switching frequencies.

Linking dataset quality and MPC in buildings: impact of temporal resolution

This study aims at assessing the impact of dataset quality on the performance of Model Predictive Control (MPC). The dataset feature, which is the target of this study, is temporal resolution, which applies to both data logging and the controller time step. A high temporal resolution might result in a more accurate predictive model, but it increases the need for data storage as well as the computational load on the model training. From the controller side, increasing the temporal resolution might lead to better control performance, but it sabotages the real-time response of the system. First, predictive models are developed based on datasets with different temporal resolutions. Subsequently, these predictive models are implemented within an MPC. Results reveal that decreasing the time step lower than 1 hour does not significantly improve the performance of the MPC. However, increasing the time step of MPC above 1 hour deteriorates its performance. Real-time response of the controller is a crucial criterion which deteriorates as the time step shortens. Hence, a suitable choice of temporal resolution is essential for developing a predictive model and MPC. In our case, a resolution of 1 hour is enough to guarantee a good performance of the controller.

Maximizing performance of linear model predictive control of glycemia for T1DM subjects

Maximizing performance of linear model predictive control of glycemia for T1DM subjects

Reinforcement Learning versus Model Predictive Control on greenhouse climate control

The greenhouse system plays a crucial role to ensure an adequate supply of fresh food for the growing global population. However, maintaining an optimal growing climate within a greenhouse requires resources and operational costs. To achieve economical and sustainable crop growth, efficient climate control in greenhouse production is paramount. Model Predictive Control (MPC) and Reinforcement Learning (RL) are the two approaches representing model-based and learning-based control, respectively. Each one has its own way to formulate control problems, define control objectives, and seek for optimal control actions that provide sustainable crop growth. Although certain forms of MPC and RL have been applied to greenhouse climate control, limited research has comprehensively analyzed the connections, differences, advantages, and disadvantages between these two approaches, both mathematically and in terms of performance. Therefore, this paper aims to address this gap by: (1) introducing a novel RL approach that utilizes Deep Deterministic Policy Gradient (DDPG) for large and continuous state–action space environments; (2) formulating the MPC and RL approaches for greenhouse climate control within a unified framework; (3) exploring the mathematical connections and differences between MPC and RL; (4) conducting a simulation study to analyze and compare the performance of MPC and RL; (5) presenting and interpreting the comparative results to provide valuable insights for the application of these control approaches in different scenarios. By undertaking these objectives, this paper seeks to contribute to the understanding and advancement of both MPC and RL methods in greenhouse climate control, fostering more informed decision-making regarding their selection and implementation based on specific requirements and constraints.

Experimental Validation for Distributed Control of Energy Hubs

As future energy systems become more decentralised due to the integration of renewable energy resources and storage technologies, several autonomous energy management and peer-to-peer trading mechanisms have been recently proposed for the operation of energy hub networks based on optimization and game theory. However, most of these strategies have been tested either only in simulated environments or small prosumer units as opposed to larger energy hubs. This simulation reality gap has hindered large-scale implementation and practical application of these method. In this paper, we aim to experimentally validate the performance of a novel multi-horizon distributed model predictive controller for an energy hub network by implementing the controller on a complete network of hubs comprising of a real energy hub inter-faced with multiple virtual hubs. The experiments are done using two different network topologies and the controller shows promising results in both setups.

Model Predictive Control for Automatic Transmission Upshift Inertia Phase

This article deals with model predictive control (MPC) design for automatic transmission (AT) upshift inertia phase, which aims to optimally coordinate the actions of oncoming (ONC) and off-going (OFG) clutches and engine and to facilitate calibration. The designed MPC strategy accounts for clutch actuation dynamics and constraints, while setting the tradeoff between three key and conflicting shift quality criteria: comfort; duration; and efficiency. The shift comfort and duration are ensured by minimizing output shaft torque and ONC clutch slip speed tracking errors, and the shift efficiency is reflected in clutch energy loss minimization on a prediction horizon. This allows for the calibration of the MPC performance through setting the inertia phase duration, the output shaft torque reference, cost function weighting coefficients, and constraints, rather than optimizing the shift control profiles themselves. The MPC problem is formulated as a constrained quadratic programming problem and efficiently solved online by an interior-point solver. The proposed MPC strategy is applicable to other transmissions with multiple actuators, such as parallel hybrid transmissions. The MPC system is examined through nonlinear powertrain model simulations for one to three shift and its performance is compared with an offline, multiobjective optimization-based control strategy. The MPC design flexibility and ease of calibration are demonstrated for different shift comfort and duration targets, as well as cost function tuning, and robustness with respect to clutch actuation parameter uncertainties is examined.

Fast Trajectory Tracking Control Algorithm for Autonomous Vehicles Based on the Alternating Direction Multiplier Method (ADMM) to the Receding Optimization of Model Predictive Control (MPC).

In order to improve the real-time performance of the trajectory tracking of autonomous vehicles, this paper applies the alternating direction multiplier method (ADMM) to the receding optimization of model predictive control (MPC), which improves the computational speed of the algorithm. Based on the vehicle dynamics model, the output equation of the autonomous vehicle trajectory tracking control system is constructed, and the auxiliary variable and the dual variable are introduced. The quadratic programming problem transformed from the MPC and the vehicle dynamics constraints are rewritten into the solution of the ADMM form, and a decreasing penalty factor is used during the solution process. The simulation verification is carried out through the joint simulation platform of Simulink and Carsim. The results show that, compared with the active set method (ASM) and the interior point method (IPM), the algorithm proposed in this paper can not only improve the accuracy of trajectory tracking, but also exhibits good real-time performance in different prediction time domains and control time domains. When the prediction time domain increases, the calculation time shows no significant difference. This verifies the effectiveness of the ADMM in improving the real-time performance of MPC.

A practically implementable reinforcement learning‐based process controller design

AbstractThe present article enables reinforcement learning (RL)‐based controllers for process control applications. Existing instances of RL‐based solutions have significant challenges for online implementation since the training process of an RL agent (controller) presently requires practically impossible number of online interactions between the agent and the environment (process). To address this challenge, we propose an implementable model‐free RL method developed by leveraging industrially implemented model predictive control (MPC) calculations (often designed using a simple linear model identified via step tests). In the first step, MPC calculations are used to pretrain an RL agent that can mimic the MPC performance. Specifically, the MPC calculations are used to pretrain the actor, and the objective function is used to pretrain the critic(s). The pretrained RL agent is then employed within a model‐free RL framework to control the process in a way that initially imitates MPC behavior (thus not compromising process performance and safety), but also continuously learns and improve its performance over the nominal linear MPC. The effectiveness of the proposed approach is illustrated through simulations on a chemical reactor example.

RBF-ARX model-based MPC approach to inverted pendulum: An event-triggered mechanism

The event-triggered predictive control based on RBF-ARX model (state-independent autoregressive with exogenous inputs constructed by radial basis function network) is designed in this paper, which is aiming at the issue of accumulating computational burden associated with solving the optimization problems at each sampling time. To alleviate the computational burden while the performance of model predictive control (MPC) cannot be simultaneously sacrificed, from the perspective of model building, the data-driven RBF-ARX model presented by a new state space form is combined into the event-triggered model predictive control (ETMPC) method instead of the mechanistic model. Therefore, we attempt to expand the results of ETMPC to the complex industrial processes rather than the restricted model described by mechanism in terms of proposing a new event-triggering condition (ETC) with a data-driven model. Additionally, the stability analysis of RBF-ARX model-based ETMPC is demonstrated with a focus on the boundedness of the model’s output. The simulation results presented in this paper serve to illustrate the effectiveness of the proposed method by acting on an inverted pendulum system in this paper, which can be showed that the computational burden is significantly reduced, and the predictive control performance of ETMPC is as well as the MPC roughly.

Robust virtual‐vector model predictive control of permanent‐magnet motor considering D‐Q axis inductance parameter uncertainty

AbstractTo suppress parameter mismatch and improve the output performance of model predictive control (MPC), a new robust virtual‐vector MPC strategy is proposed for a five‐phase permanent‐magnet motor in this study. Firstly, the incremental predictive model is applied to remove the impact of flux mismatch. Then the d‐q axis inductance parameter sensitivity of MPC is analysed, which produces the predictive current error between the normal parameter and mismatched parameter. Based on the current error, a new cost function is designed to select the voltage vector more accurately when the d‐q axis inductance parameter mismatch occurs. Thus, good robustness to the parameter's variation can be guaranteed. Afterwards, the duty cycle modulation technology is applied to allocate the duration time of two adjacent vector‐vectors and zero vector. So the motor current harmonics and torque ripple can be considerably suppressed. Finally, the experimental results are provided to show the effectiveness of this proposed method.

A robustly stabilizing fault‐tolerantMPC: A performance improvement‐based approach

AbstractIn terms of model predictive control (MPC) performance degradation caused by operational faults, in this article, a robust MPC strategy with active fault tolerance properties is proposed. The proposed strategy incorporates a fault supervision layer into the structure of conventional cost‐contracting formulation‐based robust MPC for the online update of the nominal controller model in the event of faults. The robust MPC is based on multiplant uncertainty, while the supervisory layer consists of a bank of unknown input observers and a decision‐making algorithm. Simulation results in a nonlinear polymerization reactor subject to process faults demonstrate that the proposed approach offers superior performance compared to the conventional strategy.

Modular Modeling and Statistical Validation for Grid-Connected FS-MPC-Controlled Matrix Converters

In recent publications statistical model checking (SMC) has been proposed as a method for verifying the performance of finite-set model predictive control (FS-MPC) algorithms applied to power electronics converters. One of the reasons the full potential of the method in the power electronics systems (PES) has not yet been explored is the time consuming modelling process. In this paper we propose a modular method of modelling the power electronics system components by providing simple building blocks, which can be connected to build different PES. The modelling method is here demonstrated on a direct matrix converter, which operates in a stochastic grid with different harmonic distortion levels and voltage sags. By applying the SMC, the performance of the control algorithm in terms of the output current distortion, effects of the weighting factor selection and grid distortions on the device utilization can be evaluated. The obtained results confirm, that high grid distortions and voltage sags will increase the stress of several devices. This information can be of great importance to identify the most stressed components and how the control algorithm can be adapted to extend the lifetime of the components and thereby the system during different grid conditions. The verified FS-MPC algorithm has also been implemented in an experimental set-up.

A Simplified Multivector-Based Model Predictive Current Control for PMSM With Enhanced Performance

To address the challenges of poor steady-state performance and large amount calculation of conventional model predictive current control (MPCC), a simplified multi-vector-based MPCC with enhanced performance is proposed in this paper. Firstly, an effective voltage vectors (VVs) selection method based on current-error is proposed, which can directly determine the optimal VVs without cost function enumeration. Compared with the standard method, the number of vectors that need to be evaluated is reduced from 7 to 1. Meanwhile, the duty cycle of each VV is calculated based on the current error to realize error-free control. In addition, to alleviate the deleterious effect of dead-time on the control performance, an improved MPCC scheme with optimal utilization of dead-time is proposed. The key is that the dead-time existing in MPCC is regarded as a dead-time voltage vector (DVV), which can be rationally utilized to improve the control performance. Both theoretical analysis and experimental results are given to verify the effectiveness of the proposed MPCC schemes.

Model-Predictive Control of Multilevel Inverters: Challenges, Recent Advances, and Trends

Model predictive control (MPC) has emerged as a promising control method in power electronics, particularly for multi-objective control problems such as multilevel inverter (MLI) applications. Over the past two decades, improving the performance of MPC and tackling its technical challenges such as computational load, modeling accuracy, cost function design and weighting factors selection have attracted great interest in power electronics. This article aims to discuss the current state of MPC strategies for MLI applications, describing the significance of each challenge with the reported effective solutions. Through this review, the MPC methods are categorized into two groups, direct MPC (without modulator) and indirect MPC (with modulator). The recent advances of each category are presented and analyzed, focusing on direct MPC as the most applied method for MLI topologies. In addition, some of the important concepts are experimentally validated through a case study and compared under the same operating conditions to evaluate the performance and highlight their features. Finally, the future trends of MPC for MLI applications are discussed based on the current state and reported developments.

Advanced process automation of a pharmaceutical continuous wet granulation line: Perspectives on the application of a model predictive control from solid feeders to dryer

Pharmaceutical continuous manufacturing provides the appropriate tools (e.g. the understanding of process dynamics and appropriate and adaptable control strategy) in order to deal with Quality-by-Design expectations and even to the future smart manufacturing described by Quality-by-Control. Those tools form part of the given framework of the regulatory agencies led by an effective quality risk management. Soft sensors and control algorithms such as model predictive control are stepping stones for more agile processes and increased robustness by keeping the quality attributes of the final drug product in their acceptable ranges and by mitigating undesired events. The implementation of a model predictive control (MPC) system on a pharmaceutical continuous manufacturing plant for the wet granulation process is described. The control objectives and strategy are presented as well as the selected variables, the process dynamics identification, the MPC performance and its specific tuning where a commercial software has been used, setting the framework of this study. MPCs have been applied successfully on two pharmaceutical drug products (Diclofenac and Paracetamol): an accurate control of the API content and of the LOD was achieved in order to produce a constant quality of tablets on both drug products. In addition, some of the process parameters have been identified as mandatory to be step tested for each change of drug product, leading to a simplified MPC implementation.

Simultaneous multistep transformer architecture for model predictive control

Transformer neural networks have revolutionized natural language processing by effectively addressing the vanishing gradient problem. This study focuses on applying Transformer models to time-series forecasting and customizing them for a simultaneous multistep-ahead prediction model in surrogate model predictive control (MPC). The proposed method showcases improved control performance and computational efficiency compared to LSTM-based MPC and one-step-ahead prediction models using both LSTM and Transformer networks. The study introduces three key contributions: (1) a new MPC system based on a Transformer time-series architecture, (2) a training method enabling multistep-ahead prediction for time-series machine learning models, and (3) validation of the enhanced time performance of multistep-ahead Transformer MPC compared to one-step-ahead LSTM networks. Case studies demonstrate a significant fifteen-fold improvement in computational speed compared to one-step-ahead LSTM, although this improvement may vary depending on MPC factors like the lookback window and prediction horizon.

Hybrid Model Predictive Control of a Two-Body Wave Energy Converter with Mechanically Driven Power Take-Off

In this paper, a variable damper is proposed to regulate the efficiency of a two-body wave energy converter (WEC) with mechanically driven power take-off (PTO). The variable damper introduces logic constraints into the WEC system, which can be translated into a mixed logical dynamical form with the dynamics of real-valued variables, the dynamics of logic variables, and their interactions. A hybrid model predictive control (MPC) method is used to determine the control inputs, which has the capacity to handle various constraints. The performance is assessed through simulations to evaluate the effectiveness of the proposed method. The achievable performance improvements of the proposed hybrid MPC are shown by means of comparative analysis with uncontrolled WEC devices. The results show that the proposed hybrid MPC has a high requirement on the lower bound of the variable damper and the maximum damping is used only at low relative velocities to achieve the optimum phase, like latching control. The hybrid MPC performs exceptionally well under wave conditions with a small significant wave height and long wave period, improving the power generation of the uncontrolled system up to 22.5%. And, the prediction error has a significant effect on hybrid MPC performance, especially for long prediction horizon.

Online model-based predictive control with smart thermostats: application to an experimental house in Québec

This paper tests the impact of model resolution and structure on the performance of Model Predictive Control (MPC) implementation in an unoccupied research house in Québec equipped with smart thermostats. Two low-order models and a high-order multi-zone model were calibrated with measured data, with the structure of the multi-zone model being generated automatically during the calibration procedure. The three models were used to apply real-time MPC to an experimental house in Québec using the established dynamic tariffs for morning and evening peaks. MPC with any of the three models successfully preheated the house before the demand-response events, outperforming the reference reactive controller, reducing cost and thermal discomfort. The high-order multi-zone model performed the best, reducing average cost of electricity by 55% and high-price energy consumption by 71%, compared to the low-order models, which achieved cost reductions of 40% and 44% and energy consumption reductions of 48% and 54% respectively.

Model predictive control of nonlinear processes using neural ordinary differential equation models

Neural Ordinary Differential Equation (NODE) is a recently proposed family of deep learning models that can perform a continuous approximation of a linear/nonlinear dynamic system using time-series data by integrating the neural network model with classical ordinary differential equation solvers. Modeling nonlinear dynamic processes using data has historically been a critical challenge in the field of chemical engineering research. With the development of computer science and neural network technology, recurrent neural networks (RNN) have become a popular black-box approach to accomplish this task and have been utilized to design model predictive control (MPC) systems. However, as a discrete-time approximation model, RNN requires strictly uniform step sequence data to operate, which makes it less robust to an irregular sampling scenario, such as missing data points during operation due to sensor failure or other types of random errors. This paper aims to develop a NODE model and construct an MPC based on this novel continuous-time neural network model. In this context, closed-loop stability is established and robustness to noise is addressed using a variety of data analysis techniques. An example of a chemical process is utilized to evaluate the performance of NODE-based MPC. Furthermore, the performance of the NODE-based MPC under Gaussian and non-Gaussian noise is investigated, and the subsampling method is found to be effective in building models for MPC that are suitable for handling the presence of non-Gaussian noise in the data.

Generalized Data-Driven Model-Free Predictive Control for Electrical Drive Systems

The performance of model predictive control has a strong correlation to the precision of the physical parameters of the plant, and these parameters are hard to determine since they are continuously changing during the operation process. To fully eliminate the influence of the physical parameters and enhance robustness, a model-free predictive control is proposed in this article to suit the electrical drive systems. The plant model is designed as several discrete-time transfer functions used to decouple the input and output signals and to describe their relationships, and the coefficients of these functions are online designed based on the recursive least square algorithm. An observer is designed to obtain accurately sampled current components considering the delays. The proposed method is applied to a permanent magnet synchronous motor speed control system as the stator current controller, and the simulation and experimental results show the advantages of the improved dynamics, stator current quality, and robustness compared with the conventional model-free predictive current control strategy.