Pei Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Pei Sun has a total of 30 co-authors. Here are the top co-authors by number of shared publications:
Hongye Su
Institute of Cyber‐Systems and Control, National Key Laboratory of Industrial Control Technology Zhejiang University Zhejiang China
4 shared publications- Haoyi Que
School of Artificial Intelligence, Shenzhen Polytechnic University, 7098 Liuxian Boulevard, Nanshan District, Shenzhen 518055, China
3 shared publications - Hongye Su
Zhejiang University
3 shared publications - Junghui Chen
Department of Chemical Engineering, Chung-Yuan Christian University, Taoyuan, Taiwan
3 shared publications Zhiming Zhang
State Key Laboratory of Industrial Control Technology, Zhejiang University, Hangzhou, China
3 shared publicationsLei Xie
Institute of Cyber‐Systems and Control, National Key Laboratory of Industrial Control Technology Zhejiang University Zhejiang China
2 shared publications
Latest Publications
- 24 July 2026
- Published by Elsevier in ISA Transactions
- Vol. 177, 512-526
Smart PID generator for nonlinear systems: zero-shot reinforcement learning based on virtual environments
- 24 July 2026
- Published by Elsevier in ISA Transactions
- Vol. 177, 512-526
Smart PID generator for nonlinear systems: zero-shot reinforcement learning based on virtual environments
- 1 May 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 20 (1) , 899-910
A Priori Knowledge-Based Dual Hierarchical RNN for Spatial–Temporal Process Modeling: Using a Multitubular Reactor as a Case Study
Deep learning methods have been rapidly developed in recent decades. In this article, they are extended to model spatial-temporal industrial processes. Instead of pure black-box data-driven modeling approaches, the proposed model encodes the domain knowledge and physical rules governing the spatiotemporal system, called a dual-hierarchical recurrent neural network (DH-RNN). Both spatial and temporal relationships are modeled by multiple RNNs with diverse structures, which need correct specifications of all the interactions between spatial and temporal variables with a priori knowledge of the real process. A more accurate prediction can be obtained with fewer parameters employed in the network. And the effectiveness of the proposed DH-RNN is verified via a real ethylene oxychlorination process.
- 1 May 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 20 (1) , 899-910
A Priori Knowledge-Based Dual Hierarchical RNN for Spatial–Temporal Process Modeling: Using a Multitubular Reactor as a Case Study
Deep learning methods have been rapidly developed in recent decades. In this article, they are extended to model spatial-temporal industrial processes. Instead of pure black-box data-driven modeling approaches, the proposed model encodes the domain knowledge and physical rules governing the spatiotemporal system, called a dual-hierarchical recurrent neural network (DH-RNN). Both spatial and temporal relationships are modeled by multiple RNNs with diverse structures, which need correct specifications of all the interactions between spatial and temporal variables with a priori knowledge of the real process. A more accurate prediction can be obtained with fewer parameters employed in the network. And the effectiveness of the proposed DH-RNN is verified via a real ethylene oxychlorination process.
- Open Access
Thermal-Imaging-Based PCA Method for Monitoring Process Temperature
To overcome the shortage of traditional temperature sensors, this paper adopts infrared thermal imaging technology for temperature measurement. To avoid the spatial information loss issue during the image data vectorization process, this paper adopted the spatial relationship between pixels in principal component analysis (PCA) model training, which is called spatial information-based PCA (SIPCA). Then, spatial information is also used in the fault localization method to enhance the fault location performance. Tested by an experimental tank system, the proposed method achieves better performance than the traditional PCA approach, and it can detect heat leakage faults on the surface of the equipment.
Open AccessThermal-Imaging-Based PCA Method for Monitoring Process Temperature
To overcome the shortage of traditional temperature sensors, this paper adopts infrared thermal imaging technology for temperature measurement. To avoid the spatial information loss issue during the image data vectorization process, this paper adopted the spatial relationship between pixels in principal component analysis (PCA) model training, which is called spatial information-based PCA (SIPCA). Then, spatial information is also used in the fault localization method to enhance the fault location performance. Tested by an experimental tank system, the proposed method achieves better performance than the traditional PCA approach, and it can detect heat leakage faults on the surface of the equipment.
- 5 February 2023
- Published by Wiley in The Canadian Journal of Chemical Engineering
- Vol. 101 (8) , 4587-4605
Active design of dynamic GP models for model predictive control using expected improvement
Modelling is a basic and key requirement for model‐based controlling, monitoring, or other process strategies. In non‐linear model predictive control (NMPC), although data‐driven models can be more easily established than first‐principle ones, representative data may not be adequately included in advance to train a complete model, which is an attractive research topic. An actively improved Gaussian process (GP) model building strategy is developed, especially for incomplete models based on the idea of Bayesian optimization. The GP model can be used online as the internal model of model predictive control (MPC) directly. The model‐building objective is based on the expected improvement strategy, which can exploit information gained from the currently gathered data as well as explore uncharted regions. The proposed method is a real‐time design of experiments based on variance information of GP for efficient model building with insufficient initial training data for NMPC. Multi‐step ahead prediction model is considered to give full play to predicting features of NPMC. Besides, a novel disturbance rejection strategy is also proposed based on GP outputs. Two simulation results, including comparisons with some traditional algorithms, are presented to demonstrate the effectiveness of the proposed method.
- 5 February 2023
- Published by Wiley in The Canadian Journal of Chemical Engineering
- Vol. 101 (8) , 4587-4605
Active design of dynamic GP models for model predictive control using expected improvement
Modelling is a basic and key requirement for model‐based controlling, monitoring, or other process strategies. In non‐linear model predictive control (NMPC), although data‐driven models can be more easily established than first‐principle ones, representative data may not be adequately included in advance to train a complete model, which is an attractive research topic. An actively improved Gaussian process (GP) model building strategy is developed, especially for incomplete models based on the idea of Bayesian optimization. The GP model can be used online as the internal model of model predictive control (MPC) directly. The model‐building objective is based on the expected improvement strategy, which can exploit information gained from the currently gathered data as well as explore uncharted regions. The proposed method is a real‐time design of experiments based on variance information of GP for efficient model building with insufficient initial training data for NMPC. Multi‐step ahead prediction model is considered to give full play to predicting features of NPMC. Besides, a novel disturbance rejection strategy is also proposed based on GP outputs. Two simulation results, including comparisons with some traditional algorithms, are presented to demonstrate the effectiveness of the proposed method.
- 24 May 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cybernetics
- Vol. 53 (12) , 7560-7571
SVD-Based Robust Distributed MPC for Tracking Systems Coupled in Dynamics With Global Constraints
This article presents a novel singular value decomposition (SVD)-based robust distributed model predictive control (SVD-RDMPC) strategy for linear systems with additive uncertainties. The system is globally constrained and consists of multiple interrelated subsystems with bounded disturbances, each of whom has local constraints on states and inputs. First, we integrate the steady-state target optimizer into the MPC problem through the offset cost function to formulate a modified single optimization problem for tracking changing targets from real-time optimization. Then, the concept of constraint tightening is utilized to enhance the robustness and ensure robust constraint satisfaction in the presence of interferences. On this basis, the SVD method is introduced to decompose the new optimization problem into several independent subsystems on the orthogonal projection space, and a distributed dual gradient algorithm with convergence proved is implemented to obtain the control of each nominal subsystem. The recursive feasibility is then ensured and the tracking ability of the strategy is analyzed. It is verified that for a target, the system can be steered to a neighborhood of the closest possible steady setpoint. At last, the effectiveness of the raised SVD-RDMPC strategy is established in two simulations on building temperature control and load frequency control.
- 24 May 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cybernetics
- Vol. 53 (12) , 7560-7571
SVD-Based Robust Distributed MPC for Tracking Systems Coupled in Dynamics With Global Constraints
This article presents a novel singular value decomposition (SVD)-based robust distributed model predictive control (SVD-RDMPC) strategy for linear systems with additive uncertainties. The system is globally constrained and consists of multiple interrelated subsystems with bounded disturbances, each of whom has local constraints on states and inputs. First, we integrate the steady-state target optimizer into the MPC problem through the offset cost function to formulate a modified single optimization problem for tracking changing targets from real-time optimization. Then, the concept of constraint tightening is utilized to enhance the robustness and ensure robust constraint satisfaction in the presence of interferences. On this basis, the SVD method is introduced to decompose the new optimization problem into several independent subsystems on the orthogonal projection space, and a distributed dual gradient algorithm with convergence proved is implemented to obtain the control of each nominal subsystem. The recursive feasibility is then ensured and the tracking ability of the strategy is analyzed. It is verified that for a target, the system can be steered to a neighborhood of the closest possible steady setpoint. At last, the effectiveness of the raised SVD-RDMPC strategy is established in two simulations on building temperature control and load frequency control.
- 28 April 2022
- Published by Elsevier in Control Engineering Practice
- Vol. 124
An analytical partial least squares method for process monitoring
- 28 April 2022
- Published by Elsevier in Control Engineering Practice
- Vol. 124
An analytical partial least squares method for process monitoring
- 14 May 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1396-1399
A 2-Dimensional Linear Non-minimum Phase Estimator Based on Dynamic Exceeded Regressor
In the paper, we proposed a 2-dimensional non-minimum phase linear regressor, to verify that in the procedure of dynamic extended regressor and mixing, with bounded functions of time, which persistence of excitation(PE) condition is absent, the parameter estimation errors could converge to 0. And a proposition of available form is given for the systems to meet the requirements of the dynamical extended regressor. Numeral and simple pendulum examples examine the results and simulations show the effectiveness.
- 14 May 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1396-1399
A 2-Dimensional Linear Non-minimum Phase Estimator Based on Dynamic Exceeded Regressor
In the paper, we proposed a 2-dimensional non-minimum phase linear regressor, to verify that in the procedure of dynamic extended regressor and mixing, with bounded functions of time, which persistence of excitation(PE) condition is absent, the parameter estimation errors could converge to 0. And a proposition of available form is given for the systems to meet the requirements of the dynamical extended regressor. Numeral and simple pendulum examples examine the results and simulations show the effectiveness.
- 1 December 2020
- Published by Elsevier in Journal of the Franklin Institute
- Vol. 357 (18) , 13766-13784
An event-based interaction method for consensus of multiple complex networks
- 1 December 2020
- Published by Elsevier in Journal of the Franklin Institute
- Vol. 357 (18) , 13766-13784
An event-based interaction method for consensus of multiple complex networks