Yao Shi
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
Yao Shi has a total of 20 co-authors. Here are the top co-authors by number of shared publications:
Hongye Su
5 shared publicationsLei Xie
3 shared publicationsXialai Wu
3 shared publications- Zhe Wu
Department of Chemical and Biomolecular Engineering National University of Singapore Singapore Singapore
3 shared publications Entao Sun
2 shared publications- Lei Xie
State Key Laboratory of Industrial Control Technology, Zhejiang University, Hangzhou, China
2 shared publications
Latest Publications
- 21 August 2026
- Published by Wiley in AIChE Journal
- Vol. 72 (12) , e70602
Data and physics requirements for reliable physics‐informed neural networks in process modeling
Physics‐informed neural networks (PINNs) have gained increasing attention in chemical process modeling since they can embed first‐principles knowledge into neural network training. However, in practice, the embedded physics is often inaccurate, and experimental data are costly to obtain, raising fundamental questions about the required physics‐model accuracy and data volume required to achieve a target prediction accuracy. This work develops a theoretical framework for analyzing the generalization error of PINNs under model misspecification. We establish both an architecture‐independent error bound and an explicit bound for a specific PINN architecture. The bound is further related to the solution error with respect to the true system, yielding quantitative conditions on admissible model discrepancy, data requirements, and loss‐weight selection. Based on these conditions, two adaptive algorithms are proposed to guide physics‐model refinement and data collection. The theoretical findings are demonstrated using a chemical process network.
- 21 August 2026
- Published by Wiley in AIChE Journal
- Vol. 72 (12) , e70602
Data and physics requirements for reliable physics‐informed neural networks in process modeling
Physics‐informed neural networks (PINNs) have gained increasing attention in chemical process modeling since they can embed first‐principles knowledge into neural network training. However, in practice, the embedded physics is often inaccurate, and experimental data are costly to obtain, raising fundamental questions about the required physics‐model accuracy and data volume required to achieve a target prediction accuracy. This work develops a theoretical framework for analyzing the generalization error of PINNs under model misspecification. We establish both an architecture‐independent error bound and an explicit bound for a specific PINN architecture. The bound is further related to the solution error with respect to the true system, yielding quantitative conditions on admissible model discrepancy, data requirements, and loss‐weight selection. Based on these conditions, two adaptive algorithms are proposed to guide physics‐model refinement and data collection. The theoretical findings are demonstrated using a chemical process network.
- 20 July 2026
- Published by Elsevier in Chemical Engineering Science
- Vol. 338
A computationally efficient data-driven control scheme for chemical processes with operational constraints
- 20 July 2026
- Published by Elsevier in Chemical Engineering Science
- Vol. 338
A computationally efficient data-driven control scheme for chemical processes with operational constraints
- Open Access
- 7 February 2026
- Published by Elsevier in SSRN Electronic Journal
Lipschitz bounded Deep Koopman for robust modeling and offset-free predictive control of disturbed ORC system
The organic rankine cycle (ORC) has gained wide popularity in the field of low-temperature waste heat recovery. Despite this potential, effective modeling and control of ORC systems remains challenging because of its strong nonlinear characteristics, coupled with the variability of waste heat sources in practical applications. To address this issue, we propose the Lipschitz-bounded deep Koopman modeling framework, facilitating the construction of models that are fully compatible with linear model predictive control (MPC). The proposed framework enhances data-driven modeling by leveraging historical control input and system output data to identify a Koopman-based linear approximation of the underlying nonlinear ORC dynamics. By embedding a Lipschitz-constrained recurrent neural network within the Koopman framework, the method captures nonlinear dynamics in a linear representation while guaranteeing bounded outputs under measurement noise and variability of waste heat sources. Furthermore, a state observer is incorporated to eliminate steady-state offsets and to mitigate the impact of variability. Numerical simulations confirm that the proposed approach effectively captures the nonlinear dynamics and achieves robust offset-free performance under realistic operating conditions with variability and measurement noise.
- 7 February 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessLipschitz bounded Deep Koopman for robust modeling and offset-free predictive control of disturbed ORC system
The organic rankine cycle (ORC) has gained wide popularity in the field of low-temperature waste heat recovery. Despite this potential, effective modeling and control of ORC systems remains challenging because of its strong nonlinear characteristics, coupled with the variability of waste heat sources in practical applications. To address this issue, we propose the Lipschitz-bounded deep Koopman modeling framework, facilitating the construction of models that are fully compatible with linear model predictive control (MPC). The proposed framework enhances data-driven modeling by leveraging historical control input and system output data to identify a Koopman-based linear approximation of the underlying nonlinear ORC dynamics. By embedding a Lipschitz-constrained recurrent neural network within the Koopman framework, the method captures nonlinear dynamics in a linear representation while guaranteeing bounded outputs under measurement noise and variability of waste heat sources. Furthermore, a state observer is incorporated to eliminate steady-state offsets and to mitigate the impact of variability. Numerical simulations confirm that the proposed approach effectively captures the nonlinear dynamics and achieves robust offset-free performance under realistic operating conditions with variability and measurement noise.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Control of Network Systems
- Vol. 13 (3) , 1-12
Plug-and-Play Machine Learning-Based Model Predictive Control
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Control of Network Systems
- Vol. 13 (3) , 1-12
Plug-and-Play Machine Learning-Based Model Predictive Control
- 13 September 2023
- Published by Elsevier in Control Engineering Practice
- Vol. 141
Data-driven identification and fast model predictive control of the ORC waste heat recovery system by using Koopman operator
- 13 September 2023
- Published by Elsevier in Control Engineering Practice
- Vol. 141
Data-driven identification and fast model predictive control of the ORC waste heat recovery system by using Koopman operator