The implementation of sophisticated control strategies for building energy systems is crucial for improving energy efficiency and occupant comfort.While nonlinear model predictive control offers promising benefits,its...The implementation of sophisticated control strategies for building energy systems is crucial for improving energy efficiency and occupant comfort.While nonlinear model predictive control offers promising benefits,its application to large-scale building systems remains challenging due to computational complexity and system coupling.This work presents a comprehensive study of Nonlinear Distributed Model Predictive Control(NDMPC)implementation for building energy systems,comparing Alternating Direction Method of Multipliers(ADMM)and Augmented Lagrangian Alternating Direction Inexact Newton(ALADIN)algorithms alongside different modeling approaches.We examine a multi-zone heating system with thermal storage and multiple producers,investigating both Ordinary Differential Equation(ODE)-based and Artificial Neural Network(ANN)based modeling strategies.Through systematic parameter tuning using Bayesian optimization and closedloop scaling analysis with up to 40 thermal zones,we demonstrate that ALADIN-based NDMPC can achieve performance comparable to centralized model predictive control,showing greater robustness to parameter variations than ADMM.Our results reveal that ANN-based models effectively mitigate distributed integration errors and significantly reduce computation time compared to ODE-based approaches.Detailed computational profiling identifies specific bottlenecks in different NDMPC components.These findings advance the practical implementation of NDMPC in building energy systems,offering concrete strategies for modeling choices,parameter tuning,and system architecture design.展开更多
https://www.sciencedirect.com/journal/energy-and-buildings/vol/338/suppl/C Volume 338,1 July 2025[OA](1)Real long-term performance evaluation of an improved office building operation involving a Data-driven model pred...https://www.sciencedirect.com/journal/energy-and-buildings/vol/338/suppl/C Volume 338,1 July 2025[OA](1)Real long-term performance evaluation of an improved office building operation involving a Data-driven model predictive control by Peter Klanatsky,Fran ois Veynandt,Christian Heschl,et al,Article 115590 Abstract:Data-driven Model Predictive Control(DMPC)strategies,coupled with holistically optimized HVAC system control,represent a promising approach to achieve climate targets through significant reductions in building energy consumption and associated emissions.To validate this potential in a real-world environment,a comprehensive optimization study was conducted on an office building serving as a living laboratory.Through systematic analysis of historical operational data,multiple Energy Conservation Measures(ECMs)were identified and implemented.The cornerstone of these improvements was the development and deployment of a centralized adaptive DMPC system,which was operated and evaluated over a full year.展开更多
Stability of a networked predictive control system subject to network-induced delay and data dropout is investigated in this study. By modeling the closed-loop system as a switched system with an upper-triangular stru...Stability of a networked predictive control system subject to network-induced delay and data dropout is investigated in this study. By modeling the closed-loop system as a switched system with an upper-triangular structure, a necessary and sufficient stability criterion is developed. From the criterion, it also can be seen that separation principle holds for networked predictive control systems. A numerical example is provided to confirm the validity and effectiveness of the obtained results.展开更多
The growing penetration of Electric Vehicles(EVs)in transportation brings challenges to power distribution systems due to uncertain usage patterns and increased peak loads.Effective EV fleet charging management strate...The growing penetration of Electric Vehicles(EVs)in transportation brings challenges to power distribution systems due to uncertain usage patterns and increased peak loads.Effective EV fleet charging management strategies are needed to minimize network impacts,such as peak charging power.While existing studies have addressed uncertainties in future arrivals,they often overlook the uncertainties in user-provided inputs of current ongoing charging EVs,such as estimated departure time and energy demand.This paper analyzes the impact of these uncertainties and evaluates three management strategies:a baseline Model Predictive Control(MPC),a data-hybrid MPC,and a fully data-driven Deep Reinforcement Learning(DRL)approach.For data-hybrid MPC,we adopted a diffusion model to handle user input uncertainties and a Gaussian Mixture Model for modeling arrival/departure scenarios.Additionally,the DRL method is based on a Partially Observable Markov Decision Process(POMDP)to manage uncertainty and employs a Convolutional Neural Network(CNN)for feature extraction.Robustness tests under different user uncertainty levels show that the data hybrid MPC performs better on the baseline MPC by 20%,while the DRL-based method achieves around 10%improvement.展开更多
基金financial support provided by the BMWK(Federal Ministry for Economic Affairs and Climate Action,Germany),promotional reference 03EN1006A.
文摘The implementation of sophisticated control strategies for building energy systems is crucial for improving energy efficiency and occupant comfort.While nonlinear model predictive control offers promising benefits,its application to large-scale building systems remains challenging due to computational complexity and system coupling.This work presents a comprehensive study of Nonlinear Distributed Model Predictive Control(NDMPC)implementation for building energy systems,comparing Alternating Direction Method of Multipliers(ADMM)and Augmented Lagrangian Alternating Direction Inexact Newton(ALADIN)algorithms alongside different modeling approaches.We examine a multi-zone heating system with thermal storage and multiple producers,investigating both Ordinary Differential Equation(ODE)-based and Artificial Neural Network(ANN)based modeling strategies.Through systematic parameter tuning using Bayesian optimization and closedloop scaling analysis with up to 40 thermal zones,we demonstrate that ALADIN-based NDMPC can achieve performance comparable to centralized model predictive control,showing greater robustness to parameter variations than ADMM.Our results reveal that ANN-based models effectively mitigate distributed integration errors and significantly reduce computation time compared to ODE-based approaches.Detailed computational profiling identifies specific bottlenecks in different NDMPC components.These findings advance the practical implementation of NDMPC in building energy systems,offering concrete strategies for modeling choices,parameter tuning,and system architecture design.
文摘https://www.sciencedirect.com/journal/energy-and-buildings/vol/338/suppl/C Volume 338,1 July 2025[OA](1)Real long-term performance evaluation of an improved office building operation involving a Data-driven model predictive control by Peter Klanatsky,Fran ois Veynandt,Christian Heschl,et al,Article 115590 Abstract:Data-driven Model Predictive Control(DMPC)strategies,coupled with holistically optimized HVAC system control,represent a promising approach to achieve climate targets through significant reductions in building energy consumption and associated emissions.To validate this potential in a real-world environment,a comprehensive optimization study was conducted on an office building serving as a living laboratory.Through systematic analysis of historical operational data,multiple Energy Conservation Measures(ECMs)were identified and implemented.The cornerstone of these improvements was the development and deployment of a centralized adaptive DMPC system,which was operated and evaluated over a full year.
基金supported by the National Natural Science Foundation of China(Grant Nos.6110409761321002+3 种基金61120106010&61522303)the Research Fund for the Doctoral Program of Higher Education of China(Grant No.20111101120027)the Program for New Century Excellent Talents in University(Grant No.NCET-13-0045)Beijing Higher Education Young Elite Teacher Project
文摘Stability of a networked predictive control system subject to network-induced delay and data dropout is investigated in this study. By modeling the closed-loop system as a switched system with an upper-triangular structure, a necessary and sufficient stability criterion is developed. From the criterion, it also can be seen that separation principle holds for networked predictive control systems. A numerical example is provided to confirm the validity and effectiveness of the obtained results.
基金supported by the National Research Foundation,Prime Minister’s Office,Singapore under its Campus for Research Excellence and Technological Enterprise(CREATE)programme.
文摘The growing penetration of Electric Vehicles(EVs)in transportation brings challenges to power distribution systems due to uncertain usage patterns and increased peak loads.Effective EV fleet charging management strategies are needed to minimize network impacts,such as peak charging power.While existing studies have addressed uncertainties in future arrivals,they often overlook the uncertainties in user-provided inputs of current ongoing charging EVs,such as estimated departure time and energy demand.This paper analyzes the impact of these uncertainties and evaluates three management strategies:a baseline Model Predictive Control(MPC),a data-hybrid MPC,and a fully data-driven Deep Reinforcement Learning(DRL)approach.For data-hybrid MPC,we adopted a diffusion model to handle user input uncertainties and a Gaussian Mixture Model for modeling arrival/departure scenarios.Additionally,the DRL method is based on a Partially Observable Markov Decision Process(POMDP)to manage uncertainty and employs a Convolutional Neural Network(CNN)for feature extraction.Robustness tests under different user uncertainty levels show that the data hybrid MPC performs better on the baseline MPC by 20%,while the DRL-based method achieves around 10%improvement.
