IEEE Transactions on Industrial Cyber-Physical Systems最新文献_第9页

APPO: Task Scheduling With Deep Reinforcement Learning in Cloud–Fog Collaborative Computing for Industrial Cyber-Physical Systems 基于深度强化学习的工业信息物理系统云雾协同计算任务调度

IEEE Transactions on Industrial Cyber-Physical Systems Pub Date : 2026-01-01 Epub Date: 2026-03-26 DOI: 10.1109/TICPS.2026.3697169

Yuzhe Huang;Huahu Xu;Honghao Gao

{"title":"APPO: Task Scheduling With Deep Reinforcement Learning in Cloud–Fog Collaborative Computing for Industrial Cyber-Physical Systems","authors":"Yuzhe Huang;Huahu Xu;Honghao Gao","doi":"10.1109/TICPS.2026.3697169","DOIUrl":"https://doi.org/10.1109/TICPS.2026.3697169","url":null,"abstract":"With the large-scale deployment of industrial cyber-physical system (ICPS) terminals such as drones and robots, edge computing often falls short in multidevice collaboration scenarios. Fog computing overcomes this issue by integrating different resources across a distributed node network, thereby enabling more efficient system-wide coordination. However, the limited resources of fog nodes make independently addressing the efficient scheduling demands of massive devices and complex tasks difficult. To address this problem, an attention-enhanced proximal policy optimization (APPO) algorithm is proposed in this paper. First, a dual-branch attention-based architecture is designed to model the complex dependencies between task requirements and multilayer resource states, thereby effectively capturing the coupling relationships in a hierarchical environment. Next, the PPO framework is leveraged with a centralized critic that integrates global system views, which enables coordinated scheduling decisions that are optimized across a latency-sensitive fog layer and a resource-rich cloud layer. Finally, a fault tolerance strategy is constructed to address the higher failure probabilities of distributed fog nodes through real-time monitoring and dynamic rescheduling, which significantly increases the robustness of the system. The experimental results demonstrate that the proposed APPO algorithm outperforms the existing baseline approaches and achieves a 28.15% higher task success rate while reducing virtual machine (VM) costs, energy consumption, and average delays by 20.94%, 21.33%, and 56.12%, respectively.","PeriodicalId":100640,"journal":{"name":"IEEE Transactions on Industrial Cyber-Physical Systems","volume":"4 ","pages":"380-391"},"PeriodicalIF":0.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148236112","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Resilient Prediction Event-Triggered Control Based on Attack Detection for Cyber-Physical Systems 基于攻击检测的网络物理系统弹性预测事件触发控制

IF 7.4

IEEE Transactions on Industrial Cyber-Physical Systems Pub Date : 2026-01-01 Epub Date: 2026-07-31 DOI: 10.1109/TICPS.2026.3719244

Zhen Wang;Wei Chen;Yingkang Xie;Guozeng Cui;Xia Huang

{"title":"Resilient Prediction Event-Triggered Control Based on Attack Detection for Cyber-Physical Systems","authors":"Zhen Wang;Wei Chen;Yingkang Xie;Guozeng Cui;Xia Huang","doi":"10.1109/TICPS.2026.3719244","DOIUrl":"https://doi.org/10.1109/TICPS.2026.3719244","url":null,"abstract":"This article focuses on cyber-physical systems subject to unknown disturbances and denial-of-service attacks. To ensure input-to-state stability, an event-triggered predictive control update scheme based on attack detection and a predictor is proposed. Firstly, a attack detection strategy is adopted, which makes full use of historical signals to detect whether the system is under attack at the current moment. Then, an attack detection-based control update scheme is proposed to compensate for state loss, and an event-triggered mechanism integrating detection and prediction is established to reduce resource consumption while ensuring system stability. The results show that the closed-loop cyber-physical systems can achieve input-to-state stability under the proposed control update scheme and event-triggered mechanism. An important advantage of the proposed control update scheme is that the cyber-physical systems select different state inputs according to the attack status at the event-triggered moment, thereby enabling the system to tolerate more adverse denial-of-service attacks. Finally, a simulation case is provided to verify the effectiveness of the proposed method.","PeriodicalId":100640,"journal":{"name":"IEEE Transactions on Industrial Cyber-Physical Systems","volume":"4 ","pages":"762-770"},"PeriodicalIF":7.4,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148836817","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A Multi-Channel Transmission Schedule in CPSs With the Energy Harvesting Defender Subject to DoS Attacks DoS攻击下能量收集防御器的cps多通道传输调度

IF 7.4

IEEE Transactions on Industrial Cyber-Physical Systems Pub Date : 2026-01-01 Epub Date: 2026-08-19 DOI: 10.1109/TICPS.2026.3725647

Wei Xing;Jindong Cheng;Junfeng Zhang;Haoyue Yang;Xudong Zhao

{"title":"A Multi-Channel Transmission Schedule in CPSs With the Energy Harvesting Defender Subject to DoS Attacks","authors":"Wei Xing;Jindong Cheng;Junfeng Zhang;Haoyue Yang;Xudong Zhao","doi":"10.1109/TICPS.2026.3725647","DOIUrl":"https://doi.org/10.1109/TICPS.2026.3725647","url":null,"abstract":"This paper investigates the optimal power scheduling for accurate remote state estimation in cyber-physical systems (CPSs) subject to denial-of-service attacks. A critical scenario is examined where sensor data is transmitted to a remote estimator over a multi-channel network that is vulnerable to a malicious attacker. At each time step, the sensor—equipped with an energy harvester—is to select a channel for transmitting data packets with the objective of minimizing the expected estimation error at the remote estimator. Concurrently, the attacker needs to decide which channel to target in order to maximize the expected estimation error, under constraints imposed by limited energy resources. To model the interactive decision-making process between the sensor and the attacker, a two-player Markov game framework is employed, capturing the conflicting objectives of the two players. Furthermore, an adaptive reinforcement learning algorithm named Q-learning algorithm is proposed to address the complexity of finding Nash equilibrium. Finally, simulation results demonstrate the effectiveness of the proposed approach.","PeriodicalId":100640,"journal":{"name":"IEEE Transactions on Industrial Cyber-Physical Systems","volume":"4 ","pages":"816-824"},"PeriodicalIF":7.4,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148925338","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Coordinated Defense Against Cyber-Physical Attacks in Power System Restoration 电力系统恢复中协同防御网络物理攻击

IF 7.4

IEEE Transactions on Industrial Cyber-Physical Systems Pub Date : 2026-01-01 Epub Date: 2026-09-07 DOI: 10.1109/TICPS.2026.3731126

Yansong Bai;Shaoyan Li;Jingzhe Tu;Xueping Gu

{"title":"Coordinated Defense Against Cyber-Physical Attacks in Power System Restoration","authors":"Yansong Bai;Shaoyan Li;Jingzhe Tu;Xueping Gu","doi":"10.1109/TICPS.2026.3731126","DOIUrl":"https://doi.org/10.1109/TICPS.2026.3731126","url":null,"abstract":"In recent years, geopolitical conflicts have dramatically increased the likelihood of major blackouts triggered by deliberate attacks. During power system restoration (PSR), sustained deliberate attacks can further target the vulnerable restored systems, exacerbating power outage losses. It is of great significance to enhance the defense capability of the restored systems under deliberate attack scenarios. To this end, this paper proposes a decision-making method for enhancing the resilience of PSR under the risk of cyber-physical coordinated attacks. First, the risk caused by deliberate attacks on restoration components during PSR is analyzed. From the aspects of topology and system operation, indicators such as restored network connectivity, scale-free property, outward expansion capability, and loading rate balance are extracted to reflect the attack risk level of the restored systems. Next, based on the measurement characteristics of PMUs, the observability constraints of sequential PSR schemes are established. The observability redundancy indicator is introduced to enhance the resilience of restored cyber systems against attacks. Then, a sequential PSR model considering the above cyber-physical defense indicators is established, the deloading characteristics of wind farms (WFs) and energy storage stations (ESSs) for frequency response are modeled, and a rolling solution framework adaptable to attack scenarios is designed. Finally, the objective function and constraints of the PSR model are transformed by weight coefficient selection and big-M method. The New England 39-bus system and an actual system case validate the effectiveness of the proposed method for significantly enhancing cyber-physical coordinated defense capability while ensuring restoration benefits during PSR.","PeriodicalId":100640,"journal":{"name":"IEEE Transactions on Industrial Cyber-Physical Systems","volume":"4 ","pages":"983-996"},"PeriodicalIF":7.4,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149130722","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A Lightweight Cognitive Digital Twin for Fault-Aware Control of Cyber–Physical Systems: Stochastic Modeling, Anomaly Detection, and Experimental Validation 用于信息物理系统故障感知控制的轻量级认知数字孪生：随机建模、异常检测和实验验证

IF 7.4

IEEE Transactions on Industrial Cyber-Physical Systems Pub Date : 2026-01-01 Epub Date: 2026-09-08 DOI: 10.1109/TICPS.2026.3732274

Mfonobong Uko

{"title":"A Lightweight Cognitive Digital Twin for Fault-Aware Control of Cyber–Physical Systems: Stochastic Modeling, Anomaly Detection, and Experimental Validation","authors":"Mfonobong Uko","doi":"10.1109/TICPS.2026.3732274","DOIUrl":"https://doi.org/10.1109/TICPS.2026.3732274","url":null,"abstract":"This paper presents a lightweight Cognitive Digital Twin (CDT) framework for fault-aware monitoring and control of industrial Cyber-Physical Systems (CPS). The framework integrates: (i) a stochastic fault progression model using coupled differential equations with physically interpretable parameters, (ii) a computationally efficient anomaly detection method based on information-geometric divergence measures with provable false alarm bounds, and (iii) a fault-aware optimal control strategy with input-to-state practical stability guarantees. Complete proofs are provided for all theoretical results. Validation is conducted through two complementary approaches: Monte Carlo simulation using a turbine-inspired CPS simulator (20 trials per scenario across four fault types), and a synthetic vibration benchmark modelled on the Case Western Reserve University bearing fault paradigm. Comparisons against Extended Kalman Filter-based fault detection, robust Model Predictive Control, Proportional-Integral, and reactive baseline controllers demonstrate that the CDT achieves 85–98% true positive detection rate with detection lead times of 46–542 steps (0.5–5.4 s). Under bearing fault conditions, the CDT maintains a stability index of 0.902 compared to 0.516 for the reactive baseline, a 75% improvement. The framework maintains robustness under Student-<inline-formula><tex-math>$t$</tex-math></inline-formula> and Gaussian mixture noise distributions, detects faults across a fourfold range of progression severities with 100% success rate, and degrades gracefully under progressive sensor loss. Computational cost remains below 0.1 ms per control cycle on standard hardware.","PeriodicalId":100640,"journal":{"name":"IEEE Transactions on Industrial Cyber-Physical Systems","volume":"4 ","pages":"1026-1036"},"PeriodicalIF":7.4,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149130706","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Dual-Data-Protection-Based Self-Regulating Cooperative Reinforcement Learning Control for Multiagent Systems 基于双数据保护的多智能体系统自调节协同强化学习控制

IF 7.4

IEEE Transactions on Industrial Cyber-Physical Systems Pub Date : 2026-01-01 Epub Date: 2026-09-22 DOI: 10.1109/TICPS.2026.3736752

Min Wang;Yan Lei;Xiao-Jie Peng;Hongyi Li;Tingwen Huang

{"title":"Dual-Data-Protection-Based Self-Regulating Cooperative Reinforcement Learning Control for Multiagent Systems","authors":"Min Wang;Yan Lei;Xiao-Jie Peng;Hongyi Li;Tingwen Huang","doi":"10.1109/TICPS.2026.3736752","DOIUrl":"https://doi.org/10.1109/TICPS.2026.3736752","url":null,"abstract":"Based on the cooperative simplified reinforcement learning framework, this paper proposes a dual-data-protection control algorithm for multiagent systems (MASs) with nonaffine nonlinear faults. The proposed algorithm is capable of ensuring the privacy of dual-channel data while enhancing the privacy-preserving capability. Specifically, a self-preservation mechanism is presented to protect the data inside the agent, and the constructed auxiliary system effectively mitigates the performance degradation of MASs. For inter-agent communication security, a homomorphic encryption scheme is implemented, which consistently maintains a high level of privacy protection without compromising the system performance. The cooperative simplified reinforcement learning control strategy incorporating self-regulating neighbor weights is presented to improve the robustness of MASs. Furthermore, nonaffine nonlinear faults are addressed by a Butterworth low-pass filter. A simulation example on torsional pendulum systems is utilized to demonstrate the effectiveness of the proposed control scheme.","PeriodicalId":100640,"journal":{"name":"IEEE Transactions on Industrial Cyber-Physical Systems","volume":"4 ","pages":"1063-1073"},"PeriodicalIF":7.4,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149188342","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A Novel MMD-Critic-XAI-Based Intrusion Detection System for Cyber-Physical Microgrids 一种基于mmd - critical - xai的网络物理微电网入侵检测系统

IF 7.4

IEEE Transactions on Industrial Cyber-Physical Systems Pub Date : 2026-01-01 Epub Date: 2026-09-07 DOI: 10.1109/TICPS.2026.3731921

Gurmaan S. Panesar;Matthew Oinonen;Walid G. Morsi

{"title":"A Novel MMD-Critic-XAI-Based Intrusion Detection System for Cyber-Physical Microgrids","authors":"Gurmaan S. Panesar;Matthew Oinonen;Walid G. Morsi","doi":"10.1109/TICPS.2026.3731921","DOIUrl":"https://doi.org/10.1109/TICPS.2026.3731921","url":null,"abstract":"Cyber-physical microgrids are the target of attacks due to their reliance on internet-based automation and digital controlling devices that are not designed with built-in security. As these microgrids rely heavily on the distributed energy resources (DERs) within their systems, such cyberattacks can have consequences of high impact to such critical systems leading to power outages, equipment damage and data theft. Traditional intrusion detection systems (IDSs) used machine learning models to detect and classify such attacks; however, such models are black-boxes lacking interpretability and thus they become untrustworthy. This paper presents a novel trustworthy IDS for detecting and classifying attacks alongside disturbances in microgrids using an example-based Explainable Artificial Intelligence (XAI), namely Maximal Mean Discrepancy (MMD)-Critic, to provide trustworthy intrusion decisions. To enhance the time-frequency resolution for best feature representations in scalograms, MMD-Critic is utilized to identify the optimal voice parameters in the Continuous Wavelet Transform (CWT), enabling the Convolutional Neural Network (CNN) to accurately classify each class. The proposed IDS has been tested in real-time using OPAL-RT and IEC61850 compliant Siemens-SIPROTEC Intelligent Electronic Device (IED) to generate real Generic Object-Oriented Substation Events (GOOSE) messages. The results have shown that the use of MMD-Critic increases the intrusion detection accuracy to 98.85% by identifying individual cases where the CNN misclassified. Finally, a user interface is developed to allow grid operators to understand the model's decisions.","PeriodicalId":100640,"journal":{"name":"IEEE Transactions on Industrial Cyber-Physical Systems","volume":"4 ","pages":"1037-1048"},"PeriodicalIF":7.4,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149188343","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Age of Information in Industrial Cyber-Physical Systems: A Survey on Co-Design, Optimization, and Future Directions 工业信息物理系统的信息时代：协同设计、优化和未来方向的调查

IEEE Transactions on Industrial Cyber-Physical Systems Pub Date : 2026-01-01 Epub Date: 2026-06-03 DOI: 10.1109/TICPS.2026.3700093

Xiaoling Zhang;Xiao Xie;Shuhao Qiang;Huanfei Yan;Zhibo Pang;Kan Yu

{"title":"Age of Information in Industrial Cyber-Physical Systems: A Survey on Co-Design, Optimization, and Future Directions","authors":"Xiaoling Zhang;Xiao Xie;Shuhao Qiang;Huanfei Yan;Zhibo Pang;Kan Yu","doi":"10.1109/TICPS.2026.3700093","DOIUrl":"https://doi.org/10.1109/TICPS.2026.3700093","url":null,"abstract":"The Age of Information (AoI) has emerged as a critical metric for quantifying information freshness in Industrial Cyber-Physical Systems (ICPS), where timely data is paramount for control performance and safety. While numerous optimization strategies exist, the field lacks a systematic overview structured from a functional co-design perspective. Our work provides a comprehensive survey that synthesizes and categorizes the state-of-the-art in AoI optimization by structuring the field around the core operational functionalities of ICPS: sampling and estimation, scheduling and optimization, as well as control and communication co-design. We first establish the theoretical foundations of AoI and its key variants. Then, for each functional block, we critically review and compare both foundational model-based strategies and advanced learning-based paradigms. A particular emphasis is placed on complex, real-world scenarios, offering a multi-dimensional evaluation of optimization strategies to highlight the trade-offs between different methodologies. Finally, we identify critical unresolved challenges, spanning from the theory-practice gap to the need for lightweight, robust algorithms, and outline a concrete roadmap for future research. This survey serves as a structured guide for researchers and practitioners, aiming to advance the field by bridging theoretical frameworks with the practical demands of next-generation industrial systems.","PeriodicalId":100640,"journal":{"name":"IEEE Transactions on Industrial Cyber-Physical Systems","volume":"4 ","pages":"429-441"},"PeriodicalIF":0.0,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148236103","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Deep Learning-Based Intelligent Intrusion Detection With Optimized Feature Selection for IoT Network Security 基于深度学习优化特征选择的物联网网络安全智能入侵检测

IF 7.4

IEEE Transactions on Industrial Cyber-Physical Systems Pub Date : 2026-01-01 Epub Date: 2026-07-01 DOI: 10.1109/TICPS.2026.3708811

Surendra Kumar;Avinash Kumar;Mridula Dwivedi;Mohit Kumar;Huaming Wu;Sukhpal Singh Gill

{"title":"Deep Learning-Based Intelligent Intrusion Detection With Optimized Feature Selection for IoT Network Security","authors":"Surendra Kumar;Avinash Kumar;Mridula Dwivedi;Mohit Kumar;Huaming Wu;Sukhpal Singh Gill","doi":"10.1109/TICPS.2026.3708811","DOIUrl":"https://doi.org/10.1109/TICPS.2026.3708811","url":null,"abstract":"Intrusion detection is crucial to network security, especially in Internet of Things (IoT) contexts with rising cyberthreats. Traditional Intrusion Detection Systems (IDSs) can cause false alerts and are difficult to detect new attack types. Metaheuristic optimization can enhance detection rate, model complexity, and feature selection, but it lacks scalability and flexibility to changing situations. Nature-inspired methods enable deep learning models to focus on key features. Current existing methods face challenges in securing the network from several attack types and adapting to evolving attack patterns effectively, identifying changing attack patterns in IoT networks with great accuracy. This paper presents a novel IoT-aware intrusion detection framework that integrates the Whale Optimization Algorithm (WOA), Dipper Throated Optimization (DTO), and Ant Lion Optimization (ALO) for optimized feature selection, alongside Long Short-Term Memory and eXtreme Gradient Boosting for accurate anomaly detection. The proposed method effectively balances exploration and exploitation to select informative features, while the learning models capture temporal and nonlinear patterns in network traffic. Experimental results evaluated on benchmark datasets, including NSL-KDD, KDDCup99, CICDDoS2019, and NF-UNSW-NB15-v2, demonstrate that the proposed framework achieves high detection performance with reduced false alarms. The proposed frameworkimproves computational efficiency and maintains strong detection capability across multiple attack categories,making it scalable and secure for real-time IoT intrusion detection.","PeriodicalId":100640,"journal":{"name":"IEEE Transactions on Industrial Cyber-Physical Systems","volume":"4 ","pages":"584-592"},"PeriodicalIF":7.4,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148481190","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Cross-Scenario Robustness Analysis of Deep Reinforcement Learning–Based Energy Management in Wind-Integrated Flexible Smart Traction Power Supply System 基于深度强化学习的风力集成柔性智能牵引供电系统能量管理跨场景鲁棒性分析

IF 7.4

IEEE Transactions on Industrial Cyber-Physical Systems Pub Date : 2026-01-01 Epub Date: 2026-07-07 DOI: 10.1109/TICPS.2026.3710873

Yiyan Sang;Xijiao Ren;Yichen Ying;Jiamin Sheng;Suhan Zhang;Lie Xu;Agustí Egea-Àlvarez;Jianbin Zhang

{"title":"Cross-Scenario Robustness Analysis of Deep Reinforcement Learning–Based Energy Management in Wind-Integrated Flexible Smart Traction Power Supply System","authors":"Yiyan Sang;Xijiao Ren;Yichen Ying;Jiamin Sheng;Suhan Zhang;Lie Xu;Agustí Egea-Àlvarez;Jianbin Zhang","doi":"10.1109/TICPS.2026.3710873","DOIUrl":"https://doi.org/10.1109/TICPS.2026.3710873","url":null,"abstract":"The increasing penetration of distributed renewable energy in electrified railway systems has promoted the development of flexible smart traction power supply systems (FSTPSSs). However, the strong intermittency and scenario-dependent dynamics of wind power introduce substantial uncertainty, posing significant challenges to the robustness of energy management strategies. Although deep reinforcement learning (DRL) has shown promise in addressing short-term renewable variability in FSTPSS, its robustness across heterogeneous wind power scenarios has not been systematically investigated. This paper develops a scenario-oriented robustness evaluation framework for DRL-based energy management in wind-integrated FSTPSS. Representative wind power scenarios are identified using the K-medoids clustering method, and scenario-specific stochastic fluctuation parameters are extracted to characterize intra-scenario uncertainty. On this basis, the robustness of DRL agents trained under different scenario exposures is quantitatively evaluated and compared when deployed under previously unseen operating conditions. Simulation results indicate that agents trained with diversified scenario experience achieve significantly lower mean square error (MSE) and mean absolute percentage error (MAPE) in reference power tracking when facing new wind power scenarios. These results demonstrate that explicitly incorporating scenario diversity into the training process is essential for enhancing the robustness and generalization capability of DRL-based energy management strategies under dynamic wind environments.","PeriodicalId":100640,"journal":{"name":"IEEE Transactions on Industrial Cyber-Physical Systems","volume":"4 ","pages":"665-677"},"PeriodicalIF":7.4,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148612943","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0