Hossein Hosseinalibeiki
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
Hossein Hosseinalibeiki has a total of 27 co-authors. Here are the top co-authors by number of shared publications:
- Mahdieh S. Sadabadi
The University of Manchester
9 shared publications - Reza Sepehrzad
Polytechnic of Milan; Shiraz University
9 shared publications - Ahmed Al-Durra
Department of Electrical Engineering, Advanced Power and Energy Center, Khalifa University, Abu Dhabi, UAE
7 shared publications Nima Khosravi
7 shared publications- Nima Khosravi
Department of Electrical and Instrumentation Engineering, R&D Management of NPC, Tehran, Iran
4 shared publications - Ali Kazemian3 shared publications
Latest Publications
- Open Access
- 3 September 2026
- Published by Elsevier in SSRN Electronic Journal
Aggregator-Centric Computational Coordination in Residential Energy Communities: A Review of Heterogeneous Learning, Large Language Models, and Benchmarking
Residential energy communities are becoming a key operational setting for coordinating distributed energy resources, flexible demand, energy storage, electric vehicles, and local energy exchange. Their operation is challenging because household resources differ in availability, flexibility, behavior, and technical constraints, while aggregators must preserve feasible operation under uncertain generation, changing participation, network limits, and market rules. This review synthesizes 150 peer-reviewed studies on aggregator-centric coordination, with particular focus on heterogeneous learning and large language models (LLMs) as computational interfaces for demand response and peer-to-peer energy coordination. Three connected taxonomies are developed. Taxonomy I classifies centralized, distributed, hierarchical, and transactive coordination paradigms according to market interaction, coupling, and feasibility enforcement. Taxonomy II organizes heterogeneous-agent learning according to device, behavioral, socio-economic, and constraint heterogeneity, learning architecture, and robustness to behavioral drift, seasonal variation, uncertainty, and population change. Taxonomy III positions LLMs as structured operational interfaces for participation support, tariff and rule interpretation, context-aware forecasting, and cyber-security assistance rather than autonomous dispatch mechanisms. Building on these taxonomies, the review develops a comparative evaluation framework centered on scheduling feasibility, constraint satisfaction, scalability, information parity among baselines, robustness under distribution shift, auditability, reproducibility, and validation of LLM-derived outputs. It further formulates testable research questions and minimum reporting and benchmarking requirements. Together, these contributions provide a unified basis for evaluating whether advanced computational and learning-based approaches can support reliable, reproducible, and sustainable coordination in heterogeneous residential energy communities.
- 3 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessAggregator-Centric Computational Coordination in Residential Energy Communities: A Review of Heterogeneous Learning, Large Language Models, and Benchmarking
Residential energy communities are becoming a key operational setting for coordinating distributed energy resources, flexible demand, energy storage, electric vehicles, and local energy exchange. Their operation is challenging because household resources differ in availability, flexibility, behavior, and technical constraints, while aggregators must preserve feasible operation under uncertain generation, changing participation, network limits, and market rules. This review synthesizes 150 peer-reviewed studies on aggregator-centric coordination, with particular focus on heterogeneous learning and large language models (LLMs) as computational interfaces for demand response and peer-to-peer energy coordination. Three connected taxonomies are developed. Taxonomy I classifies centralized, distributed, hierarchical, and transactive coordination paradigms according to market interaction, coupling, and feasibility enforcement. Taxonomy II organizes heterogeneous-agent learning according to device, behavioral, socio-economic, and constraint heterogeneity, learning architecture, and robustness to behavioral drift, seasonal variation, uncertainty, and population change. Taxonomy III positions LLMs as structured operational interfaces for participation support, tariff and rule interpretation, context-aware forecasting, and cyber-security assistance rather than autonomous dispatch mechanisms. Building on these taxonomies, the review develops a comparative evaluation framework centered on scheduling feasibility, constraint satisfaction, scalability, information parity among baselines, robustness under distribution shift, auditability, reproducibility, and validation of LLM-derived outputs. It further formulates testable research questions and minimum reporting and benchmarking requirements. Together, these contributions provide a unified basis for evaluating whether advanced computational and learning-based approaches can support reliable, reproducible, and sustainable coordination in heterogeneous residential energy communities.
- Open Access
- 6 June 2026
- Published by Elsevier in SSRN Electronic Journal
Safety-Filtered Multi-Agent Learning for Demand Response Under Hybrid Disruptions in Renewable-Rich Integrated Energy Systems
Demand response in integrated energy systems must remain effective under renewable and load uncertainty, volatile market conditions, and hybrid cyber disruptions affecting sensing, communication, and control channels. This paper proposes a risk-aware large language model-guided multi-agent deep reinforcement learning framework for cyber-resilient demand response in integrated energy systems. The framework integrates computational intelligence, communication-aware cyber modeling, market settlement, and real-time electrical-energy control to coordinate distributed flexible resources, including electrical and thermal loads, renewable generation, and heterogeneous storage assets. To improve robustness under stressed operating regimes, it combines cyber- and uncertainty-aware state representations, adversarial and domain-randomized training, and a conditional value-at-risk regularization term to limit adverse market and operational outcomes. It also captures the cascading interaction between operational uncertainty and cyber disruption, where uncertainty increases delivery risk and weakens observability, while cyberattacks amplify effective uncertainty in estimation, dispatch, and market delivery. In addition, a constrained action repair and feasibility projection layer is introduced to maintain executable control actions within device, comfort, communication, and network operating limits during real-time operation. The proposed method is evaluated on the IEEE 69-bus, Caracas 141-bus, and IEEE 342-node systems. Results show that, under stressed conditions, the framework adaptively reallocates flexibility across controllable resources, improving delivery stability and limiting operational violations. Under combined uncertainty and cyberattacks, it sustains 93.8–96.4% demand response delivery with 1.2–2.3% violation rates while preserving robust profitability and stronger tail performance than advanced learning-based baselines. These results demonstrate a scalable computational framework for cyber-resilient and risk-aware demand response in multi-energy distribution environments.
- 6 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessSafety-Filtered Multi-Agent Learning for Demand Response Under Hybrid Disruptions in Renewable-Rich Integrated Energy Systems
Demand response in integrated energy systems must remain effective under renewable and load uncertainty, volatile market conditions, and hybrid cyber disruptions affecting sensing, communication, and control channels. This paper proposes a risk-aware large language model-guided multi-agent deep reinforcement learning framework for cyber-resilient demand response in integrated energy systems. The framework integrates computational intelligence, communication-aware cyber modeling, market settlement, and real-time electrical-energy control to coordinate distributed flexible resources, including electrical and thermal loads, renewable generation, and heterogeneous storage assets. To improve robustness under stressed operating regimes, it combines cyber- and uncertainty-aware state representations, adversarial and domain-randomized training, and a conditional value-at-risk regularization term to limit adverse market and operational outcomes. It also captures the cascading interaction between operational uncertainty and cyber disruption, where uncertainty increases delivery risk and weakens observability, while cyberattacks amplify effective uncertainty in estimation, dispatch, and market delivery. In addition, a constrained action repair and feasibility projection layer is introduced to maintain executable control actions within device, comfort, communication, and network operating limits during real-time operation. The proposed method is evaluated on the IEEE 69-bus, Caracas 141-bus, and IEEE 342-node systems. Results show that, under stressed conditions, the framework adaptively reallocates flexibility across controllable resources, improving delivery stability and limiting operational violations. Under combined uncertainty and cyberattacks, it sustains 93.8–96.4% demand response delivery with 1.2–2.3% violation rates while preserving robust profitability and stronger tail performance than advanced learning-based baselines. These results demonstrate a scalable computational framework for cyber-resilient and risk-aware demand response in multi-energy distribution environments.
- Open Access
- 5 June 2026
- Published by Elsevier in SSRN Electronic Journal
Safety-Filtered Multi-Agent Learning for Demand Response Under Hybrid Disruptions in Renewable-Rich Integrated Energy Systems
Demand response in integrated energy systems must remain effective under renewable and load uncertainty, volatile market conditions, and hybrid cyber disruptions affecting sensing, communication, and control channels. This paper proposes a risk-aware large language model-guided multi-agent deep reinforcement learning framework for cyber-resilient demand response in integrated energy systems. The framework integrates computational intelligence, communication-aware cyber modeling, market settlement, and real-time electrical-energy control to coordinate distributed flexible resources, including electrical and thermal loads, renewable generation, and heterogeneous storage assets. To improve robustness under stressed operating regimes, it combines cyber- and uncertainty-aware state representations, adversarial and domain-randomized training, and a conditional value-at-risk regularization term to limit adverse market and operational outcomes. It also captures the cascading interaction between operational uncertainty and cyber disruption, where uncertainty increases delivery risk and weakens observability, while cyberattacks amplify effective uncertainty in estimation, dispatch, and market delivery. In addition, a constrained action repair and feasibility projection layer is introduced to maintain executable control actions within device, comfort, communication, and network operating limits during real-time operation. The proposed method is evaluated on the IEEE 69-bus, Caracas 141-bus, and IEEE 342-node systems. Results show that, under stressed conditions, the framework adaptively reallocates flexibility across controllable resources, improving delivery stability and limiting operational violations. Under combined uncertainty and cyberattacks, it sustains 93.8–96.4% demand response delivery with 1.2–2.3% violation rates while preserving robust profitability and stronger tail performance than advanced learning-based baselines. These results demonstrate a scalable computational framework for cyber-resilient and risk-aware demand response in multi-energy distribution environments.
- 5 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessSafety-Filtered Multi-Agent Learning for Demand Response Under Hybrid Disruptions in Renewable-Rich Integrated Energy Systems
Demand response in integrated energy systems must remain effective under renewable and load uncertainty, volatile market conditions, and hybrid cyber disruptions affecting sensing, communication, and control channels. This paper proposes a risk-aware large language model-guided multi-agent deep reinforcement learning framework for cyber-resilient demand response in integrated energy systems. The framework integrates computational intelligence, communication-aware cyber modeling, market settlement, and real-time electrical-energy control to coordinate distributed flexible resources, including electrical and thermal loads, renewable generation, and heterogeneous storage assets. To improve robustness under stressed operating regimes, it combines cyber- and uncertainty-aware state representations, adversarial and domain-randomized training, and a conditional value-at-risk regularization term to limit adverse market and operational outcomes. It also captures the cascading interaction between operational uncertainty and cyber disruption, where uncertainty increases delivery risk and weakens observability, while cyberattacks amplify effective uncertainty in estimation, dispatch, and market delivery. In addition, a constrained action repair and feasibility projection layer is introduced to maintain executable control actions within device, comfort, communication, and network operating limits during real-time operation. The proposed method is evaluated on the IEEE 69-bus, Caracas 141-bus, and IEEE 342-node systems. Results show that, under stressed conditions, the framework adaptively reallocates flexibility across controllable resources, improving delivery stability and limiting operational violations. Under combined uncertainty and cyberattacks, it sustains 93.8–96.4% demand response delivery with 1.2–2.3% violation rates while preserving robust profitability and stronger tail performance than advanced learning-based baselines. These results demonstrate a scalable computational framework for cyber-resilient and risk-aware demand response in multi-energy distribution environments.
- Open Access
- 22 May 2026
- Published by Elsevier in SSRN Electronic Journal
Heterogeneous Energy Storage-Supported Demand Response in Integrated Energy Systems: A Resilient Market-to-Delivery Framework under Compound Disturbances
Heterogeneous energy storage systems are central to preserving flexibility delivery in cyber-physical integrated energy systems (IES) when market-cleared commitments must remain feasible under uncertainty, cyber disturbances, communication degradation, and real-time operational deviations. This paper proposes a resilience-oriented market-to-delivery framework for storage-supported demand response (DR) in IES. The proposed model connects day-ahead flexibility clearing, multi-storage scheduling, learning-based decision support, real-time capability assessment, safety-filtered recourse, and verified service delivery within a unified hierarchical structure. In the day-ahead stage, a leader–follower market mechanism clears storage-supported load-to-grid commitments while considering distribution feasibility, settlement consistency, scenario-based risk exposure, and available storage capability. At the integrated energy system level, heterogeneous storage assets, conversion units, and comfort-aware flexible loads are coordinated to transform cleared commitments into device-feasible multi-carrier schedules. A constrained deep reinforcement learning (DRL) mechanism is used to solve the high-dimensional scheduling problem, while large language model (LLM) assistance is restricted to bounded reward calibration and storage-capability-aware repair-template selection rather than direct dispatch. During real-time execution, the cleared schedule is lifted into a fast delivery layer in which disturbance-aware estimation, storage-aware safety filtering, corrective recourse, fallback switching, and service verification determine whether the committed flexibility can be delivered, partially restored, or conservatively degraded. The framework therefore treats storage-supported DR not as a static cleared quantity, but as a physically constrained service whose credibility must be maintained from market commitment to real-time delivery. Numerical results demonstrate coherent scalability across benchmark, real-world, and large-scale systems, preserved storage-supported operational headroom, real-time delivery above 88% under compound disturbances, and bounded constraint excursions below 3%. Ablation, stability, sensitivity, and comparative analyses confirm that performance gains arise from the integrated storage-supported resilience chain rather than from isolated scheduling or control components.
- 22 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessHeterogeneous Energy Storage-Supported Demand Response in Integrated Energy Systems: A Resilient Market-to-Delivery Framework under Compound Disturbances
Heterogeneous energy storage systems are central to preserving flexibility delivery in cyber-physical integrated energy systems (IES) when market-cleared commitments must remain feasible under uncertainty, cyber disturbances, communication degradation, and real-time operational deviations. This paper proposes a resilience-oriented market-to-delivery framework for storage-supported demand response (DR) in IES. The proposed model connects day-ahead flexibility clearing, multi-storage scheduling, learning-based decision support, real-time capability assessment, safety-filtered recourse, and verified service delivery within a unified hierarchical structure. In the day-ahead stage, a leader–follower market mechanism clears storage-supported load-to-grid commitments while considering distribution feasibility, settlement consistency, scenario-based risk exposure, and available storage capability. At the integrated energy system level, heterogeneous storage assets, conversion units, and comfort-aware flexible loads are coordinated to transform cleared commitments into device-feasible multi-carrier schedules. A constrained deep reinforcement learning (DRL) mechanism is used to solve the high-dimensional scheduling problem, while large language model (LLM) assistance is restricted to bounded reward calibration and storage-capability-aware repair-template selection rather than direct dispatch. During real-time execution, the cleared schedule is lifted into a fast delivery layer in which disturbance-aware estimation, storage-aware safety filtering, corrective recourse, fallback switching, and service verification determine whether the committed flexibility can be delivered, partially restored, or conservatively degraded. The framework therefore treats storage-supported DR not as a static cleared quantity, but as a physically constrained service whose credibility must be maintained from market commitment to real-time delivery. Numerical results demonstrate coherent scalability across benchmark, real-world, and large-scale systems, preserved storage-supported operational headroom, real-time delivery above 88% under compound disturbances, and bounded constraint excursions below 3%. Ablation, stability, sensitivity, and comparative analyses confirm that performance gains arise from the integrated storage-supported resilience chain rather than from isolated scheduling or control components.
- Open Access
- 7 May 2026
- Published by Elsevier in SSRN Electronic Journal
Resilience-Oriented Market-to-Delivery Demand Response in Integrated Energy Systems under Dynamic Uncertainty and Cyber-Physical Disruptions: A Constrained Deep Reinforcement Learning Framework
Cyber-physical integrated energy systems (IESs) require demand response (DR) strategies that can preserve real-time service deliverability under dynamic uncertainty and cyber-physical disruptions. Existing studies often decouple market clearing, multi-storage scheduling, AI-enabled control, and resilient operation, leaving unclear how cleared DR commitments can be tracked, repaired, verified, or safely degraded during disturbed real-time execution. This paper proposes an AI-enabled resilience-oriented market-to-delivery DR framework that connects day-ahead commitment formation with real-time operational capability assessment and disturbance-aware recourse. In the day-ahead layer, a hierarchical leader–follower mechanism clears storage-supported DR commitments while considering distribution feasibility, settlement consistency, and the available capability of heterogeneous storage assets. At the IES layer, constrained deep reinforcement learning (DRL) supports multi-carrier scheduling through calibrated rewards and storage-capability-aware action repair, with large language model (LLM) assistance restricted to bounded calibration and repair-template selection rather than direct dispatch. In real time, cleared schedules are lifted into a fast execution layer where disturbance-aware state estimation, safety filtering, fallback switching, corrective recourse, and service verification determine whether committed DR can be delivered, partially restored, or conservatively degraded. The framework is validated on the IEEE 69-bus, Caracas 141-bus, and IEEE 342-bus systems under normal, uncertainty-only, and compound-disturbance scenarios combining dynamic operational uncertainty and cyber disruptions. Results show preserved operational headroom, coherent scaling across test systems, real-time DR delivery above 88% under compound disturbances, and bounded constraint excursions below 3%. Ablation, sensitivity, stability, and baseline comparison analyses confirm that performance gains arise from the integrated resilience chain rather than DRL alone.
- 7 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessResilience-Oriented Market-to-Delivery Demand Response in Integrated Energy Systems under Dynamic Uncertainty and Cyber-Physical Disruptions: A Constrained Deep Reinforcement Learning Framework
Cyber-physical integrated energy systems (IESs) require demand response (DR) strategies that can preserve real-time service deliverability under dynamic uncertainty and cyber-physical disruptions. Existing studies often decouple market clearing, multi-storage scheduling, AI-enabled control, and resilient operation, leaving unclear how cleared DR commitments can be tracked, repaired, verified, or safely degraded during disturbed real-time execution. This paper proposes an AI-enabled resilience-oriented market-to-delivery DR framework that connects day-ahead commitment formation with real-time operational capability assessment and disturbance-aware recourse. In the day-ahead layer, a hierarchical leader–follower mechanism clears storage-supported DR commitments while considering distribution feasibility, settlement consistency, and the available capability of heterogeneous storage assets. At the IES layer, constrained deep reinforcement learning (DRL) supports multi-carrier scheduling through calibrated rewards and storage-capability-aware action repair, with large language model (LLM) assistance restricted to bounded calibration and repair-template selection rather than direct dispatch. In real time, cleared schedules are lifted into a fast execution layer where disturbance-aware state estimation, safety filtering, fallback switching, corrective recourse, and service verification determine whether committed DR can be delivered, partially restored, or conservatively degraded. The framework is validated on the IEEE 69-bus, Caracas 141-bus, and IEEE 342-bus systems under normal, uncertainty-only, and compound-disturbance scenarios combining dynamic operational uncertainty and cyber disruptions. Results show preserved operational headroom, coherent scaling across test systems, real-time DR delivery above 88% under compound disturbances, and bounded constraint excursions below 3%. Ablation, sensitivity, stability, and baseline comparison analyses confirm that performance gains arise from the integrated resilience chain rather than DRL alone.
- Open Access
- 7 May 2026
- Published by Elsevier in SSRN Electronic Journal
Post-Disruption Resilient Recovery of Microgrid Real-Time Balancing Markets under Dynamic Uncertainty and Market-State Degradation
Real-time balancing markets coordinate distributed energy resources, energy storage systems, electric vehicle aggregators, flexible loads, and upstream-grid support under changing operating conditions. However, when dynamic uncertainty interacts with residual cyber-induced degradation after detection, partial containment, or incomplete attribution, normal market continuation may no longer be defensible, while complete suspension may waste available flexibility and interrupt balancing services. This paper proposes a trust-aware recovery-market framework for resilient microgrid (MG) real-time balancing markets. The framework addresses the intermediate operating region between unrestricted normal operation and fallback operation by modeling recovery as a degraded-trust market-continuation problem rather than as a cyberattack-detection or conventional uncertainty-aware dispatch task. Dynamic uncertainty and residual cyber-induced degradation are translated into four operational trust channels: data admissibility, cross-layer consistency, dispatch executability, and settlement-evidence confidence. These channels drive a trust-conditioned mode-transition mechanism that determines whether normal operation, bounded recovery, or fallback should be activated. At the solution level, Bayesian trust inference updates the continuation trust state under uncertain evidence, while safe deep reinforcement learning (safe DRL) generates candidate recovery actions over the reconstructed post-disruption market state. The generated actions are filtered through admissible projection and supervisory anti-oscillation governance before execution. A settlement-legitimacy layer then verifies whether the recovered outcome remains evidence-supported, sufficiently undistorted, and contractually defensible. Simulation studies on the IEEE 30-bus, Caracas 141-bus, and IEEE 300-bus systems show that the proposed framework preserves defensible recovery under residual disruption, achieving 91.8% recovery feasibility, 96.4% settlement legitimacy, and a 0.93 final defensibility score in the severe IEEE 30-bus case.
- 7 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessPost-Disruption Resilient Recovery of Microgrid Real-Time Balancing Markets under Dynamic Uncertainty and Market-State Degradation
Real-time balancing markets coordinate distributed energy resources, energy storage systems, electric vehicle aggregators, flexible loads, and upstream-grid support under changing operating conditions. However, when dynamic uncertainty interacts with residual cyber-induced degradation after detection, partial containment, or incomplete attribution, normal market continuation may no longer be defensible, while complete suspension may waste available flexibility and interrupt balancing services. This paper proposes a trust-aware recovery-market framework for resilient microgrid (MG) real-time balancing markets. The framework addresses the intermediate operating region between unrestricted normal operation and fallback operation by modeling recovery as a degraded-trust market-continuation problem rather than as a cyberattack-detection or conventional uncertainty-aware dispatch task. Dynamic uncertainty and residual cyber-induced degradation are translated into four operational trust channels: data admissibility, cross-layer consistency, dispatch executability, and settlement-evidence confidence. These channels drive a trust-conditioned mode-transition mechanism that determines whether normal operation, bounded recovery, or fallback should be activated. At the solution level, Bayesian trust inference updates the continuation trust state under uncertain evidence, while safe deep reinforcement learning (safe DRL) generates candidate recovery actions over the reconstructed post-disruption market state. The generated actions are filtered through admissible projection and supervisory anti-oscillation governance before execution. A settlement-legitimacy layer then verifies whether the recovered outcome remains evidence-supported, sufficiently undistorted, and contractually defensible. Simulation studies on the IEEE 30-bus, Caracas 141-bus, and IEEE 300-bus systems show that the proposed framework preserves defensible recovery under residual disruption, achieving 91.8% recovery feasibility, 96.4% settlement legitimacy, and a 0.93 final defensibility score in the severe IEEE 30-bus case.
- Open Access
- 28 April 2026
- Published by Elsevier in SSRN Electronic Journal
Credibility-Aware Demand Response Continuation and Regime-Switching Market Operation in Multi-Energy Microgrids under Hybrid Cyber Threats and Dynamic Uncertainty
Market-oriented multi-energy microgrids (MEMGs) are increasingly expected to sustain real-time market participation and demand response (DR) delivery under dynamic uncertainty and degraded cyber-physical conditions. Existing studies have advanced cyberattack detection, resilient control, robust scheduling, and uncertainty-aware energy management, yet they rarely formalize the intermediate operating state between fully normal market operation and complete fallback. This gap is critical because full suspension wastes flexible capacity, whereas unrestricted continuation under degraded trust can increase infeasible dispatch, unreliable commitments, and settlement exposure. This paper proposes a trust-conditioned, risk-aware, and safety-critical decision framework for MEMGs exposed to hybrid cyber threats and dynamic uncertainty. Degraded operation is modeled through measurement inconsistency, communication impairment, execution mismatch, market-signal inconsistency, and uncertainty pressure, which are converted into regime-aware trust and admissibility coefficients for market, DR, execution, and settlement-sensitive domains. A bounded continuation layer then maintains market and DR functions within contracted operating envelopes. At the solution level, a large language model (LLM)-guided multi-agent deep reinforcement learning (MADRL) architecture coordinates market, DR, conversion, and flexibility agents, with the LLM restricted to bounded advisory refinement rather than autonomous control. A CVaR-based learning objective limits upper-tail settlement and service-risk exposure, while a repair-and-proximal execution layer supports implementable actions. Simulation results show that the proposed framework improves conservative-continuation success, market participation, DR fulfillment, deviation containment, settlement exposure, infeasible-action reduction, and recovery behavior compared with advanced learning-based benchmarks. Real-feeder and scalability evaluations confirm that the framework preserves market and DR continuation under degraded conditions without relying on unsafe normal operation or unnecessary suspension.
- 28 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessCredibility-Aware Demand Response Continuation and Regime-Switching Market Operation in Multi-Energy Microgrids under Hybrid Cyber Threats and Dynamic Uncertainty
Market-oriented multi-energy microgrids (MEMGs) are increasingly expected to sustain real-time market participation and demand response (DR) delivery under dynamic uncertainty and degraded cyber-physical conditions. Existing studies have advanced cyberattack detection, resilient control, robust scheduling, and uncertainty-aware energy management, yet they rarely formalize the intermediate operating state between fully normal market operation and complete fallback. This gap is critical because full suspension wastes flexible capacity, whereas unrestricted continuation under degraded trust can increase infeasible dispatch, unreliable commitments, and settlement exposure. This paper proposes a trust-conditioned, risk-aware, and safety-critical decision framework for MEMGs exposed to hybrid cyber threats and dynamic uncertainty. Degraded operation is modeled through measurement inconsistency, communication impairment, execution mismatch, market-signal inconsistency, and uncertainty pressure, which are converted into regime-aware trust and admissibility coefficients for market, DR, execution, and settlement-sensitive domains. A bounded continuation layer then maintains market and DR functions within contracted operating envelopes. At the solution level, a large language model (LLM)-guided multi-agent deep reinforcement learning (MADRL) architecture coordinates market, DR, conversion, and flexibility agents, with the LLM restricted to bounded advisory refinement rather than autonomous control. A CVaR-based learning objective limits upper-tail settlement and service-risk exposure, while a repair-and-proximal execution layer supports implementable actions. Simulation results show that the proposed framework improves conservative-continuation success, market participation, DR fulfillment, deviation containment, settlement exposure, infeasible-action reduction, and recovery behavior compared with advanced learning-based benchmarks. Real-feeder and scalability evaluations confirm that the framework preserves market and DR continuation under degraded conditions without relying on unsafe normal operation or unnecessary suspension.
- Open Access
- 28 April 2026
- Published by Elsevier in SSRN Electronic Journal
Energy Storage System-Based Demand Response with Time-Varying Uncertainties, Cyber-Degraded Observability, and Risk-Aware Settlement in Renewable-Rich Integrated Energy Systems
Energy storage systems (ESSs) are becoming critical flexibility resources in renewable-rich integrated energy systems (IESs), where demand response (DR) must support renewable energy source (RES) variability, network feasibility, and market-oriented operation. However, ESS-supported DR faces a practical but underexplored inconsistency: the flexibility physically delivered by storage assets may differ from the delivery observed, measured, and settled by the market when uncertainty and cyber-communication disturbances occur together. RES forecast errors, load and price deviations, false data injection, denial-of-service events, packet loss, and latency can jointly distort ESS dispatch, metered DR delivery, imbalance exposure, and settlement reliability. To address this problem, this paper proposes a storage-centric, uncertainty-aware, and cyber-resilient DR framework for RES-rich IESs. The proposed model connects ESS dynamics, multi-carrier operational constraints, DR commitments, cyber-degraded observability, metered delivery, and risk-aware settlement within a unified operational chain. A large language model-guided hierarchical multi-agent deep reinforcement learning architecture coordinates distributed ESS-backed DR decisions, while a constrained action repair and projection layer converts raw control actions into feasible dispatch commands satisfying storage, network, comfort, ramping, and power-quality limits before execution. The framework is evaluated on the IEEE 69-bus system, the Caracas 141-bus feeder, and the IEEE 342-node system under nominal, uncertainty-only, attack-only, and combined uncertainty-cyber regimes. In the combined regime, the proposed method maintains 93.8–96.4% DR delivery with only 1.2–2.3% violation rates and consistently outperforms the benchmark methods in profit, constraint satisfaction, and CVaR0.95, confirming its ability to preserve feasible and economically reliable ESS-supported DR under degraded real-time conditions.
- 28 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessEnergy Storage System-Based Demand Response with Time-Varying Uncertainties, Cyber-Degraded Observability, and Risk-Aware Settlement in Renewable-Rich Integrated Energy Systems
Energy storage systems (ESSs) are becoming critical flexibility resources in renewable-rich integrated energy systems (IESs), where demand response (DR) must support renewable energy source (RES) variability, network feasibility, and market-oriented operation. However, ESS-supported DR faces a practical but underexplored inconsistency: the flexibility physically delivered by storage assets may differ from the delivery observed, measured, and settled by the market when uncertainty and cyber-communication disturbances occur together. RES forecast errors, load and price deviations, false data injection, denial-of-service events, packet loss, and latency can jointly distort ESS dispatch, metered DR delivery, imbalance exposure, and settlement reliability. To address this problem, this paper proposes a storage-centric, uncertainty-aware, and cyber-resilient DR framework for RES-rich IESs. The proposed model connects ESS dynamics, multi-carrier operational constraints, DR commitments, cyber-degraded observability, metered delivery, and risk-aware settlement within a unified operational chain. A large language model-guided hierarchical multi-agent deep reinforcement learning architecture coordinates distributed ESS-backed DR decisions, while a constrained action repair and projection layer converts raw control actions into feasible dispatch commands satisfying storage, network, comfort, ramping, and power-quality limits before execution. The framework is evaluated on the IEEE 69-bus system, the Caracas 141-bus feeder, and the IEEE 342-node system under nominal, uncertainty-only, attack-only, and combined uncertainty-cyber regimes. In the combined regime, the proposed method maintains 93.8–96.4% DR delivery with only 1.2–2.3% violation rates and consistently outperforms the benchmark methods in profit, constraint satisfaction, and CVaR0.95, confirming its ability to preserve feasible and economically reliable ESS-supported DR under degraded real-time conditions.
- Open Access
- 11 April 2026
- Published by Elsevier in SSRN Electronic Journal
Risk-Aware Multi-Agent Reinforcement Learning and Blockchain Enforcement for Peer-to-Peer Trading Market and Demand Response in Residential Energy Communities
This paper proposes a risk and uncertainty-aware energy management and market framework for residential energy communities (REC) operating in electric distribution networks under high-frequency uncertainty. The REC model captures aggregated household demand, photovoltaic generation, battery energy storage, electric vehicles with controlled charging and vehicle-to-grid capability, and flexible electrical appliances coordinated through demand response (DR). To operationalize DR while preserving privacy and accountability, we integrate a multi-agent reinforcement learning (MARL) decision layer with a blockchain-enforced peer-to-peer (P2P) markets layer. Each community agent observes local electrical states, market prices, and reputation signals, and selects dispatch and bidding actions that jointly maximize social welfare and DR delivery while explicitly penalizing tail-risk and infeasibility. Smart-contract commitments, settlement, penalties, and reputation updates provide verifiable compliance and deter strategic manipulation without exposing household-level profiles. Extensive simulations on three benchmark systems, including IEEE 33-bus, Caracas 141-bus, and IEEE 342-bus networks, demonstrate scalable performance across small, real-data-driven, and large-scale settings. Under nominal operation, the proposed method achieves welfare of 4,280/16,450/58,900 $/day with DR fulfillment of 96.1/95.0/93.6% and low downside exposure, and it consistently outperforms multi-agent proximal policy optimization (MAPPO) and centralized soft actor critic (SAC). Under hybrid cyber-physical disturbance, the framework exhibits controlled degradation, sustaining higher welfare and lower CVaR than baselines while maintaining zero hard constraint violations.
- 11 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessRisk-Aware Multi-Agent Reinforcement Learning and Blockchain Enforcement for Peer-to-Peer Trading Market and Demand Response in Residential Energy Communities
This paper proposes a risk and uncertainty-aware energy management and market framework for residential energy communities (REC) operating in electric distribution networks under high-frequency uncertainty. The REC model captures aggregated household demand, photovoltaic generation, battery energy storage, electric vehicles with controlled charging and vehicle-to-grid capability, and flexible electrical appliances coordinated through demand response (DR). To operationalize DR while preserving privacy and accountability, we integrate a multi-agent reinforcement learning (MARL) decision layer with a blockchain-enforced peer-to-peer (P2P) markets layer. Each community agent observes local electrical states, market prices, and reputation signals, and selects dispatch and bidding actions that jointly maximize social welfare and DR delivery while explicitly penalizing tail-risk and infeasibility. Smart-contract commitments, settlement, penalties, and reputation updates provide verifiable compliance and deter strategic manipulation without exposing household-level profiles. Extensive simulations on three benchmark systems, including IEEE 33-bus, Caracas 141-bus, and IEEE 342-bus networks, demonstrate scalable performance across small, real-data-driven, and large-scale settings. Under nominal operation, the proposed method achieves welfare of 4,280/16,450/58,900 $/day with DR fulfillment of 96.1/95.0/93.6% and low downside exposure, and it consistently outperforms multi-agent proximal policy optimization (MAPPO) and centralized soft actor critic (SAC). Under hybrid cyber-physical disturbance, the framework exhibits controlled degradation, sustaining higher welfare and lower CVaR than baselines while maintaining zero hard constraint violations.