The DOI matches its Crossref record; 69 of the 70 references it cites were found in a registry or an index; 38 of the 85 sources its claims cite were checked for support.
Explainable AI and Multi-Agent Systems for Energy Management in IoT-Edge Environments: A State of the Art Review
Open accessAbstract
This paper reviews Artificial Intelligence techniques for distributed energy management, focusing on integrating machine learning, reinforcement learning, and multi-agent systems within IoT-Edge-Cloud architectures.
Overview
CoChat Summary
TL;DR
This review synthesizes AI, multi-agent systems, and explainable AI for energy management across IoT-Edge-Cloud architectures, covering forecasting, optimization/control, communication protocols, edge hardware, and evaluation. It proposes a conceptual explainable distributed framework and identifies open challenges in RL explainability, latency/complexity trade-offs, and privacy-preserving federated learning.
Key findings
- Reinforcement learning-based control frameworks can yield energy reductions averaging around 22% relative to standard automation, with CO₂ emission cuts often estimated at up to 25% alongside that consumption decrease.
- For single-building forecasting with limited historical data, SVR achieved R² = 0.92 and outperformed Transformer models (R² = 0.82); for multi-building/urban-scale scenarios, Transformers achieved up to 23.7% accuracy improvements over ARIMA/LSTM baselines.
- Edge feasibility: local CNN inference for thermal anomaly detection can run within 100 ms; standard deep learning is advised only when the inference latency budget exceeds 100 ms, while sub-millisecond control loops favor heuristics or lightweight decision trees.
- The review proposes KPIs and a framework with examples including daily RMSE below 5% of average load, peak demand reduction exceeding 15% during tariff-sensitive intervals, median decision latency under 200 ms, and 8-bit quantization as a balance for LSTM-based federated predictors.
Methods
- Design: state-of-the-art review and conceptual synthesis of AI/MAS/XAI for distributed energy management; no primary experimental sample or dataset is reported.
- Scope: reviews statistical time series, ML regression, deep learning, heuristic/metaheuristic, mathematical optimization, MPC, RL/MARL, MAS coordination, communication protocols (MQTT, BACnet, semantic interoperability), edge AI/TinyML, federated learning, and metrics/KPIs.
- Analysis/output: comparative tables and a proposed Explainable and Distributed IoT-Edge-Cloud Framework with teacher-student co-evolution, local TinyML/SHAP explanations, and federated cloud aggregation.
Limitations
- Paper states key challenges remain in stabilizing explanations for RL policies, balancing model complexity with latency budgets, and ensuring scalable, privacy-preserving learning under non-stationary conditions.
- Paper states XAI transfer to time series is challenged by feature-independence violations and semantically meaningless raw attributions; RL explainability is complicated by stochastic exploration, partial observability, temporal credit assignment, and multi-agent interactions.
- Evident from text: as a review/conceptual paper, it does not report primary empirical validation of the proposed framework; many quantitative claims are drawn from cited studies rather than new experiments.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
29 Valid · 26 Not valid · 44 Not checked · 140 Own results and interpretations
At least one cited source doesn’t support this claim or points to a different paper, and none confirmed it — see the verdicts below.
In the paper
“When paired with fuzzy Q-learning adaptations [43], agents can navigate uncertain partner behaviors or incomplete operational data by adjusting policy confidence based on interaction outcomes.”
§ Explainable AI and Multi-Agent Systems for Energy Management in IoT-Edge Environments: A State of the Art Review > Distributed Coordination Strategies
What the sources say
- Misattributedcites [43] Xu, 2020, A Multi-Agent Reinforcement Learning-Based Data-Driven Method for Home Energy Management, IEEE Trans. Smart Grid sourceExperimental — accuracy not yet measured
Our copy of Xu et al. (2020) uses plain tabular multi-agent Q-learning (Section III.B, Eq. 8, with learning rate θ and ε-greedy exploration; Algorithm 3), never fuzzy Q-learning. The only mention of fuzzy Q-learning is in the literature review (Introduction/Abstract-adjacent paragraph), where the point is credited to another work: 'Ref. [17] proposes a residential energy management method considering peer-to-peer trading mechanism, where the model-free decision-making process is enhanced by the fuzzy Q-learning algorithm.' The paper's own adaptive behaviour is described only as the RL agent 'adapts to the learning environment and adjusts its policy via exploration and exploitation mechanism' (Section IV.D, comparison with GA) — no policy-confidence adjustment, no uncertain partner behaviours, and no incomplete-operational-data handling. So the fuzzy-Q-learning attribution belongs to Ref. [17] as our copy names it, not to this paper. (LLM, conf=0.72) It matches “Ref. [17] of Xu et al. (2020): a residential energy management method considering peer-to-peer trading mechanism, where the model-free decision-making process is enhanced by the fuzzy Q-learning algorithm” instead.
Checked by a reasoning model — may be wrong.
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Misattributed
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Consequently, Explainable AI (XAI) has transitioned from a theoretical preference to an operational necessity, essential for ensuring compliance and fostering trust in automated decision-making [8].”
§ Explainable AI and Multi-Agent Systems for Energy Management in IoT-Edge Environments: A State of the Art Review > 1. Introduction
What the sources say
- Partly supportedcites [8] Moraliyage, H., Dahanayake, S., De Silva, D., Mills, N., Rathnayaka, P., Nguyen, S., Alahakoon, D., and Jennings, A. (2022). A Robust Artificial Intelligence Approach with Explainability for Measurement and Verification of Energy Efficient Infrastructure for Net Zero Carbon Emissions. Sensors, 22. sourceExperimental — accuracy not yet measured
Section 2.4, “Explainability Layer” (page unavailable), calls interpretability a major need for understanding predictive models used in energy measurement and verification. The Conclusions (page unavailable) recommend that energy-market policymakers and regulators consider adopting explainable AI. The paper does not establish a general transition to operational necessity or say that XAI ensures compliance and fosters trust in automated decision-making. (LLM, conf=0.86)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“The introduction of multi-agent frameworks further enhances coordination among decentralized units by embedding decision-making capabilities directly into agents situated at different network layers [18].”
§ Explainable AI and Multi-Agent Systems for Energy Management in IoT-Edge Environments: A State of the Art Review > 2.1.2. Distributed IoT, Edge, Cloud Paradigms
What the sources say
- Partly supportedcites [18] Charbonnier, 2022, Scalable multi-agent reinforcement learning for distributed control of residential energy flexibility, Appl. Energy sourceExperimental — accuracy not yet measured
Our copy supports the general thrust: Section 2 states that 'RL-based indirect control strategies consider decision-making at the prosumer level, entering the realm of MARL', and the Abstract/Introduction describe cooperating agents controlling household flexibility (EVs, space heating, flexible loads) creating value at individual and system levels without a central scheduler. However, nothing in our copy places agents 'at different network layers' — the agents here are uniformly household/prosumer-level learners (Sec. 4.4.2 'MARL structures': each agent i learns its own Q-table, no information shared; Sec. 1: 'household-level decision-making'). The paper also does not simply assert enhanced coordination: it notes coordination performance of standard MARL drops at scale under partial observability and that purely local implicit strategies can have 'potential negative impacts for the grid' (Sec. 2). The multi-layer/hierarchical part of the claim is unsupported by this source. (LLM, conf=0.72)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“While Deep Learning and Reinforcement Learning (RL) offer superior performance in handling non-linear energy patterns, they often function as”
§ Explainable AI and Multi-Agent Systems for Energy Management in IoT-Edge Environments: A State of the Art Review > 1. Introduction
What the sources say
- Partly supportedcites [5] Noorchenarboo, 2025, Explaining deep learning-based anomaly detection in energy consumption data by focusing on contextually relevant data, Energy Build. sourceExperimental — accuracy not yet measured
Our copy supports the deep-learning half squarely: the Abstract (p. 3) states DL approaches 'are black-box approaches that do not provide transparency or explanations', and the Introduction (p. 5) pairs their ability to 'model complex non-linear relationships' and 'superior accuracy' with 'challenges in transparency and explainability'; Section 3.1 (p. 8) adds that DNNs are hard to interpret. Reinforcement learning is not characterized this way anywhere in our copy — RL appears only once, in Section 2.1 (p. 6), where Zhang et al. [26] used SHAP to explain deep RL models for power system emergency control, a report of another work rather than a statement that RL is an opaque black box or that it outperforms on non-linear energy patterns. So the RL attribution is unsupported by this source. (LLM, conf=0.78)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“In parallel, MAS frameworks enable distributed coordination, allowing autonomous agents to negotiate actions related to consumption, generation, and storage under uncertainty and heterogeneous objectives [4].”
§ Explainable AI and Multi-Agent Systems for Energy Management in IoT-Edge Environments: A State of the Art Review > 1. Introduction
What the sources say
- Supportedcites [4] Alhasnawi, B.N., Jasim, B.H., Esteban, M.D., and Guerrero, J.M. (2020). A Novel Smart Energy Management as a Service over a Cloud Computing Platform for Nanogrid Appliances. Sustainability, 12. source
The paper's own contribution list (Sec. 2, p. n/a) states it proposes 'a MAS for micro-grid representation that integrates IoT devices for energy management inside the buildings', with agents installed on low-performance hardware — i.e. a distributed, agent-based coordination scheme. Its problem formulation (Sec. 3.2/3.2.2) carries decision variables for device scheduling/consumption (α), self-generation with prosumer vs consumer status (β) and energy storage (γ), and notes 'community users have preferences with regard to their idle modes of operating' (heterogeneous objectives) under time-varying cost; the literature it reviews and builds on describes agents acting in a local energy market where prices are set 'based on agents' order in a decentralized way' with agents trading short-term electricity (Sec. 2). Together our copy treats exactly this topic, so the broad background claim is supported; only the word 'uncertainty' is implicit (dynamic/random pricing environment) rather than spelled out as a MAS property. (LLM, conf=0.80)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
101 more claims · 27 with their cited sources checked
Create a free account to see all 239 claims and to verify the 16 references still open.
- It may state two separate claims as one. 72
- Probably not a substantial claim. 85
- The quote only partly states this claim. 1
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Explainable AI and Multi-Agent Systems for Energy Management in IoT-Edge Environments: A State of the Art Review > Multi-Building/Urban Scale High-dimensional data; complex spatio-temporal dependencies
“Ultra-Low Power Edge Battery-operated MCUs (<10 mW); strict memory limits TinyML/Heuristics Only feasible option for inference on microcontrollers; trades precision for autonomy [33,34].”
§ Explainable AI and Multi-Agent Systems for Energy Management in IoT-Edge Environments: A State of the Art Review > 3.1.1. Statistical Time Series Approaches
“Consequently, these methods are best suited for Level 1 (Edge) deployment in microcontrollers or gateways where memory is scarce, serving as reliable fallback mechanisms when cloud connectivity is lost.”
Citations
Cites
A Perspective on Explainable Artificial Intelligence Methods: SHAP and LIME
Salih, A. M., Raisi‐Estabragh, Z., Galazzo, I. B., Radeva, P., Petersen, S. E., Lekadir, K., & Menegaz, G. (2025) · Advanced Intelligent Systems · doi:10.1002/aisy.202400304 · Cited by 897
On-Line Building Energy Optimization Using Deep Reinforcement Learning
Mocanu, E., Mocanu, D. C., Nguyen, P. H., Liotta, A., Webber, M. E., Gibescu, M., & Slootweg, J. G. (2019) · IEEE Transactions on Smart Grid · doi:10.1109/tsg.2018.2834219 · Cited by 649
A Multi-Agent Reinforcement Learning-Based Data-Driven Method for Home Energy Management
Xu, X., Jia, Y., Xu, Y., Xu, Z., Chai, S., & Lai, C. S. (2020) · IEEE Transactions on Smart Grid · doi:10.1109/tsg.2020.2971427 · Cited by 420
DeepComfort: Energy-Efficient Thermal Comfort Control in Buildings Via Reinforcement Learning
Gao, G., Li, J., & Wen, Y. (2020) · IEEE Internet of Things Journal · doi:10.1109/jiot.2020.2992117 · Cited by 283
Aguilar, J., Garces-Jimenez, A., R-Moreno, M. D., & García, R. (2021) · Renewable and Sustainable Energy Reviews · doi:10.1016/j.rser.2021.111530 · Cited by 278
Predicting Energy Consumption Using LSTM, Multi-Layer GRU and Drop-GRU Neural Networks
Mahjoub, S., Chrifi-Alaoui, L., Marhic, B., & Delahoche, L. (2022) · Sensors · doi:10.3390/s22114062 · Cited by 271
Multi-Agent Systems Applications in Energy Optimization Problems: A State-of-the-Art Review
González-Briones, A., De La Prieta, F., Mohamad, M. S., Omatu, S., & Corchado, J. M. (2018) · Energies · doi:10.3390/en11081928 · Cited by 169
Self-citationEnergy Optimization Using a Case-Based Reasoning Strategy
González‐Briones, A., Prieto, J., la Prieta, F. D., Herrera‐Viedma, E., & Corchado, J. M. (2018) · Sensors · doi:10.3390/s18030865 · Cited by 152
Self-citationReinforcement learning for HVAC control in intelligent buildings: A technical and conceptual review
Al Sayed, K., Boodi, A., Sadeghian Broujeny, R., & Beddiar, K. (2024) · Journal of Building Engineering · doi:10.1016/j.jobe.2024.110085 · Cited by 140
Multi-agent microgrid energy management based on deep learning forecaster
Afrasiabi, M., Mohammadi, M., Rastegar, M., & Kargarian, A. (2019) · Energy · doi:10.1016/j.energy.2019.115873 · Cited by 131
Cited by
Sulaima, M. F., Pradhan, A. K., Lenort, R., Frischer, R., Krejcar, O., & Namazi, H. (2026) · Energy Conversion and Management X · doi:10.1016/j.ecmx.2026.101920 · Cited by 3
Arinze, S. N., & Nwajana, A. O. (2026) · World Electric Vehicle Journal · doi:10.3390/wevj17070380 · Cited by 4
Krzywański, J., Piotrowska-Woroniak, J., Otwinowski, H., et al. (2026) · Energy · doi:10.1016/j.energy.2026.140935 · Cited by 1
Haneef, Z., Umar, M. I., Saleem, F., et al. (2026) · Information · doi:10.3390/info17090930 · Cited by 1
“Although this provides straightforward interpretation, it ignores shared information across related outputs and becomes computationally expensive for large fleets [22].”
Chen, Z., Zhen, H., Zhu, Z., Jiang, J., Kong, M., & Gu, X. (2026) · International Conference on Communications, Information System and Computer Engineering (CISCE) · doi:10.1109/cisce69494.2026.11504945 · Cited by 3
“When the SF Score monitors that the code similarity between adjacent rounds tends toward 1.0 (i.e., reasoning fixation occurs), the system automatically triggers a dynamic routing protocol to introduce a high-order Mediator [5] node for non-linear intervention.”
Influential
Metadata
- Authors
- Álvarez-López, C., González-Briones, A., & Li, T.
- Affiliations
- BISITE Research Group, University of Salamanca, 37007 Salamanca, Spain
- Northwestern Polytechnical University (CN)
- Venue
- Electronics
- Publisher
- MDPI AG
- Published
- 2026-01-15
- Volume
- 15
- Issue
- 2
- Pages
- 385
- ISSN
- 2079-9292 (ISSN-L)
- OpenAlex
- W7124310453
- Cited by
- 23
- References
- 65
- Citations by year
- 2026: 23
- FWCI
- 36.64
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Electronics (publishedVersion, cc-by)
- Funders
- Secretary of State for Digitalization and Artificial Intelligence and by the European Union (TSI-100933-2023-0001)
- Topics
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Smart Grid Energy Management
- Physical Sciences › Computer Science › Computer Networks and Communications › IoT and Edge/Fog Computing
- Physical Sciences › Computer Science › Information Systems › Internet of Things and AI
- Keywords
- explainable AI, multi-agent reinforcement learning, distributed energy management, energy prediction, federated learning, model predictive control, metaheuristic optimization, edge computing, low-power hardware design, privacy-preserving machine learning, interpretability, communication protocols, benchmarking, smart energy systems
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Explainable AI and Multi-Agent Systems for Energy Management in IoT-Edge Environments: A State of the Art Review. https://app.cochat.ai/papers/10.3390/electronics15020385
- The paper
- Álvarez-López, C., González-Briones, A., & Li, T. (2026). Explainable AI and Multi-Agent Systems for Energy Management in IoT-Edge Environments: A State of the Art Review. Electronics. https://doi.org/10.3390/electronics15020385
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.