Source-linked AI summary
Intelligent Building Control Systems for Thermal Comfort and Energy-Efficiency: A Systematic Review of Artificial Intelligence-Assisted Techniques
Ghezlane Halhoul Merabet, Mohamed Essaaidi, Mohamed Ben Haddou, Basheer Qolomany, Junaid Qadir, Muhammad Anan, Ala Al-Fuqaha, Mohamed Riduan Abid, Driss Benhaddou
TL;DR
Building control must reconcile energy consumption with occupants’ thermal comfort, a challenge complicated by differences between standardized predictions and real-world experience. This paper systematically reviews AI-assisted building-control techniques, their implementations, and reported performance. The review finds broad use and promise, but concludes that AI-based control remains not completely satisfactory, partly because high-quality real-world data are scarce.
Problem
Building control must balance energy consumption with thermal comfort, while standardized comfort predictions can differ from real-world occupant experience.
Method
The paper conducts a systematic review of AI techniques for building control, examining comfort assessment, control applications, implementations, inputs, outputs, and performance.
Results
Multiple AI-based techniques were used across building-control functions, including optimization, predictive control, and comfort or energy management.
Takeaways & Limitations
AI-assisted building control is a promising and ongoing research area for improving energy efficiency while maintaining thermal comfort.
Takeaways & Limitations
AI-based control commonly requires large quantities of high-quality real-world data, which are lacking in the building and energy sectors.
Abstract
from arXiv · showhide
Building operations represent a significant percentage of the total primary energy consumed in most countries due to the proliferation of Heating, Ventilation and Air-Conditioning (HVAC) installations in response to the growing demand for improved thermal comfort. Reducing the associated energy consumption while maintaining comfortable conditions in buildings are conflicting objectives and represent a typical optimization problem that requires intelligent system design. Over the last decade, different methodologies based on the Artificial Intelligence (AI) techniques have been deployed to find the sweet spot between energy use in HVAC systems and suitable indoor comfort levels to the occupants. This paper performs a comprehensive and an in-depth systematic review of AI-based techniques used for building control systems by assessing the outputs of these techniques, and their implementations in the reviewed works, as well as investigating their abilities to improve the energy-efficiency, while maintaining thermal comfort conditions. This enables a holistic view of (1) the complexities of delivering thermal comfort to users inside buildings in an energy-efficient way, and (2) the associated bibliographic material to assist researchers and experts in the field in tackling such a challenge. Among the 20 AI tools developed for both energy consumption and comfort control, functions such as identification and recognition patterns, optimization, predictive control. Based on the findings of this work, the application of AI technology in building control is a promising area of research and still an ongoing, i.e., the performance of AI-based control is not yet completely satisfactory. This is mainly due in part to the fact that these algorithms usually need a large amount of high-quality real-world data, which is lacking in the building or, more precisely, the energy sector.
1. Introduction
Building control must balance thermal comfort with energy conservation as HVAC use and climate pressures increase. The review motivates AI-assisted control because standard comfort models may diverge from occupants’ real experiences.
- Energy and comfort challenge: HVAC growth and warmer periods increase concerns about peak electricity demand and building energy use.The text links widespread air-conditioning, climate change, and rising peak-demand concerns.
- Energy and comfort challenge: Thermal comfort and energy conservation are the two main objectives of intelligent building control.The paper frames their joint management as the central building-control problem.
- Thermal comfort assessment: PMV assesses average thermal sensation on a standardized scale, with ISO 7730 recommending PMV 0 ± 0.5 for best comfort.This analytical approach represents comfort through physical and physiological variables, including heat exchange between the body and environment.
- Thermal comfort assessment: In-situ studies report discrepancies between standardized comfort predictions and occupants’ subjective behavior.These findings motivate adaptive approaches that account for occupants’ reactions to their thermal environment.
- AI-assisted control: AI-based building control research includes evolutionary algorithms, fuzzy logic, neural networks, and adaptive control for comfort and energy management.Reported applications include neural-network controllers deployed in Japanese air-conditioning and fan installations.
- AI-assisted control: The review addresses a gap by bringing together AI-based methods for thermal comfort and energy control while including individual occupant interactions.Its scope covers methods, system components, inputs, outputs, and prior survey literature.
REF. OF THE WORK
Previous surveys addressed intelligent building control, optimization, HVAC methods, and comfort prediction from different perspectives. This work distinguishes itself by combining AI-based energy-comfort control with occupant interaction and system-level analysis.
- Prior survey limitations: Several prior works did not focus specifically on AI methods for building control or omitted thermal comfort, human factors, or occupant preferences.These exclusions limited their coverage of integrated energy-comfort control.
- Prior survey coverage: Earlier reviews examined agent-based control, computational optimization, HVAC control, and intelligent control for energy or comfort management.Their scopes varied across sustainable-building design, fuzzy cognitive maps, and smart-building services.
- Prior survey limitations: Other reviews focused on machine learning, big-data analytics, residential buildings, or comfort-prediction models rather than integrated AI energy-comfort control.The reviewed distinctions concern both application scope and classification of AI techniques.
- This review’s contribution: The paper presents itself as the first review to cover personal occupant interaction in AI-assisted energy-efficient thermal-comfort optimization.Its stated contributions include investigating AI/ML tools, identifying challenges, and mapping control-system components and functionalities.
- This review’s contribution: The review analyzes AI-based energy control, thermal-comfort assessment, case studies, model inputs and outputs, and reported performance.It also synthesizes key insights, research directions, and related challenges.
2. Systematic Review Methodology
The study follows a PRISMA-guided systematic review process covering AI-assisted thermal-comfort and HVAC control research. It searches multiple scholarly databases, filters studies using explicit criteria, and organizes the included evidence by technique and application.
- Search strategy: The search used terms related to thermal comfort, artificial intelligence, big data, occupants, and HVAC systems across five scholarly databases and manual sources.The databases included ACM Digital Library, Scopus, Google Scholar, IEEE Xplore, Web of Science, and ScienceDirect.
- Search strategy: The review covered scientific publications from 1992 to 2020 and followed PRISMA guidelines for systematic reviews and meta-analyses.The protocol was represented through an article-selection flowchart.
- Study selection: 1,198 articles were initially gathered, and 125 studies satisfied the inclusion criteria after filtering.Eligible studies addressed indoor conditions, AI-based HVAC or thermal-comfort control, and reported system performance.
- Evidence organization: The included studies were arranged chronologically to analyze innovative AI techniques and their usability for building energy savings and thermal comfort.The review considered occupant preferences, comfort inference, control devices, building components, and occupant interactions.
- Evidence organization: The review summarizes study cases, deployed AI techniques, comfort-inference methods, inputs, controlled parameters, and reported comfort and energy-savings performance.These outputs are compiled in Tables 3 to 9.
3. AI Development for Thermal Comfort and Energy-Efficiency in Buildings
The review identifies approximately 20 AI techniques for jointly managing building thermal comfort and energy consumption, including pattern recognition, optimization, and predictive control. Reported studies generally improve energy performance while maintaining or improving comfort, although outcomes vary across methods and settings.
- AI control techniques: Approximately 20 AI techniques were identified for controlling both thermal comfort and energy consumption in buildings.The reviewed techniques include functions such as pattern identification and recognition, optimization, and predictive control.
- Implementations: The reviewed implementations span IoT-based HVAC control, deep-learning thermal-load optimization, hybrid data-driven VAV prediction, and machine-learning thermal-environment control.The review also reports frameworks for smart thermostats, heating set-point scheduling, ACMV optimization, and network-based building control.
- Fuzzy and hybrid control: Fuzzy-logic and hybrid approaches achieved energy savings while maintaining or improving occupants’ thermal comfort across several building-control studies.Reported savings included 18%–40%, 25%, 18.9%, and up to 38%, depending on the method and comparison baseline.
- Neural-network and learning-based control: Neural-network and learning-based controllers reduced energy use while satisfying comfort conditions or maintaining temperature set-points.Examples include approximately 27.12% lower energy consumption, approximately 20% lower heating energy, and 25% average energy savings under PMV constraints.
- Optimization and predictive control: Optimization and predictive strategies also improved comfort or reduced energy and operating costs relative to conventional, rule-based, or baseline controllers.Reported outcomes included up to 45% more energy savings, 44.3% better comfort performance than the PMV model, and up to 12.8% cost savings versus a rule-based strategy.
- Prediction performance: Several AI models improved prediction or control quality, including 89.2% average personal-comfort prediction accuracy and 88% occupancy-estimation accuracy.Other reported results included load-prediction errors around 7.9% and 7.2%, and a type-2 fuzzy model RMSE of 12.55 versus 17.64 for linear regression.
FUZZY LOGIC
The reviewed AI control techniques include reinforcement learning, predictive control, and related approaches, each offering capabilities alongside data, computation, modeling, or implementation constraints.
- Reinforcement Learning: Reinforcement learning and deep reinforcement learning address complex control problems, support long-term results, and can learn behavior that maximizes performance.These methods use sensory data for energy and comfort optimization in HVAC systems, heat pumps, and water heaters.
- Reinforcement Learning: Reinforcement learning requires substantial data and computation, while state-value storage can make memory use expensive.The reviewed limitations also note that building-control states may be difficult to determine completely.
- Predictive Control: Predictive control faces online-computation barriers, centralized complexity in large buildings, and costs associated with obtaining and implementing process models.Difficulty obtaining a mathematical model has long penalized predictive control in buildings.
- Predictive Control: Predictive control can handle disturbances, multivariable systems, future changes, and constraints during command synthesis.Model predictive control is described as preferable for large processes composed of several subsystems.
ADVANCED MPC
Advanced model predictive control is presented as useful for energy and comfort management, while hybrid regulators combine complementary regulators but require substantial setup and training effort.
- Advanced MPC: MPC can save building energy, reduce peak electricity demand, account for future climate changes, and outperform PID or FLC alone in heating and cooling.MPC energy-performance prediction commonly uses simplified resistance-capacity models.
- Advanced MPC: Hybrid-regulator design requires user experience and substantial training data, while configuring the conventional or advanced component is difficult.The passage identifies both the intelligent and non-intelligent portions as setup challenges.
- Advanced MPC: Hybrid regulators can solve problems that individual regulators cannot solve separately.The benefit is attributed to combining regulators rather than relying on one regulator alone.
HYBRID METHODS
The review describes hybrid and optimization-oriented AI control methods, emphasizing combined models, evolutionary optimization, swarm optimization, and co-simulation for building energy and comfort management.
- Hybrid Methods: Combined models perform better than single models in reducing energy consumption, maintaining comfort conditions, and incorporating user preferences.The review reports significant results for combined approaches across these objectives.
- Co-simulation: Co-simulators provide real-time parameters for AI controllers by combining a global system view with the controller’s local subsystem view.This complementary arrangement supports optimization of building energy and thermal comfort.
- Optimization Methods: The review focuses on advanced optimization techniques adopted in AI-assisted building control, particularly genetic algorithms and particle swarm optimization.These techniques are presented as central optimization methods rather than a complete review of all numerical optimization methods.
- Genetic Algorithms: Genetic algorithms evolve populations through selection and recombination, evaluating individuals as candidate solutions to generate improved solutions.A classical genetic-algorithm flowchart is provided in Figure 7.
- Optimization Methods: Optimization methods in the review are summarized by their characteristics in Table 11.The review notes renewed interest in numerical optimization methods since the 2000s.
4. Theoretical Analysis of the AI Applied for Building Control
The review examines AI-assisted building control through its energy–comfort objectives, model inputs and outputs, control functions, and thermal-comfort assessment approaches.
- Objectives and Systems: AI-assisted building control targets energy efficiency and indoor comfort while accounting for user preferences and lower-level variables such as temperature, humidity, and air speed.The reviewed systems include intelligent Building Energy Management Systems for tracking building microclimates and reducing energy use and operating costs.
- Control Architecture: Neural networks and fuzzy logic form the core of AI-assisted building control, with AI tools also applied to sensor-based control.Multiple sensors collect environmental and personal variables, store them in databases, and support intelligent decisions.
- Control Functions: Optimized setting and predictive control are the most adopted functions, with optimized setting mainly implemented through genetic algorithms or particle swarm optimization.Reviewed methods commonly operate heating and cooling loads while respecting authorized power rates and thermal comfort.
- Thermal Comfort Assessment: The review distinguishes general and individual comfort models, noting that individual models can provide individualized treatment and better satisfaction.The conventional PMV-PPD model is identified as a general comfort model and is based on climatic-chamber studies involving 1,300 subjects.
5. Trend Analysis and Discussions
The review finds broad use of AI/ML techniques for building control, with neural networks, fuzzy logic, optimization, predictive control, and hybrid approaches supporting energy and comfort objectives. Reported results include substantial energy savings, comfort improvements, and cost reductions, although outcomes vary across methods and applications.
- AI/ML technique trends: Neural networks are the most popular AI approach among the reviewed building-control studies.
- Energy savings and predictive control: AI/ML techniques achieved up to 31% in the reviewed energy-saving results, while Bayesian-network models were used to estimate acceptable thermal conditions.
- Energy savings and predictive control: ~50% energy reduction was reported for learning-based model predictive control on an HVAC testbed, while RBF-network predictive control saved more than 50% with good thermal-sensation coverage.
- Thermal comfort: Average comfort improvement was around 50%, with maximum comfort reaching 100% through neural networks, DAI, MAS, and genetic algorithms.
- Cost outcomes: AI/ML methods reduced average costs by up to 34%, while the maximum reported energy-saving cost reduction was 58%.
6. Open Challenges and Future Research Directions
The review identifies persistent challenges in delivering personalized thermal comfort and energy efficiency because real-world conditions, occupants, connected systems, and security requirements introduce substantial complexity. Future directions include dynamic, human-centered, data-rich, and networked building-control approaches.
- Research scope: Providing thermal comfort while improving building energy productivity remains an open research problem with numerous research challenges.
- Connected buildings: Connected buildings can support smart-grid demand response, net-zero communities, and improved energy efficiency and resilience through community microgrids and district energy.
- Context-sensitive comfort: Real-world comfort conditions vary with climate and building management, and naturally ventilated buildings exhibit broader comfort ranges than PMV predicts.
- Context-sensitive comfort: PMV and rational chamber-based models can fail to predict in-situ comfort because occupant–environment interactions and psychological or sociological factors affect thermal experience.
- Personalization and human factors: Existing thermal comfort standards do not distinguish variability between occupants or dynamically track relevant parameters for changing building-control settings.
- Data and validation: Future work calls for genuine, heterogeneous sustainable-building data and further evaluation of promising techniques such as EACRA in smart-building environments.
7. Conclusions
The review examines AI techniques for building control systems that address energy efficiency and thermal comfort, assessing their implementation and outputs. It finds an active but still ongoing research area, with data availability, privacy, and smart-building integration remaining important challenges.
- Review scope: The paper presents a comprehensive review of AI techniques used in building control systems.The review assesses implementations and outputs reported in published works.
- Evidence base: Reviewed studies primarily used empirical case studies based on occupant feedback, measurements, and publicly accessible datasets.Data on thermal comfort and energy usage were collected through questionnaires, interviews, measurements, or existing datasets.
- AI techniques: Artificial neural networks supported data recognition, classification, thermal-comfort description, and Predicted Mean Vote estimation.Fuzzy Logic treated thermal comfort as a subjective or fuzzy parameter and commonly used the PMV comfort index.
- AI techniques: Fuzzy Logic approaches targeted high occupant satisfaction and optimum energy efficiency through simple, scalable regulation without requiring a system model.Their efficiency was generally compared with traditional controls, while other approaches incorporated learning or optimization assumptions.
- Challenges and future directions: AI and machine-learning applications in building control remain an ongoing research endeavor because they typically require massive quantities of high-quality real-world data that the energy sector has lacked.Smart meters, IoT, and cloud storage are increasing data quantity and sophistication, while future work must also address privacy, security, context awareness, and human-in-the-loop comfort modeling.