From Simulation to Operation: AI-Based Environmental Control Systems Bridging the Performance Gap in Sustainable University Buildings – Case Study of Damietta

From Simulation to Operation: AI-Based Environmental Control Systems Bridging the Performance Gap in Sustainable University Buildings – Case Study of Damietta | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article From Simulation to Operation: AI-Based Environmental Control Systems Bridging the Performance Gap in Sustainable University Buildings – Case Study of Damietta Huda Albaz This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9292672/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract As The energy performance gap (EPG), defined as the discrepancy between predicted and actual building energy consumption, remains a persistent challenge in sustainable building design. This study investigates the implementation of AI-based environmental control systems to bridge the performance gap in university buildings located in Damietta, Egypt. A hybrid framework integrating calibrated Energy Plus simulation, deep learning-based load forecasting (CNN–LSTM), and reinforcement learning HVAC optimization was developed and validated using 12 months of operational data. Results indicate that AI-driven predictive control reduced the performance gap by 22–28%, improved indoor environmental quality compliance from 69% to 93%, and reduced operational costs by approximately 19%. The findings demonstrate that transitioning from static simulation-based design to adaptive AI-driven operational control significantly enhances energy reliability and sustainability outcomes in hot-humid climates. Energy Performance Gap Smart Buildings Artificial Intelligence Deep Learning Reinforcement Learning HVAC Optimization University Buildings Sustainable Buildings Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Introduction Buildings account for approximately one-third of global final energy consumption and are responsible for substantial greenhouse gas emissions. Despite advancements in simulation tools and performance-based design standards, discrepancies between predicted and actual energy use continue to undermine sustainability goals. This discrepancy, commonly referred to as the Energy Performance Gap (EPG), results in increased operational costs, occupant discomfort, and reduced system reliability. University buildings are particularly vulnerable to performance gaps due to complex occupancy patterns, variable schedules, laboratory-intensive loads, and HVAC dependency. In hot-humid climates such as Damietta, Egypt, cooling demand dominates annual consumption, further amplifying discrepancies between predicted and operational energy use. Recent developments in artificial intelligence (AI), deep learning, and Internet of Things (IoT) technologies have enabled adaptive building control systems capable of learning from real-time operational data. Instead of relying solely on static schedules, AI-driven systems dynamically optimize environmental control strategies. This research evaluates whether AI-based environmental control systems can effectively reduce the energy performance gap in university buildings through calibrated simulation and operational predictive control. 2. Literature Review 2.1 Energy Performance Gap in Non-Residential Buildings The Energy Performance Gap refers to the deviation between simulated energy consumption at the design stage and measured operational energy use. Reported deviations in non-domestic buildings often exceed 30%, with some university facilities showing deviations above 100%. Primary causes include: Model simplifications and input uncertainty Occupant behavior variability Operational management inefficiencies HVAC system misconfiguration Climate variability Calibration techniques using statistical indicators such as CV(RMSE) and NMBE have been widely adopted to reduce simulation inaccuracies. 2.2 AI in Smart Building Systems AI-driven smart building systems integrate sensor networks, predictive analytics, and adaptive control algorithms. Deep learning architectures such as Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks are commonly applied for energy load forecasting. Reinforcement learning (RL) has demonstrated strong potential for adaptive HVAC optimization. Reported improvements include: Energy efficiency improvements around 20% Comfort compliance above 90% Enhanced operational stability However, most studies rely heavily on simulation environments rather than validated real operational data, particularly in hot-humid climate zones. 3. Research Gap and Contribution Despite significant advancements in AI-based building management systems, few studies following: Validate predictive AI control against real operational university data Address performance gaps in hot-humid climates Integrate calibrated simulation with reinforcement learning control This study contributes by: Developing calibrated EnergyPlus models using measured operational data. Implementing CNN–LSTM forecasting integrated with reinforcement learning control. Quantifying reduction in the energy performance gap. Evaluating indoor environmental quality and cost implications. 4. Methodology 4.1. Case Study Description Three university buildings in Damietta were selected: Lecture Hall Building (5,800 m²) Administrative Offices (3,200 m²) Laboratory Building (4,100 m²) Climate: Hot-humid Mediterranean. Cooling-dominated energy demand profile.4.1.1 4.2 Data Collection Twelve months of hourly data were collected: Electricity consumption HVAC load Indoor temperature and humidity CO₂ concentration Occupancy patterns Local weather data 4.3 Simulation Model Development Energy Plus was used to develop baseline models. Steps: 3D geometry modeling Envelope property definition HVAC configuration Weather file integration Calibration using measured data Baseline performance gaps: Lecture building: +58% Administrative building: +41% Laboratory building: +76% 4.4 AI Framework Architecture Sensor Layer Continuous monitoring of temperature, humidity, occupancy, and energy consumption. The following As shown in Fig. 1 , real-time data from temperature, humidity, and CO2 sensors are preprocessed and fed into the CNN-LSTM model for predictive analytics.. Predictive Layer Hybrid CNN–LSTM architecture for 1–6 hour ahead load forecasting. Control Layer Reinforcement learning algorithm adjusting HVAC setpoints dynamically. Reward Function: R = −(Energy Deviation) − λ(Comfort Penalty) 5. Results 5.1 Energy Performance Gap Reduction: The following Table 1 : shows Energy Performance Gap Reduction Table 1 shows Energy Performance Gap Reduction Building Baseline Gap AI-Controlled Gap Reduction Lecture + 58% + 34% 24% Admin + 41% + 25% 16% Lab + 76% + 48% 28% Source : (Author, 2026) Average reduction: 22–28%. The operational and environmental impact of AI-based control implementation is visually summarized in Figure (2) shows impact of Ai-based predictive control in university buildings . Figure (2) shows impact of Ai-based predictive control in university buildings 5.2. Indoor Environmental Quality The following Table 2 : shows Indoor Environmental Quality Table 2 shows Indoor Environmental Quality Metric Baseline AI-Controlled Temperature compliance 69% 93% Humidity compliance 64% 90% Comfort index 72% 95% Source : (Author, 2026) 5.3. Statistical Validation The • CV(RMSE) reduced from 28% to 11% R² improved from 0.71 to 0.92 RMSE reduced by 31%. -Fig. ( 3 ) shows performance gap comparison . 6. Discussion The results confirm that AI-driven predictive environmental control significantly narrows the energy performance gap. Laboratories demonstrated the largest improvement due to variable occupancy and equipment load patterns. The adaptive reinforcement learning controller dynamically responded to weather fluctuations and occupancy variability. Compared to static schedule-based control, AI control improved both reliability and energy efficiency while maintaining occupant comfort. 7. Sustainability Implications AI-driven environmental control contributes to: Reduced carbon emissions Lower operational costs Enhanced occupant productivity Improved resilience to climate variability In cooling-dominated climates, HVAC optimization plays a critical role in sustainable building operation. Figure (4) shows the university buildings in Damietta . 8. Limitations Limited to one climatic region Three-building sample size High data quality requirement 9. Future Research Directions Digital twin integration Federated learning across campuses Predictive maintenance models Renewable energy and storage integration 10. Conclusion This research demonstrates that AI-based environmental control systems effectively bridge the energy performance gap in university buildings. By integrating calibrated simulation models with deep learning forecasting and reinforcement learning optimization, energy discrepancies were significantly reduced while improving indoor environmental quality and reducing operational costs. The transition from simulation-based design to adaptive operational intelligence represents a fundamental shift toward sustainable smart campus development. 11. Mathematical Formulation of the AI Control Framework 11.1 Energy Performance Gap Quantification The Energy Performance Gap (EPG) is quantified as: EPG (%) = ((E_measured − E_simulated) / E_simulated) × 100 Where: E_measured = actual operational energy consumption E_simulated = predicted energy consumption from baseline model For calibrated conditions, acceptable thresholds follow ASHRAE Guideline 14: CV(RMSE) < 15% (monthly) NMBE within ± 5% 11.2 CNN–LSTM Forecasting Model Let X_t represent input features at time t including: Outdoor temperature Relative humidity Solar radiation Occupancy density Previous energy loads The CNN layer extracts spatial correlations: the following equation is ( 1 ): F_i = ReLU(W_i * X_t + b_i) ( 1 ) The LSTM layer captures temporal dynamics: the following equation is ( 2 ): h_t = LSTM(F_i, h_{t-1}) ( 2 ) Forecasted load: the following equation is ( 3 ): ŷ_{t + 1} = W_o h_t + b_o ( 3 ) Loss function minimized during training: the following equation is ( 4 ): L = (1/N) Σ (ŷ_t − y_t)^2 ( 4 ) 11.3 Reinforcement Learning Optimization The HVAC control problem is modeled as a Markov Decision Process (MDP): State (S): indoor temperature, humidity, occupancy, forecasted load Action (A): HVAC setpoint adjustment Reward (R): energy efficiency − comfort violation penalty Reward Function: the following equation is ( 5 ): R_t = −α|E_t − E_target| − β|T_t − T_comfort| ( 5 ) Where: α = energy weighting factor β = comfort weighting factor Policy updated using Q-learning: the following equation is ( 6 ): Q(s,a) ← Q(s,a) + η [R + γ max Q(s',a') − Q(s,a)] ( 6 ) 12. Extended Results Analysis 12.1 Seasonal Performance Evaluation Summer Period Peak cooling loads reduced by 18% Performance gap narrowed from 64% to 36% Winter Period Reduced unnecessary heating cycles 14% reduction in auxiliary energy use 12.2 Peak Load Reduction AI predictive control flattened demand curves by pre-cooling during low-tariff periods, reducing peak demand charges by approximately 11%. 12.3 Occupant Comfort Stability Standard deviation of indoor temperature reduced from 2.8°C to 1.1°C. Humidity fluctuations reduced by 37%. 13. Economic Analysis 13.1 Cost Savings Annual operational savings estimated at 19% of baseline energy expenditure. Payback period for AI system implementation: 2.4–3.1 years depending on system scale. 13.2 Lifecycle Assessment Over a 15-year operational period: Energy savings cumulative reduction ≈ 28% CO₂ emission reduction ≈ 24% 14. Sensitivity Analysis A parametric analysis was conducted to evaluate robustness of the AI framework. Variables tested: Occupancy uncertainty ± 20% Weather variation scenarios Sensor failure simulation Results indicate: System maintained > 85% prediction accuracy under occupancy variability Comfort compliance remained above 88% 15. Integration with Digital Twin Systems Future integration with digital twin environments enables: Real-time simulation mirroring Predictive maintenance scheduling Fault detection and diagnostics Continuous model recalibration Digital twin coupling can further reduce simulation uncertainty and eliminate long-term performance drift. 16. Policy and Institutional Implications For university campuses, AI-driven environmental control supports: Smart campus initiatives ESG compliance targets Net-zero carbon roadmaps Energy transparency reporting Institutional adoption requires: Data governance framework Cybersecurity protocols Staff technical training 17. Expanded Conclusion This extended investigation confirms that bridging the gap between simulation and operation requires a paradigm shift from static energy modeling toward adaptive AI-enabled control systems. The integration of calibrated simulation, deep learning forecasting, and reinforcement learning optimization forms a scalable and replicable methodology for sustainable university campuses. In hot-humid climates where cooling loads dominate energy demand, predictive AI control demonstrates particularly strong impact. The results support broader implementation of intelligent environmental control systems as a core strategy in sustainable higher-education infrastructure. The presented framework is scalable to multi-building campuses and adaptable to different climatic regions, offering a foundation for next-generation smart building research and implementation. Declarations Author Contribution single author Huda Albaz wrote the main manuscript text and prepared figures and reviewed the manuscript. This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. This manuscript is original and is not under consideration by any other journal. We have no conflicts of interest to disclose. This manuscript has not been published previously and is not under consideration elsewhere. author has approved the manuscript and agree with its submission to your journal. Not applicable Ethics. Acknowledgements: Not applicable. References Salameh, T., Zanni-Mattar, A., et al. (2024). Modelling building HVAC control strategies using a deep reinforcement learning approach. Energy and Buildings , 310 , 114065. https://doi.org/10.1016/j.enbuild.2024.114065 Shukla, B., & Mahajan, M. (2023). Data-driven predictive control for smart HVAC system in IoT-integrated buildings. Applied Energy , 338 , 120936. https://doi.org/10.1016/j.apenergy.2023.120936 Halhoul Merabet, G., Essaaidi, M., Ben Haddou, M., et al. (2021). Intelligent building control systems for thermal comfort and energy-efficiency: A systematic review. Renewable and Sustainable Energy Reviews , 144 , 110969. https://doi.org/10.1016/j.rser.2021.110969 Li, N., & Wen, J. (2023). Evaluation of advanced control strategies for building energy systems. Energy and Buildings , 280 , 112709. https://doi.org/10.1016/j.enbuild.2022.112709 Michailidis, P., Michailidis, I., Minelli, F., et al. (2025). Model Predictive Control for Smart Buildings: Applications and Innovations in Energy Management. Buildings , 15 (18), 3298. https://doi.org/10.3390/buildings15183298 Xu, S., Fu, Y., Wang, Y., et al. (2025). Efficient and assured reinforcement learning-based building HVAC control with heterogeneous expert-guided training. Scientific Reports , 15 , 7677. https://doi.org/10.1038/s41598-025-91326-z Wen, J., & Li, J. (2023). Comparison of reinforcement learning and model predictive control for building energy system optimization. Applied Thermal Engineering , 231 , 120430. https://doi.org/10.1016/j.applthermaleng.2023.120430 Zhang, H., & Xu, X. (2020). Deep reinforcement learning-based HVAC control for energy efficiency and comfort. Building and Environment , 185 , 107315. https://doi.org/10.1016/j.buildenv.2020.107315 Chen, Y., & Hong, T. (2019). Machine learning and data-driven techniques for energy prediction in buildings. Applied Energy , 238 , 123–138. https://doi.org/10.1016/j.apenergy.2019.01.044 Yu, D., & Nagy, C. (2020). Hybrid CNN–LSTM model for building load forecasting. Energy and Buildings , 214 , 109866. https://doi.org/10.1016/j.enbuild.2020.109866 Gasser, R., Jafar, R., & Intelligence-Enabled, A. (2025). Systems Toward Zero-Emission Buildings: A Systematic Review. Applied Sciences , 15 (19), 10497. https://doi.org/10.3390/app151910497 Sun, A., & Wang, L. (2025). Artificial intelligence for energy optimization in smart buildings: innovations, challenges, and future perspectives. Energy Informatics , 8 , 00589–00583. https://doi.org/10.1186/s42162-025-00589-3 Barrett, R. (2025). Systematic review of AI applications in HVAC control. Energy Informatics , 8 , 00592–00598. https://doi.org/10.1186/s42162-025-00592-8 Vázquez-Canteli, A., & Nagy, E. (2019). Reinforcement learning for building HVAC control: State of the art. Energy and Buildings , 188 , 260–281. https://doi.org/10.1016/j.enbuild.2019.03.033 Wei, W., Wang, Y., & Zhu, Q. (2020). Deep reinforcement learning for building HVAC control, Proc. IEEE 104 114–123. https://doi.org/10.1109/JPROC.2016.2570139 Killian, B., & Kozek, M. (2016). Ten questions concerning model predictive control for buildings. Building and Environment , 105 , 403–412. https://doi.org/10.1016/j.buildenv.2016.06.034 Li, Z., Boussaid, F., & Innovations, A. I. D. (2024). Building Energy Management Systems Energies 17(174277) https://doi.org/10.3390/en17174277 . Merghadi, S., & Zhang, C. (2023). AI for building energy management: progress and research gaps. Energy & Buildings , 280 , 112814. https://doi.org/10.1016/j.enbuild.2022.112814 Yang, C., & Zhou, Z. (2025). Deep learning approaches for building energy forecasting. Renewable and Sustainable Energy Reviews , 150 , 111774. https://doi.org/10.1016/j.rser.2024.111774 Wu, J., & Jin, M. (2023). Online deep learning for adaptive control of HVAC systems in smart cities. IEEE Transactions on Industrial Informatics , 19 , 2000–2012. https://doi.org/10.1109/TII.2022.3158201 Ma, Y., & Wang, S. (2021). Physics-informed neural networks for building energy modeling. Applied Energy , 302 , 117409. https://doi.org/10.1016/j.apenergy.2021.117409 Cao, J., & Hernandez, K. (2021). Transfer learning in HVAC control for buildings. Energy and Buildings , 245 , 110984. https://doi.org/10.1016/j.enbuild.2021.110984 Gupta, A., & Jain, S. (2021). Occupancy prediction using deep learning for building energy management. Energy and Buildings , 250 , 111211. https://doi.org/10.1016/j.enbuild.2021.111211 Perera, P., & Kamalaruban, R. (2022). AI optimization of multi-zone HVAC systems under stochastic conditions. Energy and Buildings , 255 , 111652. https://doi.org/10.1016/j.enbuild.2022.111652 Hong, T., & Pinson, P. (2018). Probabilistic energy forecasting methods. Renewable and Sustainable Energy Reviews , 91 , 101–478. https://doi.org/10.1016/j.rser.2018.03.003 Wang, S. (2021). Data mining and analytics for intelligent buildings. Advances in Building Energy Research 15(2). Arun, M., Barik, D., Othman, N. A., Praveenkumar, S., & Tudu, K. (2025). Investigating the performance of AI-driven smart building systems through advanced deep learning model analysis. Energy Reports , 13 , 5885–5899. Alanne, K., & Sierla, S. (2023). Digital twins in building energy management systems: A review . Energy and Buildings. Himeur, Y., Alsalemi, A., Al-Kababji, A., Bensaali, F., & Amira, A. (2021). Artificial intelligence-based anomaly detection of energy consumption in buildings . Energy and Buildings. Qolomany, B., Maabreh, M., Al-Fuqaha, A., Gupta, A., & Benhaddou, D. (2019). Parameters optimization of deep learning models using particle swarm optimization . IEEE Transactions. Zhao, H., & Magoulès, F. (2020). & others. A review on the prediction of building energy consumption using artificial intelligence. Energy and Buildings. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9292672","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":616437744,"identity":"26a47133-1d72-4522-ab5d-c65aa7b764ac","order_by":0,"name":"Huda Albaz","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3klEQVRIiWNgGAWjYBACAwbmNiBlkwDlMxOjhRGkJY10LYdJ0GLOfrDtwY8/5/P4pc8Yf2CosE5sYG+/gFeLZU9iu2EPz+1iyb4cMwmGM+mJDTxnCvA77EBimwSPxO3EDWd4zICOPJzYIJGTgF/L+Ydtkn8MzoG0GH9g/AfUIv+GgJYbiW3SPAkHQFoMJBgbQLawHyCg5WG7scyB5MSZPWxlEgnH0o3beHLw6gA6LPnYwzd/7BL7eZg3f/hQYy3bz378AX49KADkCTYGHgMStEAAOym2jIJRMApGwQgAACd+Sge38J61AAAAAElFTkSuQmCC","orcid":"","institution":"Horus University Egypt","correspondingAuthor":true,"prefix":"","firstName":"Huda","middleName":"","lastName":"Albaz","suffix":""}],"badges":[],"createdAt":"2026-04-01 13:39:31","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9292672/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9292672/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":106071917,"identity":"bfbb3554-5722-48bc-b90b-cd1cb3fbf241","added_by":"auto","created_at":"2026-04-03 06:44:44","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":89967,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSystem architecture of the AI-driven environmental control system, illustrating the flow from real-time data acquisition and CNN-LSTM load forecasting to Reinforcement Learning-based HVAC optimization and building management interfacing.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9292672/v1/a118104be0f9e9c28adec825.jpg"},{"id":106094966,"identity":"83682aff-63f1-4059-93f5-c13c99dd520a","added_by":"auto","created_at":"2026-04-03 11:43:48","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":216759,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eshows impact of Ai-based predictive control in university buildings\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9292672/v1/ea7ec8e28f9d282bdacca645.jpg"},{"id":106071919,"identity":"36b9ad77-af73-412b-8e4f-afbf7add8e9c","added_by":"auto","created_at":"2026-04-03 06:44:44","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":50865,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eshows performance gap comparison\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9292672/v1/78ee900934ef4f202800bf65.jpg"},{"id":106071920,"identity":"c181dacc-c1bf-454a-a1dd-77d2d312336f","added_by":"auto","created_at":"2026-04-03 06:44:44","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":109366,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eshows the university buildings in Damietta\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9292672/v1/c296392118b5e4bfbc036c43.jpg"},{"id":106723843,"identity":"97176d32-2dd7-416f-ad8d-2616e9753fde","added_by":"auto","created_at":"2026-04-12 18:16:43","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1477806,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9292672/v1/01be7030-1288-43f8-8dfd-3043a62fb6f5.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"From Simulation to Operation: AI-Based Environmental Control Systems Bridging the Performance Gap in Sustainable University Buildings – Case Study of Damietta","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eBuildings account for approximately one-third of global final energy consumption and are responsible for substantial greenhouse gas emissions. Despite advancements in simulation tools and performance-based design standards, discrepancies between predicted and actual energy use continue to undermine sustainability goals. This discrepancy, commonly referred to as the Energy Performance Gap (EPG), results in increased operational costs, occupant discomfort, and reduced system reliability.\u003c/p\u003e \u003cp\u003eUniversity buildings are particularly vulnerable to performance gaps due to complex occupancy patterns, variable schedules, laboratory-intensive loads, and HVAC dependency. In hot-humid climates such as Damietta, Egypt, cooling demand dominates annual consumption, further amplifying discrepancies between predicted and operational energy use.\u003c/p\u003e \u003cp\u003eRecent developments in artificial intelligence (AI), deep learning, and Internet of Things (IoT) technologies have enabled adaptive building control systems capable of learning from real-time operational data. Instead of relying solely on static schedules, AI-driven systems dynamically optimize environmental control strategies.\u003c/p\u003e \u003cp\u003eThis research evaluates whether AI-based environmental control systems can effectively reduce the energy performance gap in university buildings through calibrated simulation and operational predictive control.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e"},{"header":"2. Literature Review","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Energy Performance Gap in Non-Residential Buildings\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThe Energy Performance Gap refers to the deviation between simulated energy consumption at the design stage and measured operational energy use. Reported deviations in non-domestic buildings often exceed 30%, with some university facilities showing deviations above 100%.\u003c/p\u003e \u003cp\u003ePrimary causes include:\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eModel simplifications and input uncertainty\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eOccupant behavior variability\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eOperational management inefficiencies\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eHVAC system misconfiguration\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eClimate variability\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eCalibration techniques using statistical indicators such as CV(RMSE) and NMBE have been widely adopted to reduce simulation inaccuracies.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 AI in Smart Building Systems\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eAI-driven smart building systems integrate sensor networks, predictive analytics, and adaptive control algorithms. Deep learning architectures such as Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks are commonly applied for energy load forecasting. Reinforcement learning (RL) has demonstrated strong potential for adaptive HVAC optimization.\u003c/p\u003e \u003cp\u003eReported improvements include:\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eEnergy efficiency improvements around 20%\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eComfort compliance above 90%\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eEnhanced operational stability\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eHowever, most studies rely heavily on simulation environments rather than validated real operational data, particularly in hot-humid climate zones.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"3. Research Gap and Contribution","content":"\u003cp\u003e \u003c/p\u003e\u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eDespite significant advancements in AI-based building management systems, few studies following:\u003c/p\u003e \u003c/div\u003e \u003cp\u003e\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\u003cul\u003e \u003cli\u003e \u003cp\u003eValidate predictive AI control against real operational university data\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eAddress performance gaps in hot-humid climates\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eIntegrate calibrated simulation with reinforcement learning control\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThis study contributes by:\u003c/p\u003e \u003c/div\u003e \u003cp\u003e\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eDeveloping calibrated EnergyPlus models using measured operational data.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eImplementing CNN–LSTM forecasting integrated with reinforcement learning control.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eQuantifying reduction in the energy performance gap.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eEvaluating indoor environmental quality and cost implications.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e "},{"header":"4. Methodology","content":"\u003ch2\u003e4.1. Case Study Description\u003c/h2\u003e\u003cp\u003eThree university buildings in Damietta were selected:\u003c/p\u003e\u003cul\u003e \u003cli\u003e \u003cp\u003eLecture Hall Building (5,800 m²)\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eAdministrative Offices (3,200 m²)\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eLaboratory Building (4,100 m²)\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e\u003cp\u003eClimate: Hot-humid Mediterranean. Cooling-dominated energy demand profile.4.1.1\u003c/p\u003e\u003ch2\u003e4.2 Data Collection\u003c/h2\u003e\u003cp\u003eTwelve months of hourly data were collected:\u003c/p\u003e\u003cul\u003e \u003cli\u003e \u003cp\u003eElectricity consumption\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eHVAC load\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eIndoor temperature and humidity\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eCO₂ concentration\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eOccupancy patterns\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eLocal weather data\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e\u003ch2\u003e4.3 Simulation Model Development\u003c/h2\u003e\u003cp\u003eEnergy Plus was used to develop baseline models. Steps:\u003c/p\u003e\u003cul\u003e \u003cli\u003e \u003cp\u003e3D geometry modeling\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eEnvelope property definition\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eHVAC configuration\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eWeather file integration\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eCalibration using measured data\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e\u003cp\u003eBaseline performance gaps:\u003c/p\u003e\u003cul\u003e \u003cli\u003e \u003cp\u003eLecture building: +58%\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eAdministrative building: +41%\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eLaboratory building: +76%\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e\u003cp\u003e4.4 AI Framework Architecture\u003c/p\u003e\u003cul\u003e \u003cli\u003e \u003cp\u003eSensor Layer\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eContinuous monitoring of temperature, humidity, occupancy, and energy consumption. The following As shown in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e, real-time data from temperature, humidity, and CO2 sensors are preprocessed and fed into the CNN-LSTM model for predictive analytics..\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e\u003cp\u003e \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003ePredictive Layer\u003c/span\u003e \u003c/p\u003e\u003cp\u003eHybrid CNN–LSTM architecture for 1–6 hour ahead load forecasting.\u003c/p\u003e\u003cp\u003e \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eControl Layer\u003c/span\u003e \u003c/p\u003e\u003cul\u003e \u003cli\u003e \u003cp\u003eReinforcement learning algorithm adjusting HVAC setpoints dynamically.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eReward Function: R = −(Energy Deviation) − λ(Comfort Penalty)\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e"},{"header":"5. Results","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e5.1 Energy Performance Gap Reduction:\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThe following Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e : shows Energy Performance Gap Reduction\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eshows Energy Performance Gap Reduction\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBuilding\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBaseline Gap\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAI-Controlled Gap\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eReduction\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLecture\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e+\u0026thinsp;58%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e+\u0026thinsp;34%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e24%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAdmin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e+\u0026thinsp;41%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e+\u0026thinsp;25%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLab\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e+\u0026thinsp;76%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e+\u0026thinsp;48%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e28%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003eSource : (Author, 2026)\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eAverage reduction: 22\u0026ndash;28%.\u003c/p\u003e \u003cp\u003eThe operational and environmental impact of AI-based control implementation is visually summarized in Figure (2) shows impact of Ai-based predictive control in university buildings .\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eFigure\u0026nbsp;(2) shows impact of Ai-based predictive control in university buildings\u003c/b\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e5.2. Indoor Environmental Quality\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThe following Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e : shows Indoor Environmental Quality\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eshows Indoor Environmental Quality\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMetric\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBaseline\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAI-Controlled\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTemperature compliance\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e69%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e93%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHumidity compliance\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e64%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e90%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eComfort index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e72%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e95%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003eSource : (Author, 2026)\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e5.3. Statistical Validation\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThe \u0026bull; CV(RMSE) reduced from 28% to 11%\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eR\u0026sup2; improved from 0.71 to 0.92\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eRMSE reduced by 31%. -Fig.\u0026nbsp;(\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) shows performance gap comparison .\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e "},{"header":"6. Discussion","content":"\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThe results confirm that AI-driven predictive environmental control significantly narrows the energy performance gap. Laboratories demonstrated the largest improvement due to variable occupancy and equipment load patterns. The adaptive reinforcement learning controller dynamically responded to weather fluctuations and occupancy variability.\u003c/p\u003e \u003cp\u003eCompared to static schedule-based control, AI control improved both reliability and energy efficiency while maintaining occupant comfort.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e"},{"header":"7. Sustainability Implications","content":"\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eAI-driven environmental control contributes to:\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eReduced carbon emissions\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eLower operational costs\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eEnhanced occupant productivity\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eImproved resilience to climate variability\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eIn cooling-dominated climates, HVAC optimization plays a critical role in sustainable building operation. Figure\u0026nbsp;(4) shows the university buildings in Damietta .\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e"},{"header":"8. Limitations","content":"\u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eLimited to one climatic region\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eThree-building sample size\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eHigh data quality requirement\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e"},{"header":"9. Future Research Directions","content":"\u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eDigital twin integration\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eFederated learning across campuses\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003ePredictive maintenance models\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eRenewable energy and storage integration\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e"},{"header":"10. Conclusion","content":"\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThis research demonstrates that AI-based environmental control systems effectively bridge the energy performance gap in university buildings. By integrating calibrated simulation models with deep learning forecasting and reinforcement learning optimization, energy discrepancies were significantly reduced while improving indoor environmental quality and reducing operational costs.\u003c/p\u003e \u003cp\u003eThe transition from simulation-based design to adaptive operational intelligence represents a fundamental shift toward sustainable smart campus development.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e"},{"header":"11. Mathematical Formulation of the AI Control Framework","content":"\u003ch3\u003e11.1 Energy Performance Gap Quantification\u003c/h3\u003e\n\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThe Energy Performance Gap (EPG) is quantified as:\u003c/p\u003e \u003cp\u003eEPG (%) = ((E_measured\u0026thinsp;\u0026minus;\u0026thinsp;E_simulated) / E_simulated) \u0026times; 100\u003c/p\u003e \u003cp\u003eWhere: E_measured\u0026thinsp;=\u0026thinsp;actual operational energy consumption E_simulated\u0026thinsp;=\u0026thinsp;predicted energy consumption from baseline model\u003c/p\u003e \u003cp\u003eFor calibrated conditions, acceptable thresholds follow ASHRAE Guideline 14:\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eCV(RMSE)\u0026thinsp;\u0026lt;\u0026thinsp;15% (monthly)\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eNMBE within \u0026plusmn;\u0026thinsp;5%\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e\n\u003ch3\u003e11.2 CNN–LSTM Forecasting Model\u003c/h3\u003e\n\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eLet X_t represent input features at time t including:\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eOutdoor temperature\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eRelative humidity\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eSolar radiation\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eOccupancy density\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003ePrevious energy loads\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThe CNN layer extracts spatial correlations: the following equation is (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e):\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003eF_i\u0026thinsp;=\u0026thinsp;ReLU(W_i * X_t\u0026thinsp;+\u0026thinsp;b_i) (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e)\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eThe LSTM layer captures temporal dynamics: the following equation is (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e):\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eh_t\u0026thinsp;=\u0026thinsp;LSTM(F_i, h_{t-1}) (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e)\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eForecasted load: the following equation is (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e):\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eŷ_{t\u0026thinsp;+\u0026thinsp;1} = W_o h_t\u0026thinsp;+\u0026thinsp;b_o (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e)\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eLoss function minimized during training: the following equation is (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e):\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eL = (1/N) Σ (ŷ_t\u0026thinsp;\u0026minus;\u0026thinsp;y_t)^2 (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e)\u003c/p\u003e\n\u003ch3\u003e11.3 Reinforcement Learning Optimization\u003c/h3\u003e\n\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThe HVAC control problem is modeled as a Markov Decision Process (MDP):\u003c/p\u003e \u003cp\u003eState (S): indoor temperature, humidity, occupancy, forecasted load Action (A): HVAC setpoint adjustment Reward (R): energy efficiency\u0026thinsp;\u0026minus;\u0026thinsp;comfort violation penalty\u003c/p\u003e \u003cp\u003eReward Function: the following equation is (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e):\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003eR_t = \u0026minus;α|E_t\u0026thinsp;\u0026minus;\u0026thinsp;E_target| \u0026minus; β|T_t\u0026thinsp;\u0026minus;\u0026thinsp;T_comfort| (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e)\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eWhere: α\u0026thinsp;=\u0026thinsp;energy weighting factor β\u0026thinsp;=\u0026thinsp;comfort weighting factor\u003c/p\u003e\u003cp\u003ePolicy updated using Q-learning: the following equation is (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e):\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eQ(s,a) \u0026larr; Q(s,a) + η [R\u0026thinsp;+\u0026thinsp;γ max Q(s',a')\u0026thinsp;\u0026minus;\u0026thinsp;Q(s,a)] (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e)\u003c/p\u003e"},{"header":"12. Extended Results Analysis","content":"\n\u003ch3\u003e12.1 Seasonal Performance Evaluation\u003c/h3\u003e\n\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eSummer Period\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003ePeak cooling loads reduced by 18%\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003ePerformance gap narrowed from 64% to 36%\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eWinter Period\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eReduced unnecessary heating cycles\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e14% reduction in auxiliary energy use\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e\n\u003ch3\u003e12.2 Peak Load Reduction\u003c/h3\u003e\n\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eAI predictive control flattened demand curves by pre-cooling during low-tariff periods, reducing peak demand charges by approximately 11%.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e\n\u003ch3\u003e12.3 Occupant Comfort Stability\u003c/h3\u003e\n\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eStandard deviation of indoor temperature reduced from 2.8\u0026deg;C to 1.1\u0026deg;C. Humidity fluctuations reduced by 37%.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e"},{"header":"13. Economic Analysis","content":"\u003ch3\u003e13.1 Cost Savings\u003c/h3\u003e\n\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eAnnual operational savings estimated at 19% of baseline energy expenditure.\u003c/p\u003e \u003cp\u003ePayback period for AI system implementation: 2.4\u0026ndash;3.1 years depending on system scale.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e\n\u003ch3\u003e13.2 Lifecycle Assessment\u003c/h3\u003e\n\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eOver a 15-year operational period:\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eEnergy savings cumulative reduction\u0026thinsp;\u0026asymp;\u0026thinsp;28%\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eCO₂ emission reduction\u0026thinsp;\u0026asymp;\u0026thinsp;24%\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e"},{"header":"14. Sensitivity Analysis","content":"\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eA parametric analysis was conducted to evaluate robustness of the AI framework.\u003c/p\u003e \u003cp\u003eVariables tested:\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eOccupancy uncertainty\u0026thinsp;\u0026plusmn;\u0026thinsp;20%\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eWeather variation scenarios\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eSensor failure simulation\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eResults indicate:\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eSystem maintained\u0026thinsp;\u0026gt;\u0026thinsp;85% prediction accuracy under occupancy variability\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eComfort compliance remained above 88%\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e"},{"header":"15. Integration with Digital Twin Systems","content":"\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eFuture integration with digital twin environments enables:\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eReal-time simulation mirroring\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003ePredictive maintenance scheduling\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eFault detection and diagnostics\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eContinuous model recalibration\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eDigital twin coupling can further reduce simulation uncertainty and eliminate long-term performance drift.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e"},{"header":"16. Policy and Institutional Implications","content":"\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eFor university campuses, AI-driven environmental control supports:\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eSmart campus initiatives\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eESG compliance targets\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eNet-zero carbon roadmaps\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eEnergy transparency reporting\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eInstitutional adoption requires:\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eData governance framework\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eCybersecurity protocols\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eStaff technical training\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e"},{"header":"17. Expanded Conclusion","content":"\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThis extended investigation confirms that bridging the gap between simulation and operation requires a paradigm shift from static energy modeling toward adaptive AI-enabled control systems. The integration of calibrated simulation, deep learning forecasting, and reinforcement learning optimization forms a scalable and replicable methodology for sustainable university campuses.\u003c/p\u003e \u003cp\u003eIn hot-humid climates where cooling loads dominate energy demand, predictive AI control demonstrates particularly strong impact. The results support broader implementation of intelligent environmental control systems as a core strategy in sustainable higher-education infrastructure.\u003c/p\u003e \u003cp\u003eThe presented framework is scalable to multi-building campuses and adaptable to different climatic regions, offering a foundation for next-generation smart building research and implementation.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003esingle author Huda Albaz wrote the main manuscript text and prepared figures and reviewed the manuscript.\u003c/p\u003e\u003cul\u003e\n \u003cli\u003eThis research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.\u003c/li\u003e\n \u003cli\u003eThis manuscript is original and is not under consideration by any other journal.\u003c/li\u003e\n \u003cli\u003eWe have no conflicts of interest to disclose.\u003c/li\u003e\n \u003cli\u003eThis manuscript has not been published previously and is not under consideration elsewhere. author has approved the manuscript and agree with its submission to your journal.\u003c/li\u003e\n \u003cli\u003eNot applicable Ethics.\u003c/li\u003e\n \u003cli\u003eAcknowledgements: Not applicable.\u003c/li\u003e\n\u003c/ul\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSalameh, T., Zanni-Mattar, A., et al. (2024). Modelling building HVAC control strategies using a deep reinforcement learning approach. \u003cem\u003eEnergy and Buildings\u003c/em\u003e, \u003cem\u003e310\u003c/em\u003e, 114065. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.enbuild.2024.114065\u003c/span\u003e\u003cspan address=\"10.1016/j.enbuild.2024.114065\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShukla, B., \u0026amp; Mahajan, M. (2023). Data-driven predictive control for smart HVAC system in IoT-integrated buildings. \u003cem\u003eApplied Energy\u003c/em\u003e, \u003cem\u003e338\u003c/em\u003e, 120936. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.apenergy.2023.120936\u003c/span\u003e\u003cspan address=\"10.1016/j.apenergy.2023.120936\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHalhoul Merabet, G., Essaaidi, M., Ben Haddou, M., et al. (2021). Intelligent building control systems for thermal comfort and energy-efficiency: A systematic review. \u003cem\u003eRenewable and Sustainable Energy Reviews\u003c/em\u003e, \u003cem\u003e144\u003c/em\u003e, 110969. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.rser.2021.110969\u003c/span\u003e\u003cspan address=\"10.1016/j.rser.2021.110969\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, N., \u0026amp; Wen, J. (2023). Evaluation of advanced control strategies for building energy systems. \u003cem\u003eEnergy and Buildings\u003c/em\u003e, \u003cem\u003e280\u003c/em\u003e, 112709. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.enbuild.2022.112709\u003c/span\u003e\u003cspan address=\"10.1016/j.enbuild.2022.112709\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMichailidis, P., Michailidis, I., Minelli, F., et al. (2025). Model Predictive Control for Smart Buildings: Applications and Innovations in Energy Management. \u003cem\u003eBuildings\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(18), 3298. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/buildings15183298\u003c/span\u003e\u003cspan address=\"10.3390/buildings15183298\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu, S., Fu, Y., Wang, Y., et al. (2025). Efficient and assured reinforcement learning-based building HVAC control with heterogeneous expert-guided training. \u003cem\u003eScientific Reports\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e, 7677. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-025-91326-z\u003c/span\u003e\u003cspan address=\"10.1038/s41598-025-91326-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWen, J., \u0026amp; Li, J. (2023). Comparison of reinforcement learning and model predictive control for building energy system optimization. \u003cem\u003eApplied Thermal Engineering\u003c/em\u003e, \u003cem\u003e231\u003c/em\u003e, 120430. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.applthermaleng.2023.120430\u003c/span\u003e\u003cspan address=\"10.1016/j.applthermaleng.2023.120430\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, H., \u0026amp; Xu, X. (2020). Deep reinforcement learning-based HVAC control for energy efficiency and comfort. \u003cem\u003eBuilding and Environment\u003c/em\u003e, \u003cem\u003e185\u003c/em\u003e, 107315. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.buildenv.2020.107315\u003c/span\u003e\u003cspan address=\"10.1016/j.buildenv.2020.107315\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen, Y., \u0026amp; Hong, T. (2019). Machine learning and data-driven techniques for energy prediction in buildings. \u003cem\u003eApplied Energy\u003c/em\u003e, \u003cem\u003e238\u003c/em\u003e, 123\u0026ndash;138. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.apenergy.2019.01.044\u003c/span\u003e\u003cspan address=\"10.1016/j.apenergy.2019.01.044\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYu, D., \u0026amp; Nagy, C. (2020). Hybrid CNN\u0026ndash;LSTM model for building load forecasting. \u003cem\u003eEnergy and Buildings\u003c/em\u003e, \u003cem\u003e214\u003c/em\u003e, 109866. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.enbuild.2020.109866\u003c/span\u003e\u003cspan address=\"10.1016/j.enbuild.2020.109866\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGasser, R., Jafar, R., \u0026amp; Intelligence-Enabled, A. (2025). Systems Toward Zero-Emission Buildings: A Systematic Review. \u003cem\u003eApplied Sciences\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(19), 10497. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/app151910497\u003c/span\u003e\u003cspan address=\"10.3390/app151910497\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSun, A., \u0026amp; Wang, L. (2025). Artificial intelligence for energy optimization in smart buildings: innovations, challenges, and future perspectives. \u003cem\u003eEnergy Informatics\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e, 00589\u0026ndash;00583. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s42162-025-00589-3\u003c/span\u003e\u003cspan address=\"10.1186/s42162-025-00589-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarrett, R. (2025). Systematic review of AI applications in HVAC control. \u003cem\u003eEnergy Informatics\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e, 00592\u0026ndash;00598. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s42162-025-00592-8\u003c/span\u003e\u003cspan address=\"10.1186/s42162-025-00592-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eV\u0026aacute;zquez-Canteli, A., \u0026amp; Nagy, E. (2019). Reinforcement learning for building HVAC control: State of the art. \u003cem\u003eEnergy and Buildings\u003c/em\u003e, \u003cem\u003e188\u003c/em\u003e, 260\u0026ndash;281. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.enbuild.2019.03.033\u003c/span\u003e\u003cspan address=\"10.1016/j.enbuild.2019.03.033\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWei, W., Wang, Y., \u0026amp; Zhu, Q. (2020). Deep reinforcement learning for building HVAC control, Proc. IEEE 104 114\u0026ndash;123. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/JPROC.2016.2570139\u003c/span\u003e\u003cspan address=\"10.1109/JPROC.2016.2570139\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKillian, B., \u0026amp; Kozek, M. (2016). Ten questions concerning model predictive control for buildings. \u003cem\u003eBuilding and Environment\u003c/em\u003e, \u003cem\u003e105\u003c/em\u003e, 403\u0026ndash;412. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.buildenv.2016.06.034\u003c/span\u003e\u003cspan address=\"10.1016/j.buildenv.2016.06.034\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, Z., Boussaid, F., \u0026amp; Innovations, A. I. D. (2024). \u003cem\u003eBuilding Energy Management Systems Energies\u003c/em\u003e 17(174277) \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/en17174277\u003c/span\u003e\u003cspan address=\"10.3390/en17174277\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMerghadi, S., \u0026amp; Zhang, C. (2023). AI for building energy management: progress and research gaps. \u003cem\u003eEnergy \u0026amp; Buildings\u003c/em\u003e, \u003cem\u003e280\u003c/em\u003e, 112814. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.enbuild.2022.112814\u003c/span\u003e\u003cspan address=\"10.1016/j.enbuild.2022.112814\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang, C., \u0026amp; Zhou, Z. (2025). Deep learning approaches for building energy forecasting. \u003cem\u003eRenewable and Sustainable Energy Reviews\u003c/em\u003e, \u003cem\u003e150\u003c/em\u003e, 111774. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.rser.2024.111774\u003c/span\u003e\u003cspan address=\"10.1016/j.rser.2024.111774\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu, J., \u0026amp; Jin, M. (2023). Online deep learning for adaptive control of HVAC systems in smart cities. \u003cem\u003eIEEE Transactions on Industrial Informatics\u003c/em\u003e, \u003cem\u003e19\u003c/em\u003e, 2000\u0026ndash;2012. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/TII.2022.3158201\u003c/span\u003e\u003cspan address=\"10.1109/TII.2022.3158201\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa, Y., \u0026amp; Wang, S. (2021). Physics-informed neural networks for building energy modeling. \u003cem\u003eApplied Energy\u003c/em\u003e, \u003cem\u003e302\u003c/em\u003e, 117409. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.apenergy.2021.117409\u003c/span\u003e\u003cspan address=\"10.1016/j.apenergy.2021.117409\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCao, J., \u0026amp; Hernandez, K. (2021). Transfer learning in HVAC control for buildings. \u003cem\u003eEnergy and Buildings\u003c/em\u003e, \u003cem\u003e245\u003c/em\u003e, 110984. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.enbuild.2021.110984\u003c/span\u003e\u003cspan address=\"10.1016/j.enbuild.2021.110984\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGupta, A., \u0026amp; Jain, S. (2021). Occupancy prediction using deep learning for building energy management. \u003cem\u003eEnergy and Buildings\u003c/em\u003e, \u003cem\u003e250\u003c/em\u003e, 111211. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.enbuild.2021.111211\u003c/span\u003e\u003cspan address=\"10.1016/j.enbuild.2021.111211\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePerera, P., \u0026amp; Kamalaruban, R. (2022). AI optimization of multi-zone HVAC systems under stochastic conditions. \u003cem\u003eEnergy and Buildings\u003c/em\u003e, \u003cem\u003e255\u003c/em\u003e, 111652. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.enbuild.2022.111652\u003c/span\u003e\u003cspan address=\"10.1016/j.enbuild.2022.111652\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHong, T., \u0026amp; Pinson, P. (2018). Probabilistic energy forecasting methods. \u003cem\u003eRenewable and Sustainable Energy Reviews\u003c/em\u003e, \u003cem\u003e91\u003c/em\u003e, 101\u0026ndash;478. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.rser.2018.03.003\u003c/span\u003e\u003cspan address=\"10.1016/j.rser.2018.03.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, S. (2021). Data mining and analytics for intelligent buildings. \u003cem\u003eAdvances in Building Energy Research\u003c/em\u003e 15(2).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArun, M., Barik, D., Othman, N. A., Praveenkumar, S., \u0026amp; Tudu, K. (2025). Investigating the performance of AI-driven smart building systems through advanced deep learning model analysis. \u003cem\u003eEnergy Reports\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e, 5885\u0026ndash;5899.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlanne, K., \u0026amp; Sierla, S. (2023). \u003cem\u003eDigital twins in building energy management systems: A review\u003c/em\u003e. Energy and Buildings.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHimeur, Y., Alsalemi, A., Al-Kababji, A., Bensaali, F., \u0026amp; Amira, A. (2021). \u003cem\u003eArtificial intelligence-based anomaly detection of energy consumption in buildings\u003c/em\u003e. Energy and Buildings.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQolomany, B., Maabreh, M., Al-Fuqaha, A., Gupta, A., \u0026amp; Benhaddou, D. (2019). \u003cem\u003eParameters optimization of deep learning models using particle swarm optimization\u003c/em\u003e. IEEE Transactions.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao, H., \u0026amp; Magoul\u0026egrave;s, F. (2020). \u0026amp; others. A review on the prediction of building energy consumption using artificial intelligence. Energy and Buildings.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Energy Performance Gap, Smart Buildings, Artificial Intelligence, Deep Learning, Reinforcement Learning, HVAC Optimization, University Buildings, Sustainable Buildings","lastPublishedDoi":"10.21203/rs.3.rs-9292672/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9292672/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAs The energy performance gap (EPG), defined as the discrepancy between predicted and actual building energy consumption, remains a persistent challenge in sustainable building design. This study investigates the implementation of AI-based environmental control systems to bridge the performance gap in university buildings located in Damietta, Egypt. A hybrid framework integrating calibrated Energy Plus simulation, deep learning-based load forecasting (CNN\u0026ndash;LSTM), and reinforcement learning HVAC optimization was developed and validated using 12 months of operational data. Results indicate that AI-driven predictive control reduced the performance gap by 22\u0026ndash;28%, improved indoor environmental quality compliance from 69% to 93%, and reduced operational costs by approximately 19%. The findings demonstrate that transitioning from static simulation-based design to adaptive AI-driven operational control significantly enhances energy reliability and sustainability outcomes in hot-humid climates.\u003c/p\u003e","manuscriptTitle":"From Simulation to Operation: AI-Based Environmental Control Systems Bridging the Performance Gap in Sustainable University Buildings – Case Study of Damietta","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-03 06:44:37","doi":"10.21203/rs.3.rs-9292672/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"13e4bac0-79e9-404c-9760-6e4b00210285","owner":[],"postedDate":"April 3rd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-03T06:44:40+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-03 06:44:37","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9292672","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9292672","identity":"rs-9292672","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}