Thursday, November 19, 2026
Science
No Result
View All Result
  • Login
  • HOME
  • SCIENCE NEWS
  • CONTACT US
  • HOME
  • SCIENCE NEWS
  • CONTACT US
No Result
View All Result
Scienmag
No Result
View All Result
Home Science News

Machine Learning Model Predicts Absorption Chiller Output to Optimize Thermal Performance

November 19, 2026
in Science News
Teresa Odom
By Teresa Odom Scienmag Editorial Profile - Machine Learning
Reading Time: 6 mins read
0
Machine Learning Model Predicts Absorption Chiller Output to Optimize Thermal Performance

Machine Learning Model Predicts Absorption Chiller Output to Optimize Thermal Performance

65
SHARES
587
VIEWS
Share on FacebookShare on Twitter
ADVERTISEMENT

A hybrid modeling framework combining finite-time thermodynamics, linear programming, and supervised machine learning can predict the maximum cooling capacity and coefficient of performance of single-effect absorption refrigeration systems, according to research published in the International Journal of Air-Conditioning and Refrigeration. Developed by an engineering team including Abdel-Farid Mamadou Idrissou, the approach addresses the long-standing computational challenge of dynamically optimizing heat exchanger conductances when operating conditions vary. Absorption refrigeration systems driven by waste heat or thermal storage offer an alternative to conventional vapor compression chillers, yet identifying optimal operating states across multiple heat exchangers has historically been hindered by the mathematical limitations of geometric variational methods. By integrating numerical linear programming with machine learning regression, the investigators established a procedure to maintain optimal thermodynamic control in static systems under changing thermal demands.

Classical thermodynamics, rooted in the foundational theoretical principles formulated by Sadi Carnot in 1824, establishes the theoretical upper limit of heat engine efficiency operating between thermal reservoirs. However, Carnot efficiency assumes reversible processes occurring infinitely slowly, meaning that heat and work transfer rates approach zero. Real-world thermal engineering systems operate under finite time and finite size constraints, where temperature gradients generate unavoidable thermal irreversibilities. To bridge this gap, finite-time thermodynamics, developed in 1975 by R. S. Berry, P. Salamon, and B. Andresen, introduces finite cycle durations and physical equipment boundaries into thermodynamic models. While finite-time thermodynamics provides realistic approximations of irreversible losses, classical parametric optimization methods rely on geometric variational techniques that are mathematically constrained to three-dimensional operating parameter spaces, making simultaneous optimization across four separate heat exchangers functionally intractable.

To overcome the dimensional boundary of classical three-variable parametric studies, the researchers formulated an optimal thermal conductance allocation problem in four-dimensional space. The investigated plant consists of a single-effect heating, ventilation, and air conditioning absorption refrigeration cycle coupled to a latent heat thermal energy storage battery. The thermal storage subsystem uses a eutectic mixture of lithium nitrate and sodium nitrate with a melting temperature of 194 degrees Celsius, selected for its thermal stability and congruent melting properties. This phase change material is charged via flue gas streams from low-grade waste heat and discharged to supply thermal power to the refrigeration generator. In baseline testing, the phase change material battery provided approximately 40 minutes of operational autonomy at a discharge rate of 61.41 kilowatts, delivering thermal energy while being maintained at isothermal conditions to stabilize the generator against thermal fluctuations.

The refrigeration subsystem operates on a binary ammonia-water mixture across four primary heat exchangers: the generator, condenser, evaporator, and absorber. Within this thermodynamic cycle, the greatest energy transfers and entropy generation occur across these four components, whereas the electrical consumption of the mechanical solution pump is negligible in comparison. The authors simplified the physical system into an equivalent endoreversible model where internal cycle irreversibilities such as fluid friction and internal throttling are assumed negligible, while external heat exchanges across system boundaries remain irreversible. Heat transfers across each component boundary were modeled using Newton’s cooling law, relating heat transfer rates to temperature differentials across the respective thermal conductances. Applying an energy balance and Clausius’s inequality established that the endoreversible coefficient of performance remains bounded by the product of the Carnot heat engine efficiency and Carnot refrigeration performance.

The team translated the allocation problem into a linear programming formulation, defining the objective function as the dimensionless refrigeration heat load of the evaporator subject to four governing constraints. The decision variables comprise the four dimensionless thermal conductances of the condenser, generator, evaporator, and absorber, with their sum constrained to unity. The first constraint limits the cycle performance below the theoretical Carnot boundary, the second fixes the total conductance sum to unity, the third enforces the energy conservation balance across all four exchangers, and the fourth confines each individual conductance between zero and one. To solve this constrained optimization problem, the researchers implemented the numerical interior-point method, which utilizes positive slack variables and logarithmic barrier functions to convert inequalities into equality constraints, avoiding the higher computational overhead and resolution step limitations associated with Dantzig’s traditional simplex algorithm.

To confirm that the linear programming formulation generated physically reliable data, the investigators conducted validation against previously published models and alternative numerical solvers. First, they benchmarked their method against the endoreversible model of Vargas and colleagues, which had simplified the cycle by combining the condenser and absorber into a single effective exchanger in three-dimensional space. When adapted to three conductances, the interior-point linear programming solver converged on an optimal dimensionless cooling load of 0.01597 compared to 0.01500 obtained via classical geometric parametric study, yielding an absolute difference of 0.00097 while identifying identical optimal decision variables. Additionally, comparing the interior-point method directly against the simplex algorithm on the same system yielded identical variable distributions and a negligible objective function variance of 0.00003, verifying that the linear formulation reliably characterizes optimal refrigeration configurations.

Applying the four-dimensional linear programming method to the single-effect plant demonstrated measurable performance gains over baseline operations. At a dimensionless fuel flow rate of 0.09, representing the ratio of heat transfer fluid heat capacity to total system conductance, the reference unoptimized configuration exhibited a dimensionless cooling load of 0.0131 and a coefficient of performance of 0.4200. Following numerical optimization via nine interior-point iterations, the objective cooling load increased to 0.0177 in the feasible solution space, achieving a coefficient of performance of 0.5931. This shift represents a 28.81 percent improvement in cycle efficiency compared to the unoptimized baseline. Furthermore, parametric analysis of optimal conductance distributions revealed that at maximum cooling capacity, the thermal conductances of the condenser and absorber are precisely equal, establishing a design rule for balanced heat rejection in single-effect absorption chillers.

Using the validated optimization routine, the researchers generated a database of 100 optimal operational states across varying fuel flow rates, total conductances, condenser temperatures, and storage masses to train supervised machine learning models. The dataset was split into an 80 percent training set and a 20 percent unseen test set to evaluate five regression algorithms: Random Forest Regressor, Gradient Boosting Regressor, Decision Trees Regressor, K-Nearest Neighbors Regressor, and Tweedie Regressor. Model evaluations examined two distinct input scenarios. In Situation 1, models were trained using only the four thermal conductances as predictors. In Situation 2, the dimensionless generator fuel flow rate was added as an explanatory feature alongside the conductances, while both configurations retained the coefficient of performance and maximum cooling load as target labels evaluated through coefficient of determination, mean absolute error, and root mean square error metrics.

Evaluation results demonstrated that incorporating the fuel flow rate in Situation 2 markedly improved prediction precision across all regression algorithms. Linear Tweedie regression proved ineffective, yielding coefficients of determination below 10 percent in both situations, while nonparametric K-Nearest Neighbors achieved scores of only 20 to 55 percent. When predicting cooling load in Situation 1 without the fuel flow rate, high data dispersion restricted the coefficient of determination to less than 60 percent across all models, with Random Forest achieving 58.53 percent. In Situation 2, the inclusion of fuel flow rate elevated cooling load prediction accuracy to 74.35 percent for Random Forest, 74.07 percent for Decision Trees, and 85.04 percent for Gradient Boosting. For coefficient of performance prediction, Gradient Boosting demonstrated strong predictive capacity, achieving a coefficient of determination of 96.34 percent in Situation 2 alongside the lowest error distances from zero.

To evaluate practical applicability, the authors validated the trained Gradient Boosting model against independent experimental data reported by Evola and colleagues for a solar-driven single-effect absorption chiller. Using recorded fluid mass flow rates, fluid properties, and conductance ratios, the machine learning model calculated predicted maximum cooling loads and performance coefficients, which were then coupled into transient energy balance equations to simulate water-side temperatures. The model predicted real-time evaporator and generator operating temperatures that aligned closely with experimental measurements, yielding a mean square error of 1.41 degrees Celsius and root mean square error of 1.18 degrees Celsius for evaporator temperatures, alongside corresponding errors of 0.05 degrees Celsius and 0.23 degrees Celsius for generator temperatures.

The authors note specific engineering boundaries and scope constraints associated with the current findings. The derived performance coefficients and optimal conductance distributions apply strictly to single-effect absorption systems operating within the typical activation temperature window of 70 to 120 degrees Celsius. While physical heat exchanger surface areas cannot be altered once equipment is manufactured, plant operators can adjust heat transfer coefficients during post-design operation by modulating secondary fluid flow rates to match predicted conductance targets. The authors emphasize that while their computational and experimental comparisons demonstrate the technical viability of static optimal control via machine learning, a comprehensive techno-economic feasibility study evaluating manufacturing capital against operational savings has not yet been conducted.

Subject of Research: Machine learning and linear programming optimization of endoreversible single-effect absorption refrigeration systems

Article Title: Intelligent optimal control of endoreversible single-effect HVAC-AR system using machine learning

Article References: Mamadou, A. F. I., Idrissou, M. O. K., Sanya, S. A. O., Vargas, J. V. C., & Alexandra, A. R. (2025). Intelligent optimal control of endoreversible single-effect HVAC-AR system using machine learning. International Journal of Air-Conditioning and Refrigeration, 33(1), Article 20. https://doi.org/10.1007/s44189-025-00088-w

Image Credits: AI Generated

DOI: 10.1007/s44189-025-00088-w

Keywords: absorption refrigeration, finite-time thermodynamics, linear programming, gradient boosting regressor, coefficient of performance, latent heat thermal energy storage, optimal thermal conductance allocation, machine learning, cooling capacity, phase change material, interior point method, heat exchangers, Carnot efficiency, thermodynamic optimization, waste heat recovery, HVAC systems

Cite Scienmag News

Teresa Odom. (November 19, 2026). Machine Learning Model Predicts Absorption Chiller Output to Optimize Thermal Performance. Scienmag. https://scienmag.com/machine-learning-model-predicts-absorption-chiller-output-to-optimize-thermal-performance/

Teresa Odom. "Machine Learning Model Predicts Absorption Chiller Output to Optimize Thermal Performance." Scienmag, 19 November 2026, https://scienmag.com/machine-learning-model-predicts-absorption-chiller-output-to-optimize-thermal-performance/. Accessed 19 November 2026.

Teresa Odom. "Machine Learning Model Predicts Absorption Chiller Output to Optimize Thermal Performance." Scienmag. November 19, 2026. https://scienmag.com/machine-learning-model-predicts-absorption-chiller-output-to-optimize-thermal-performance/

Tags: absorption refrigerationCarnot efficiencycoefficient of performancecooling capacityfinite-time thermodynamicsgradient boosting regressorheat exchangersHVAC systemsinterior point methodlatent heat thermal energy storagelinear programmingMachine learningoptimal thermal conductance allocationphase-change materialthermodynamic optimizationwaste heat recovery
Share26Tweet16
Previous Post

Machine Learning Model Detects Smart Home Cyber Threats Using Behavioral Deviation Analysis

Next Post

Neural Network Architecture Accurately Classifies and Captions Plots in Educational Videos

Related Posts

Causal Graph Neural Network Predicts Financial Systemic Risk and Contagion Trajectories
Technology and Engineering

Causal Graph Neural Network Predicts Financial Systemic Risk and Contagion Trajectories

November 19, 2026
High Relative Hospital Volume Linked to Reduced Colon Cancer Recurrence Risk
Medicine

High Relative Hospital Volume Linked to Reduced Colon Cancer Recurrence Risk

November 19, 2026
Guatemala Volcán de Fuego Evacuations Lag Behind Escalating Climax Timescales
Science News

Guatemala Volcán de Fuego Evacuations Lag Behind Escalating Climax Timescales

November 19, 2026
Model Reveals Design Rules for Microplates on Soft Viscoelastic Micropillar Foundations
Science News

Model Reveals Design Rules for Microplates on Soft Viscoelastic Micropillar Foundations

November 19, 2026
Alzheimer’s Review Links Lipid Droplets and Metabolism Disruption Across Brain Cells
Biology

Alzheimer’s Review Links Lipid Droplets and Metabolism Disruption Across Brain Cells

November 19, 2026
Hybrid Neural-Tree Learning Framework Improves Risk Detection Across Commercial Banking Benchmarks
Technology and Engineering

Hybrid Neural-Tree Learning Framework Improves Risk Detection Across Commercial Banking Benchmarks

November 19, 2026
Next Post
Neural Network Architecture Accurately Classifies and Captions Plots in Educational Videos

Neural Network Architecture Accurately Classifies and Captions Plots in Educational Videos

  • Mothers who receive childcare support from maternal grandparents show more optimized

    Mothers who receive childcare support from maternal grandparents show more parental warmth, finds NTU Singapore study

    27656 shares
    Share 11059 Tweet 6912
  • University of Seville Breaks 120-Year-Old Mystery, Revises a Key Einstein Concept

    1061 shares
    Share 424 Tweet 265
  • Bee body mass, pathogens and local climate influence heat tolerance

    682 shares
    Share 273 Tweet 171
  • Researchers record first-ever images and data of a shark experiencing a boat strike

    546 shares
    Share 218 Tweet 137
  • Groundbreaking Clinical Trial Reveals Lubiprostone Enhances Kidney Function

    531 shares
    Share 212 Tweet 133
Science

Embark on a thrilling journey of discovery with Scienmag.com—your ultimate source for cutting-edge breakthroughs. Immerse yourself in a world where curiosity knows no limits and tomorrow’s possibilities become today’s reality!

RECENT NEWS

  • Causal Graph Neural Network Predicts Financial Systemic Risk and Contagion Trajectories
  • High Relative Hospital Volume Linked to Reduced Colon Cancer Recurrence Risk
  • Guatemala Volcán de Fuego Evacuations Lag Behind Escalating Climax Timescales
  • Model Reveals Design Rules for Microplates on Soft Viscoelastic Micropillar Foundations

Categories

  • Agriculture
  • Anthropology
  • Archaeology
  • Athmospheric
  • Biology
  • Biotechnology
  • Blog
  • Bussines
  • Cancer
  • Chemistry
  • Climate
  • Earth Science
  • Editorial Policy
  • Marine
  • Mathematics
  • Medicine
  • Pediatry
  • Policy
  • Psychology & Psychiatry
  • Science Education
  • Science News
  • Social Science
  • Space
  • Technology and Engineering

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 5,151 other subscribers

© 2026 Scienmag - Science Magazine

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • HOME
  • SCIENCE NEWS
  • CONTACT US

© 2026 Scienmag - Science Magazine

Discover more from Science

Subscribe now to keep reading and get access to the full archive.

Continue reading