Pith. sign in

REVIEW 2 major objections 5 minor 36 references

A Central Chilled Water Plant Model for Designing Learning-Based Controllers

T0 review · 2 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Chiller-plant model caps every heat exchanger at its physical limit

desk verdict Useful open-source CCWP simulator with a real index slip in the key constraint; the central saturation idea is sound but the paper must fix (25) and acknowledge the uncalibrated bounds. read the letter →

arxiv 2508.15649 v1 pith:KVQAMEZC submitted 2025-08-21 eess.SY cs.SY

Zhong Guo , Prabir Barooah This is my paper · ORCID
classification eess.SYcs.SY
keywords centralchilledwaterplantschillerplantmodelinglearning-basedcontrolreinforcementlearningheatexchangersaturationconstrainedoptimizationthermalenergystoragecoolingtowers
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to give learning-based controllers a simulation of a central chilled water plant that remains physically plausible even when the controller explores inputs far outside normal operation. Existing chiller, cooling tower, and coil models can return nonsense—like condenser water at 90 C when flow is cut to a tenth—because they do not cap heat exchange at capacity. The authors rework each equipment model so that the delivered cooling is the closest possible to the demand while respecting hard heat-exchanger limits, and assemble the pieces into a modular plant simulation implemented in Matlab and released openly. The value of the claim is that reinforcement-learning and extremum-seeking controllers need a simulator that lets them violate normal operating ranges and then show poor, but credible, performance in response.

What carries the argument

The load-bearing mechanism is a set of embedded nonlinear programs solved at every timestep. For the chiller, the program is min over the evaporator heat rate q_evap of (T_chws_set − T_chws_next)^2, subject to the evaporator and condenser heat-balance dynamics, the empirical capacity and EIR modifier curves, and the hard bound T_cdwr ≤ 40 C; its solution fixes both next states. For the cooling tower, the program is min over range and approach temperatures of (T_app − bT_app)^2, subject to range + approach = scope and fixed lower and upper bounds; once solved, the supply temperature follows. The aggregate cooling coil uses the same saturation logic through a simple min of demand and capacity.

What would settle it

Run the model with condenser water flow reduced to a tenth of nominal under high load; it predicts condenser return capped at 40 C. Or, better, repeat that experiment on a real plant: a measured steady-state condenser return above 40 C, or a plateau at a substantially different temperature, would falsify the wide-range-validity claim.

Watch

Extended reading notes

Core claim

The paper's central claim is that capacity saturation can be made a structural property of a chiller-plant component model by writing each state-propagation step as a small constrained optimization problem. Instead of letting an empirical chiller curve push condenser-return temperature to 90 C under low condenser flow, the proposed chiller model chooses the evaporator heat rate that brings chilled-water supply as close as possible to its setpoint while keeping both evaporator and condenser within their capacity bounds—so the condenser outlet stops at 40 C and the unsatisfied load appears as warm supply water. The cooling tower is handled the same way: range and approach are chosen closest to

Load-bearing premise

The paper assumes the fixed physical bounds chosen in Table I—such as 40 C condenser outlet and 15 C cooling-coil return—match the real saturation limits of a plant, and it explicitly does not verify this against experimental data; if those bounds are wrong, the claimed wide range of validity fails.

Editorial extensions

If this is right

  • Learning-based controllers can explore aggressive actions, such as very low condenser flow at high load, and receive poor but physically plausible consequences, which is what reward-based training needs.
  • The plant supports fleets of distinct chillers and cooling towers, so the training environment can reflect real heterogeneity instead of assuming identical machines.
  • The modular separation of equipment programs means a user can swap a component model or add equipment without rewriting a single plant-wide optimizer.
  • The model is implemented in a numerical computing environment and released openly, so the three-day closed-loop example and its exogenous weather, price, and load data can be reproduced by other groups.
  • Plantwide COP in the closed-loop example is about 5.84 against chiller design COPs of 6.28 and 7.35, which the paper reads as evidence that the plant-level model behaves reasonably.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper, the same constrained-propagation template should transfer to any heat exchanger whose empirical model can overshoot capacity, such as air-side coils or heat-recovery devices.
  • If plant-specific saturation bounds were estimated from logged measurements instead of fixed constants, the claimed wider range of validity would become quantitatively testable rather than qualitative.
  • Because the model deliberately returns degraded performance outside the envelope, it could double as a testbed for constraint-aware control, fault detection, or safety-layer design, not only for RL pretraining.
  • The paper's closed-loop demonstration uses a rule-based controller; comparing the same simulator under model-predictive control or adversarial inputs would probe how far the saturation assumptions hold.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 5 minor

Summary. The paper proposes a modular, open-source Matlab model of a central chilled water plant (aggregate cooling coil, TES, heterogeneous chillers, cooling towers, pumps). The main novelty is an optimization-based treatment of heat-exchanger saturation: the chiller's evaporator heat rate is computed by an NLP that minimizes chilled-water setpoint tracking error subject to equipment relations, and the cooling tower's range and approach temperatures are computed by an NLP subject to user-specified bounds. The authors claim that this construction makes the model respect heat-exchanger capacities irrespective of inputs, thereby giving a wider range of validity than standard EnergyPlus/MBL models, which can produce physically absurd outputs (e.g., condenser water return temperature near 90 C) when operated far from nominal conditions. A closed-loop simulation with a rule-based controller and a public code release are included.

Significance. If the model were implemented as described, it would address a real need: learning-based controllers require simulators that remain physically meaningful during exploratory actions outside nominal operating ranges. The paper has concrete strengths: it makes the code publicly available, the modular structure is sensible, the demonstration of the Electric:EIR model's failure at low condenser water flow is instructive, and the optimization-based saturation idea is a reasonable way to encode heat-exchanger limits. However, the central chiller NLP as printed does not actually constrain the condenser outlet temperature, so the main claimed property is not entailed by the equations. In addition, the saturation bounds are not calibrated or validated against data, and the paper itself states that experimental verification is not addressed. These issues are load-bearing for the abstract's 'wider range of validity' claim. The paper is a useful modeling contribution in progress, but it needs correction and more careful claims before it can be accepted.

major comments (2)
  1. [§II-D, Eq. (31) and constraint (25)] The NLP (31) is said to enforce condenser capacity through constraint (25), but (25) is written as T_cdws_k ≤ T_cdwr_k ≤ T_cdwr_UB. Here T_cdwr_k is a current state, i.e., an input to the optimization, not a function of the decision variable qevap_k. The variable that depends on qevap_k is the next state T_cdwr_{k+1}, defined by (23)-(24). Nothing in the constraint set (18)-(30) caps T_cdwr_{k+1}; qevap_k is only bounded above by min{qevap_req_k, qevap_UB_k}. Under low condenser water flow, the resulting qcd_k can therefore drive T_cdwr_{k+1} above 40 C. The saturation behavior shown in Figs. 5-6 is thus not entailed by the printed equations, and Remark 2's statement that (25) 'implicitly enforces an upper bound on qcd_k' is incorrect. The authors should add an explicit constraint on T_cdwr_{k+1} (or equivalently on qcd_k via (23)) to (31), and should verify that the public code implemen
  2. [§V and Table I] The abstract claims that the proposed model 'has a wider range of validity' than existing models, but this claim is not supported by experimental calibration or validation. The paper states in Section V that 'Verification of the proposed CCWP model against experimental data is not addressed in this paper.' The saturation bounds T_cwr_UB = 40 C, T_lwr_UB = 15 C, T_ran_LB/UB, and T_app_LB/UB in Table I are imposed a priori, so the simulated saturation is a consequence of the user-selected constraints, not a measured property of a real plant. If the true physical limits differ from these constants, the claimed wider validity will not hold. At minimum, the abstract and Section IV should be softened to claim only that the model enforces the specified bounds, leaving empirical validity as future work; ideally, the constants should be calibrated against plant data or a sensitivity study should
minor comments (5)
  1. [§II-D, Remark 2] The remark says the saturation feature is 'not present in the Electric:EIR:wSat model'; presumably this should be the original Electric:EIR model. Please fix the model name.
  2. [§IV] There is a duplicated phrase: 'The OA wet bulb temperature is obtained from is obtained from the National Solar Radiation Database.'
  3. [§II-B, Eq. (7)] The variable mdot_lw_UB is used in (7) and mentioned in the text but is not included in the Nomenclature. Please add it.
  4. [Table I] T_lwr_UB is listed twice with the same value 15 C. Please check whether one entry is intended to be a different bound, such as T_lws_UB.
  5. [§IV] The plantwide COP statement '93-80% of the chillers'' is ambiguous; it should read something like 'between 80% and 93% of the individual chiller COPs.'

Circularity Check

3 steps flagged · score 6.0 of 10

The capacity-saturation behavior presented as the model's key improvement is enforced by the same bounds that define it; the numerical 'verification' is by construction, and Eq. (25) as printed does not even bound the next condenser state.

  1. self definitional [Sec. II-D, Eqs. (23)-(25), (31), numerical verification around Fig. 5]
    "We modify the Electric:EIR chiller model by enforcing saturation of the condenser water outlet temperature at the predetermined value T cwr UB , which implicitly defines a hard upper bound on the capacity of the condenser... We see that when the condenser water flow rate becomes sufficiently small, the proposed model predicts that the return temperature saturates at its upper bound, 40◦C."

    The saturation is the constraint itself: the NLP (31) contains (25), T_cdws_k ≤ T_cdwr_k ≤ T_cdwr_UB, where T_cdwr_UB is the same predetermined value shown in Fig. 5. Observing T_cdwr = 40°C at low condenser flow is not an independent prediction; it is the optimizer returning a feasible point on that bound, assuming (25) is applied to the computed next state. If, as printed, (25) only bounds the input state T_cdwr_k, the saturation claim is not even entailed. Either way, the 'wider range of validity' rests on a user-chosen constant, not on measured behavior.

  2. self definitional [Sec. II-B, Eqs. (4)-(6), Fig. 10]
    "˙qcc_k = min( ˙qL_k, ˙qcc_UB,k), (5) ˙qcc_UB,k = Cpw ˙mlw_k (T lwr_UB − T lws_k ), (6) ... Fig. 10: 'Saturation of cooling coil heat exchange during periods of high load.'"

    The cooling-coil capacity is defined through the return-water upper bound T_lwr_UB. The min() in Eq. (5) clips the delivered heat to qcc_UB, so the exit temperature cannot exceed the corresponding T_lwr_UB. Showing that the high-load simulation reaches this upper bound is a direct consequence of the clipping operation, not an empirical confirmation of coil physics.

1 more flagged steps
  1. self definitional [Sec. II-E, Eq. (45), Case 1, numerical verification in Fig. 8]
    "Case 1: T cwr_k − T oawb_k < Tran_LB + T app_LB : ... no cooling can be provided to the cooling water. Therefore, T cws_k+1 = T cwr_k . ... We see from the figure that when the OA flow rate is really low ... the proposed model correctly predicts that ... the return water from the cooling tower would be at the same temperate as that of the water supplied to it ( 35◦C)."

    The no-cooling behavior is written into the model as Case 1 and is also enforced by the box constraints in the NLP (45) (T_ran_LB ≤ T_ran ≤ T_ran_UB and T_app_LB ≤ T_app ≤ T_app_UB). The 'prediction' that extreme tower inputs produce no cooling is a restatement of the model's case logic and constraints, not a validated physical result.

full rationale

The paper is not circular in the common 'fit-then-predict' sense: no data fitting occurs and the component correlations are imported from EnergyPlus/MBL. However, the central claimed improvement—heat-exchanger saturation and a wider range of validity—is installed as hard bounds (T_lwr_UB, T_cdwr_UB, T_ran, T_app bounds). The numerical demonstrations in Figs. 5, 8, and 10 therefore show the constraints doing their defining job rather than independently confirming the model's validity. The paper explicitly acknowledges the absence of experimental verification in Sec. V: 'Verification of the proposed CCWP model against experimental data is not addressed in this paper.' That is a validation gap, but combined with the by-construction saturation it makes the 'wider range of validity' claim unsupported by the reported tests. There is also a correctness issue in the chiller formulation: constraint (25) bounds T_cdwr_k, an input state, not T_cdwr_{k+1} that depends on the decision variable qevap_k via Eq. (23); as printed, the NLP does not enforce condenser saturation. If the implementation instead bounds T_cdwr_{k+1}, then saturation is exactly an imposed constraint. Self-citations [8,24,33] are preliminary/application references and are not load-bearing for the core modeling claim.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

All capacity bounds and empirical correlations are imported from prior literature or chosen by the authors; the paper's new content is the optimization layer. The list shows the central claim rests on unvalidated bounds and the assumption that empirical fits extrapolate.

free parameters (4)
  • T_cwr_UB = 40 C
    Maximum allowable condenser water return temperature, chosen a priori (Table I); defines condenser capacity saturation. No experimental calibration.
  • T_lwr_UB = 15 C
    Maximum allowable cooling coil return water temperature (Table I), set near the top of the ASHRAE typical 12-15 C range; not derived from coil design data.
  • T_ran_LB/T_ran_UB = 2.2 / 22.2 C
    Cooling tower range bounds in the YorkCalc:wSat NLP (45), chosen for the nominal tower; no independent validation.
  • T_app_LB/T_app_UB = 0 / 40 C
    Cooling tower approach bounds in the NLP (45); chosen for the nominal tower.
assumptions (5)
  • domain assumption Heat exchanger capacity can be modeled as a hard upper bound on outlet temperature (T_cwr_UB, T_lwr_UB).
    Used in chiller (25), cooling coil (6), and Remark 1; not derived from component physics.
  • domain assumption Low-level controllers' action can be abstracted as the solution of the NLP (31) and (45).
    The actual control loops inside the chiller and cooling tower are replaced by the optimization that minimizes setpoint error while respecting constraints.
  • domain assumption Empirical correlations from Electric:EIR and YorkCalc remain valid outside the original operating range once saturation bounds are added.
    The model relies on polynomial fits (22), (27), (42) beyond their fitted envelope; the paper adds saturation but does not revalidate the correlations.
  • domain assumption TES is perfectly insulated with no mixing between warm and cold sub-tanks.
    Used in TES equations (10)-(14), following the standard thermocline tank model from [27].
  • domain assumption Primary-secondary pumping with one-way bypass valve exactly describes flows.
    Equations (54)-(55) assume fixed nominal chilled water flow through each on-chiller and a bypass for the excess; used throughout Section III-A.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Central Chilled Water Plant Model for Designing Learning-Based Controllers." pith.science (2026). https://pith.science/paper/KVQAMEZC

@misc{pith2026250815649,
  author       = {Pith},
  title        = {Pith review of: A Central Chilled Water Plant Model for Designing Learning-Based Controllers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KVQAMEZC}},
  note         = {Machine review of arXiv:2508.15649}
}
read the original abstract

We describe a framework of modeling a central chilled water plant (CCWP) that consists of an aggregate cooling coil, a number of heterogeneous chillers and cooling towers, and a chilled water-based thermal energy storage system. We improve upon existing component models from the open literature using a constrained optimization-based framework to ensure that the models respect capacities of all the heat exchangers (cooling coils, chillers, and cooling towers) irrespective of the inputs provided. As a result, the proposed model has a wider range of validity compared to existing models; the latter can produce highly erroneous outputs when inputs are not within normal operating range. This feature is essential for training learning-based controllers that can choose inputs beyond normal operating conditions and is lacking in currently available models. The overall plant model is implemented in Matlab and is made publicly available. Simulation of a CCWP with closed loop control is provided as an illustration.

Figures

Figures reproduced from arXiv: 2508.15649 by the authors.

Figure 1
Figure 1. Configuration of a Central Chilled Water Plant considered in this paper. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Evidence of limited region of validity of the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Aggregate load/cooling coil. Model inputs (including disturbance) [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Schematic of a water-cooled vapor compression chiller, showing [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Prediction of the proposed Electric:EIR:wSat chiller model: (top) steady state condenser outlet temperature and (bottom) evaporator outlet temperature (vertical axis) as a function of two inputs (horizontal axes), while other inputs are held constant (T chwr = 12◦C, m˙…
Figure 7
Figure 7. Figure 7: Cooling tower flows and their related variables. Inlet and outlet [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 6
Figure 6. Figure 6: The proposed chiller model Electric:EIR:wSat’s predic￾tion (top) condenser outlet temperature and (bottom) evaporator outlet temperature at steady state (vertical axis) as a function of two inputs (horizontal axes), while other inputs are held constant (m˙ chw = 31.86 …
Figure 8
Figure 8. Figure 8: Cooling tower model prediction: T cws ∞ as a function of the two mass flow rates while other inputs are held constant. III. CENTRAL CHILLED WATER PLANT MODEL The widely deployed primary-secondary pumping strategy is considered in our CCWP model [30]. Two subsystems — c…
Figure 9
Figure 9. Figure 9: Closed loop simulation: All three exogenous inputs [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: Closed loop simulation: Saturation of cooling coil heat exchange [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]
Figure 11
Figure 11. Figure 11: Closed loop simulation: Chilled water and cooling water temper [PITH_FULL_IMAGE:figures/full_fig_p012_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 35 canonical work pages

  1. [1]

    Improved cooling tower control of legacy chiller plants by optimizing the condenser water set point,

    S. Huang, W. Zuo, and M. D. Sohn, “Improved cooling tower control of legacy chiller plants by optimizing the condenser water set point,” Building and Environment , vol. 111, pp. 33–46, 2017. [Online]. Available: https://www.sciencedirect. com/science/article/pii/S0360132316304036

    work page 2017

  2. [2]

    Model predictive control of central chiller plant with thermal energy storage via dynamic programming and mixed-integer linear programming,

    K. Deng, Y . Sun, S. Li, Y . Lu, J. Brouwer, P. G. Mehta, M. Zhou, and A. Chakraborty, “Model predictive control of central chiller plant with thermal energy storage via dynamic programming and mixed-integer linear programming,” IEEE Transactions on Automation Science and Engineering, vol. 12, no. 2, pp. 565–579, 2015

    work page 2015

  3. [3]

    District heating and cooling systems – framework for modelica-based simulation and dynamic optimization,

    G. Schweiger, P.-O. Larsson, F. Magnusson, P. Lauenburg, and S. Velut, “District heating and cooling systems – framework for modelica-based simulation and dynamic optimization,” Energy, vol. 137, pp. 566–578, 2017. [Online]. Available: https://www.sciencedirect.com/science/ article/pii/S0360544217308691

    work page 2017

  4. [4]

    A case study of economic optimization of hvac systems based on the stanford university campus airside and waterside systems,

    N. R. Patel, M. J. Risbeck, J. B. Rawlings, C. T. Maravelias, M. J. Wenzel, and R. D. Turney, “A case study of economic optimization of hvac systems based on the stanford university campus airside and waterside systems,” in International High Performance Buildings Conference, vol. paper 253, 2018

    work page 2018

  5. [5]

    Virtual testbed for model predictive control development in district cooling systems,

    L. Zabala, J. Febres, R. Sterling, S. L ´opez, and M. Keane, “Virtual testbed for model predictive control development in district cooling systems,” Renewable and Sustainable Energy Reviews , vol. 129, p. 109920, 2020. [Online]. Available: https://www.sciencedirect.com/science/ article/pii/S1364032120302112

    work page 2020

  6. [6]

    Machine learning vs. hybrid machine learning model for optimal operation of a chiller,

    S. Park, K. U. Ahn, S. Hwang, S. Choi, and C. S. Park, “Machine learning vs. hybrid machine learning model for optimal operation of a chiller,” Science and Technology for the Built Environment, vol. 25, no. 2, pp. 209–220, 2019

    work page 2019

  7. [7]

    Model-free optimal chiller loading method based on Q-learning,

    S. Qiu, Z. Li, Z. Li, and X. Zhang, “Model-free optimal chiller loading method based on Q-learning,” Science and Technology for the Built Environment, vol. 26, no. 8, pp. 1100–1116, 2020

    work page 2020

  8. [8]

    Reinforcement learning for optimal control of a district cooling energy plant,

    Z. Guo, A. R. Coffman, and P. Barooah, “Reinforcement learning for optimal control of a district cooling energy plant,” in American Control Conference (ACC), June 2022, pp. 3329– 3334

    work page 2022

Show all 36 references
  1. [9]

    Optimizing industrial HV AC systems with hierarchi- cal reinforcement learning,

    W. Wong, P. Dutta, O. V oicu, Y . Chervonyi, C. Paduraru, and J. Luo, “Optimizing industrial HV AC systems with hierarchi- cal reinforcement learning,” arXiv:2209.08112, 2022

    2022 arXiv

  2. [10]

    Model-free MIMO control tuning of a chiller process using reinforcement learning,

    C. Rosdahl, B. Bernhardsson, and B. Eisenhower, “Model-free MIMO control tuning of a chiller process using reinforcement learning,” Science and Technology for the Built Environment , vol. 29, no. 8, pp. 782–794, 2023

    2023

  3. [11]

    Real-time optimization of a chilled water plant with parallel chillers based on extremum seeking control,

    B. Mu, Y . Li, J. M. House, and T. I. Salsbury, “Real-time optimization of a chilled water plant with parallel chillers based on extremum seeking control,”Applied Energy, vol. 208, pp. 766–781, 2017. 13

    2017

  4. [12]

    Energy optimization of an in-service commercial building chiller plant via extremum seeking control,

    H. D. Vu, B. D. Keating, A. G. Alleyne, K. S. Chai, and Z. Zhang, “Energy optimization of an in-service commercial building chiller plant via extremum seeking control,” in Inter- national High Performance Buildings Conference , 2018

    2018

  5. [13]

    2020 ASHRAE Handbook — HV AC Systems and Equipment,

    ASHRAE, “2020 ASHRAE Handbook — HV AC Systems and Equipment,” 2020

    2020

  6. [14]

    EnergyPlus: Energy simulation program,

    D. B. Crawley, L. K. Lawrie, C. O. Pedersen, and F. C. Winkel- mann, “EnergyPlus: Energy simulation program,” ASHRAE Journal, vol. 42, no. 4, pp. 49–56, 2000

    2000

  7. [15]

    Modelica buildings library,

    M. Wetter, W. Zuo, T. S. Nouidui, and X. Pang, “Modelica buildings library,” Journal of Building Performance Simulation, vol. 7, no. 4, pp. 253–270, 2014. [Online]. Available: https://doi.org/10.1080/19401493.2013.765506

    2014

  8. [16]

    Development, validation, and assessment of a high fidelity chilled water plant model,

    B. P. Baillie and G. M. Bollas, “Development, validation, and assessment of a high fidelity chilled water plant model,” Applied Thermal Engineering , vol. 111, pp. 477–488, 2017

    2017

  9. [17]

    A multi-agent control approach for optimization of central cooling plants,

    R. Jaramillo, “A multi-agent control approach for optimization of central cooling plants,” Ph.D. dissertation, Purdue Univer- sity, 2018

    2018

  10. [18]

    Open-source modelica models for the control performance simulation of chiller plants with water-side economizer,

    C. Fan, K. Hinkelman, Y . Fu, W. Zuo, S. Huang, C. Shi, N. Mamaghani, C. Faulkner, and X. Zhou, “Open-source modelica models for the control performance simulation of chiller plants with water-side economizer,” Applied Energy , vol. 299, p. 117337, 2021

    2021

  11. [19]

    Openbuildingcontrol,

    L. B. N. Lab and P. N. N. Lab, “Openbuildingcontrol,” accessed on Jun.03.2022. [Online]. Available: https://www. energy.gov/eere/buildings/openbuildingcontrol

    2022

  12. [20]

    Ibpsa project 1,

    M. W. (LBNL) and C. van Treeck (RWTH Aachen), “Ibpsa project 1,” accessed on Jun.03.2022. [Online]. Available: https://ibpsa.github.io/project1/index.html

    2022

  13. [21]

    U.S. Dept. of Energy, EnergyPlus Engineering Reference . US DEPARTMENT OF ENERGY, 2022, version 22.2.0. [Online]. Available: https://energyplus.net/documentation

    2022

  14. [22]

    J. C. Hensley, Cooling tower fundamentals . Marley Cooling Tower Company Stockton, CA, 1985

    1985

  15. [23]

    Optimal chiller sequencing control in an office building considering the variation of chiller maximum cooling capacity,

    Z. Liu, H. Tan, D. Luo, G. Yu, J. Li, and Z. Li, “Optimal chiller sequencing control in an office building considering the variation of chiller maximum cooling capacity,” Energy and Buildings, vol. 140, pp. 430–442, 2017. [Online]. Available: https://www.sciencedirect.com/scie...

    2017

  16. [24]

    Opti- mal control of district cooling energy plant with reinforcement learning and model predictive control,

    Z. Guo, A. Chaudhari, A. R. Coffman, and P. Barooah, “Opti- mal control of district cooling energy plant with reinforcement learning and model predictive control,” ASME Journal of Engineering for Sustainable Buildings and Cities, vol. 5, p. 12, February 2024

    2024

  17. [25]

    Model predictive control for energy-efficient HV AC operation with humidity and latent heat considerations,

    N. S. Raman, K. Devaprasad, B. Chen, H. A. Ingley, and P. Barooah, “Model predictive control for energy-efficient HV AC operation with humidity and latent heat considerations,” Applied Energy, vol. 279, p. 115765, December 2020

    2020

  18. [26]

    Atlanta, GA: ASHRAE, 2022

    American Society of Heating Refrigerating and Air Condi- tioning Engineers, Standard 90.1-2022-Energy Standard for Sites and Buildings Except Low-Rise Residential Buildings . Atlanta, GA: ASHRAE, 2022

    2022

  19. [27]

    Glazer, ASHRAE Design Guide for Cool Thermal Storage (2nd Edition)

    J. Glazer, ASHRAE Design Guide for Cool Thermal Storage (2nd Edition) . American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc., 2019

    2019

  20. [28]

    A mixed-integer linear programming model for real- time cost optimization of building heating, ventilation, and air conditioning equipment,

    M. J. Risbeck, C. T. Maravelias, J. B. Rawlings, and R. D. Turney, “A mixed-integer linear programming model for real- time cost optimization of building heating, ventilation, and air conditioning equipment,” Energy and Buildings , vol. 142, pp. 220–235, 2017

    2017

  21. [29]

    The ASHRAE handbook fundamentals (SI Edi- tion),

    ASHRAE, “The ASHRAE handbook fundamentals (SI Edi- tion),” 2016

    2016

  22. [30]

    American Society of Heating, Refrigerating and Air-Conditioning Engi- neers, Inc., 2019

    ——, ASHRAE Handbook- HVAC applications . American Society of Heating, Refrigerating and Air-Conditioning Engi- neers, Inc., 2019

    2019

  23. [31]

    CasADi: a software framework for nonlinear op- timization and optimal control,

    J. A. E. Andersson, J. Gillis, G. Horn, J. B. Rawlings, and M. Diehl, “CasADi: a software framework for nonlinear op- timization and optimal control,” Mathematical Programming Computation, vol. 11, no. 1, pp. 1–36, Mar 2019

    2019

  24. [32]

    On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming,

    A. W ¨achter and L. T. Biegler, “On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming,” Mathematical Programming , vol. 106, no. 1, pp. 25–57, Mar 2006

    2006

  25. [33]

    Reinforcement learning for optimal control of dis- trict cooling energy plant,

    Z. Guo, “Reinforcement learning for optimal control of dis- trict cooling energy plant,” Ph.D. dissertation, University Of Florida, December 2023

    2023

  26. [34]

    PJM Interconnection Real-Time Hourly LMPs,

    PJM data miner, “PJM Interconnection Real-Time Hourly LMPs,” https://www.pjm.com/markets-and-operations/etools/ data-miner-2, 2022, [Online, accessed 2022-10-02]

    2022

  27. [35]

    David Brooks (PE, CxA, Principal System Engineer at Akin Brooks Engineering), personal communication, 2023

    2023

  28. [36]

    From laboratory to on-site operation: Reevaluation of em- pirically based electric water chiller models,

    B. Zhang, S. You, S. Wang, X. Ding, C. Wang, and Y . Gao, “From laboratory to on-site operation: Reevaluation of em- pirically based electric water chiller models,” in Building Simulation, vol. 15, no. 2. Springer, 2022, pp. 213–232

    2022

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.