A Coupled Dynamic Model of HVAC Energy Demand, Cabin Thermal Conditions, and Air Quality in Passenger Cars
Abstract:
Ventilation strategies influence the energy demand of passenger-car heating, ventilation, and air-conditioning (HVAC) systems by changing outdoor-air loads, occupant-generated carbon dioxide (CO$_2$) accumulation, and particle ingress. Assessing these competing effects requires consistent treatment of cabin heat transfer, moisture, airflow, and filtration. This study investigates the energy–air-quality trade-offs associated with alternative cabin ventilation strategies using a reduced-order dynamic model. Transient heat and moisture balances were coupled with occupant-generated CO$_2$, fine particulate matter (PM$_{2.5}$) transport and filtration, and HVAC electrical demand. Fresh-air operation, high recirculation, and a rule-based switching strategy were compared in an illustrative 30-min summer simulation with four occupants. Under high recirculation, the predicted end-of-period CO$_2$ concentration reached approximately 7,059 ppm, compared with 1,030 ppm under fresh-air operation. The corresponding PM$_{2.5}$ concentrations were approximately 6.7 and 35.0 μg/m$^3$, while cumulative HVAC electricity use was 0.332 and 0.379 kWh, respectively. High recirculation therefore reduced simulated electricity use by 12.4% but allowed substantially greater CO$_2$ accumulation. The rule-based strategy consumed 0.378 kWh and produced final CO$_2$ and PM$_{2.5}$ concentrations close to those of continuous fresh-air operation. These model-based results indicate that cabin energy savings must be assessed alongside ventilation and pollutant exposure requirements. The framework provides a common basis for comparing HVAC electricity demand and cabin environmental conditions under prescribed ventilation configurations.
1. Introduction
Passenger-car heating, ventilation, and air-conditioning (HVAC) systems consume energy to maintain cabin environmental conditions. In battery-electric vehicles, HVAC operation draws on the same stored electrical energy as propulsion, making cabin conditioning an integral part of vehicle energy management. Lee et al. [1] examined passenger electric-vehicle energy consumption under real-world driving conditions and demonstrated the importance of ambient temperature. Zhu et al. [2] reviewed the integration of cabin conditioning, battery cooling, and powertrain thermal management, while Geleta et al. [3] surveyed control strategies for reducing cabin-conditioning energy use while maintaining thermal comfort. These studies connect cabin environmental requirements with vehicle electrical demand.
Ventilation directly affects this relationship. Outdoor air dilutes occupant-generated carbon dioxide (CO$_2$), but it can also introduce fine particulate matter (PM$_{2.5}$) and increase sensible and latent conditioning loads. Recirculation permits repeated filtration and can reduce outdoor-air conditioning demand while limiting CO$_2$ removal. Wei et al. [4] examined the effects of recirculation on cabin air quality and energy use across different climates and outdoor particle concentrations. Mohan and Varghese [5] investigated the simultaneous control of CO$_2$ and PM$_{2.5}$, addressing their opposing responses to ventilation. Russi et al. [6] combined cabin environmental measurements with HVAC electricity measurements during heating and cooling, and Lesage et al. [7] evaluated air quality and energy consumption through system simulation. Together, these studies show that ventilation strategies require a joint assessment of cabin conditions and electrical demand.
The representation of cabin thermal conditions depends on the intended engineering application. Zhao et al. [8] reviewed vehicle-cabin thermal models and discussed the relationship between model complexity, boundary conditions, and predictive accuracy. Torregrosa-Jaime et al. [9] developed a dynamic two-zone lumped thermal model and validated it under variable ambient conditions. Their work demonstrates how conservation-based models can describe transient cabin loads while retaining explicit representations of air renewal and solar radiation.
Predicting cabin thermal states and assessing occupant comfort involve different quantities. Chen et al. [10] reviewed thermal-comfort models for nonuniform and transient cabin environments. A subsequent comparative evaluation examined model accuracy, generalizability, and measurement feasibility [11]. ISO 14505-2:2026 specifies an equivalent-temperature approach for evaluating thermal conditions in vehicles [12]. These approaches distinguish mean cabin-air temperature from the thermal conditions experienced by individual occupants. The present study uses mean air temperature and humidity ratio to describe the bulk thermal and moisture states and to calculate conditioning loads.
Pollutant transport requires separate descriptions of internally generated CO$_2$ and outdoor particles because they respond differently to airflow. Wei et al. [13] developed a size-resolved particle and CO$_2$ model incorporating recirculation, infiltration, filtration, and deposition. Chang et al. [14] used computational fluid dynamics to investigate ventilation strategies for balancing CO$_2$ and suspended particle concentrations under different occupancies. These studies show why outdoor-air exchange alone cannot characterize cabin air quality: increasing outdoor airflow can dilute CO$_2$ while increasing particle ingress.
Filtration further modifies this response. Wang et al. [15] investigated vehicle cabin-filter efficiency, pressure drop, and service life. Renck and Jung [16] examined cabin filters and portable air purifiers, including the dependence of particle-removal performance on HVAC fan settings. Their findings support treating airflow and filtration as separate model inputs. Interpreting cabin air-quality predictions also requires attention to the measurements used for comparison and the meaning of the selected concentration thresholds. Silva et al. [17] reviewed the accuracy, applications, and limitations of low-cost indoor air-quality sensors. Persily [18] clarified that 1,000 ppm CO$_2$ is not a universal indoor-air-quality limit. Accordingly, the CO$_2$ thresholds used here specify a ventilation-control policy rather than a general health criterion.
These concentration thresholds connect cabin air-quality assessment with ventilation control, where outdoor-air supply and recirculation are adjusted alongside thermal requirements. Ramesh Babu et al. [19] investigated adaptive recirculation during electric-truck heating using coupled computational fluid dynamics and thermoregulation modeling, with humidity and CO$_2$ informing the recirculation strategy. Lesage et al. [20] examined CO$_2$-responsive ventilation in an electric passenger vehicle and related its operation to thermal conditions and energy use. At the integrated thermal-system level, Wang et al. [21] developed model predictive control for cabin and battery temperature management. These studies demonstrate how environmental requirements can guide the selection of energy-efficient operating inputs.
The electrical comparison must account for air circulation as well as heating or cooling. Zhang et al. [22] investigated low-temperature HVAC energy consumption while considering blower efficiency. When ventilation policies use different supply flows or fan settings, their energy difference reflects the complete operating configuration. Reporting airflow, conditioning demand, and fan power separately makes the source of that difference traceable.
Comparing the resulting energy and environmental trajectories requires both consistent physical accounting and checks on the numerical solution. Oberkampf and Roy [23] distinguished numerical verification from physical validation in scientific computing. Ordinary-differential-equation methods [24] and software implementations such as SciPy’s integration routines [25] provide tools for solving transient balances. Analytical limiting cases, step-size comparisons, and integrated conservation residuals provide complementary checks on the computed solutions.
Uncertainty propagation and parameter sensitivity provide complementary forms of model assessment. The Joint Committee for Guides in Metrology guides address uncertainty expression and propagation [26], [27], while Morris screening and global sensitivity methods characterize parameter influence over a sampled input domain [28], [29]. The present study uses local finite differences to examine responses around the prescribed summer baseline.
The preceding studies establish the importance of jointly assessing cabin conditions and HVAC energy use. For comparisons across ventilation policies, the remaining methodological task is to keep the airflow pathways, source terms, and electrical accounting consistent and explicit. Controlled outdoor-air supply, infiltration, and recirculation perform different physical functions. Outdoor air and infiltration exchange heat, moisture, and contaminants with the surroundings, whereas recirculation returns cabin air through the HVAC system and enables repeated filtration. A compact formulation that preserves these distinctions allows differences in cabin conditions and electricity use to be traced to the prescribed operating inputs.
This study develops a reduced-order dynamic model coupling cabin temperature, humidity ratio, CO$_2$ concentration, and PM$_{2.5}$ concentration with HVAC electrical demand. The formulation integrates occupant sources, heat transfer, outdoor-air exchange, recirculation filtration, and conditioning loads within four state balances and a common electrical-demand calculation. Its methodological contribution is a transparent comparison framework in which changes in cabin conditions and energy use can be traced to the same physical inputs and operating policies.
The model is applied to a prescribed 30-min summer scenario with four occupants. Continuous fresh-air operation, high recirculation, and a CO$_2$-based switching policy are compared using transient cabin conditions and cumulative electricity consumption. Conditioning and fan-power contributions are reported separately, and switching behavior is interpreted in relation to the CO$_2$ equilibrium associated with each ventilation mode. The resulting analysis provides a basis for evaluating vehicle HVAC energy efficiency alongside cabin environmental requirements under explicitly defined operating conditions.
2. Engineering Model and Assumptions
The passenger compartment is represented as a single, well-mixed control volume coupled to the HVAC system and the outdoor environment. Four state variables describe its transient response: mean air temperature, humidity ratio, CO$_2$ concentration, and PM$_{2.5}$ concentration. HVAC electrical demand is calculated from the conditioning load and air-circulation power. This representation connects cabin environmental conditions to energy use through explicit heat and mass balances, following conservation-based approaches to cabin thermal response and pollutant transport [9], [13].
The model boundary includes heat transfer through the vehicle envelope, solar gains, occupant heat and moisture release, occupant-generated CO$_2$, outdoor-air exchange, and particle filtration and deposition. The effective cabin volume is denoted by $V$, while $C_\text{th}$ represents the combined thermal capacitance assigned to the cabin air and the interior materials participating in the modeled transient. Envelope heat transfer is described by the overall conductance $UA$. These quantities are effective model parameters; the numerical values used in the summer demonstration are specified in Section 4.
The HVAC supply consists of controlled outdoor air and recirculated cabin air. For a total supply flow $Q_\text{sup}$ and recirculation fraction $r$,
$Q_{\text {oa }}=(1-r) Q_{\text {sup }}, Q_{\mathrm{r}}=r Q_{\text {sup }}, 0 \leq r \leq 1$
where, $r$ applies only to the controlled HVAC supply. Uncontrolled infiltration is represented separately by $Q_\text{inf}$, giving the total outdoor-air exchange $Q_\text{f}=Q_\text{oa}+Q_\text{inf}$. At approximately constant cabin pressure and volume, this inflow is balanced by an equal exhaust or exfiltration flow. Recirculation remains internal to the cabin–HVAC system: it enables repeated particle filtration but does not directly dilute cabin CO$_2$. Outdoor-air exchange, by contrast, transports heat, moisture, and pollutants between the cabin and its surroundings [4], [13].
Figure 1 presents the cabin boundary and HVAC airflow arrangement. Outdoor air and return air enter the mixing section, pass through the filtration and conditioning components, and are supplied to the cabin. Occupant emissions, solar gains, and envelope heat transfer act on the cabin control volume. Particle entry through uncontrolled infiltration is treated separately from filtered outdoor-air supply.

Filtration is represented by effective single-pass removal efficiencies for the outdoor-air and recirculation paths, denoted by $\eta_\text{f}$ and $\eta_\text{r}$, respectively. These coefficients account for the particle-removal performance relevant to the modeled PM$_{2.5}$ concentration. For a shared filter exposed to comparable particle distributions and operating conditions, the two efficiencies can be assigned the same value. Separate coefficients allow different filtration arrangements to be represented. Filter performance depends on airflow, particle size, and loading [13], [15], [16]. The present formulation represents these effects through effective removal efficiencies prescribed for each operating condition.
Table 1 summarizes the model parameters, prescribed inputs, and their characterization basis. The values adopted for the illustrative summer scenario are specified in Section 4.
| Parameter or Input | Physical Meaning | Characterization Basis |
|---|---|---|
| $V$, $C_\text{th}$, $UA$ | Effective cabin volume, thermal capacitance, and envelope conductance | Vehicle geometry, thermal characterization, or stated engineering assumptions |
| $Q_\text{sup}$, $r$, $Q_\text{inf}$ | Supply airflow, recirculation fraction, and uncontrolled outdoor-air exchange | HVAC airflow data, leakage characterization, or prescribed scenario values |
| $\eta_\text{f}$, $\eta_\text{r}$, $p_\text{inf}$, $k_\text{d}$ | Filter efficiencies, infiltration penetration factor, and particle deposition rate | Component data, relevant literature, or stated assumptions |
| $T_\text{o}$, $w_\text{o}$, $I_\text{sol}$ | Outdoor temperature, humidity ratio, and solar irradiance | Environmental records or prescribed boundary conditions |
| $c_\text{o}$, $p_\text{m,o}$ | Outdoor CO$_2$ and PM$_{2.5}$ concentrations | Monitoring data or prescribed concentrations |
| $n$, $\dot{q}_\text{occ}$, $\dot{m}_\text{v,occ}$, $g_{\mathrm{CO}_2}$ | Occupancy and per-person sensible heat, moisture, and CO$_2$ generation | Occupant activity assumptions or documented source rates |
| ${\rm COP}_\text{h}$, ${\rm COP}_\text{c}$, $P_\text{fan}$, $P_\text{aux}$ | Heating and cooling performance, fan power, and other auxiliary power | Equipment data or stated electrical-demand assumptions |
The model assumes well-mixed cabin air, with uniform mean temperature, humidity ratio, and pollutant concentrations. Air properties and effective envelope parameters remain constant within each simulation, and the windows are closed. Supply jets, local radiation asymmetry, breathing-zone gradients, and volatile organic compounds are outside the state description. The thermal states characterize bulk cabin conditions; local occupant-comfort assessment requires additional information on airflow, surface temperatures, and occupant exposure [9], [10], [11], [12], [19].
Temperature is expressed in kelvin in the governing equations and reported in degrees Celsius. Volumetric flow rates are converted to m$^3\cdot$s$^{-1}$ for calculation, thermal and electrical power are expressed in watts, and cumulative electrical energy is reported in kilowatt-hours. Humidity ratio is expressed in kilograms of water vapour per kilogram of dry air. CO$_2$ is calculated as a dimensionless volume fraction and reported in ppm, while PM$_{2.5}$ mass concentration is expressed in μg/m$^{3}$.
The thermal balance combines HVAC sensible conditioning, envelope heat transfer, outdoor-air exchange, solar gain, and occupant sensible heat:
where, $T$ and $T_\text{o}$ are cabin and outdoor temperatures, $\rho$ is air density, and $c_p$ is its specific heat capacity. The sensible conditioning rate ${\dot{Q}}_\text{HVAC,s}$ is positive for heating and negative for cooling. It represents the thermal energy added or removed by HVAC conditioning, with outdoor-air exchange accounted for separately by the third term. Accordingly, it excludes the ventilation heat-transfer term already included in Eq. (1).
The effective solar load is:
$\dot{Q}_{\mathrm{sol}}=\alpha_{\mathrm{sol}} A_{\mathrm{sol}} I_{\mathrm{sol}}$
where, $A_\text{sol}$ is an effective solar-exposure area and $\alpha_{\mathrm{sol}}$ combines the transmission and absorption effects represented by the lumped model. A prescribed solar heat-gain value can equivalently be applied directly. The occupant term uses the sensible heat-release rate per person, $\dot{q}_{\mathrm{occ}}$, so that moisture-related energy is not included twice.
Moisture conservation is formulated using cabin and outdoor humidity ratios, $w$ and $w_\text{o}$:
$m_{\mathrm{da}} \frac{\mathrm{~d} w}{\mathrm{~d} t}=\dot{m}_{\mathrm{da}, \mathrm{f}}\left(w_{\mathrm{o}}-w\right)+n \dot{m}_{\mathrm{v}, \mathrm{occ}}+\dot{m}_{\mathrm{v}, \mathrm{HVAC}}-\dot{m}_{\mathrm{v}, \mathrm{cond}}, \dot{m}_{\mathrm{da}, \mathrm{f}} \simeq \rho_{\mathrm{da}} Q_{\mathrm{f}}$
The approximately constant dry-air mass is $m_\text{da}$, and $\rho_\text{da}$ is the corresponding dry-air density. Occupant moisture generation is ${\dot{m}}_\text{v,occ}$. The signed HVAC moisture-transfer rate ${\dot{m}}_\text{v,HVAC}$ is negative during coil dehumidification and positive during humidification. The nonnegative term ${\dot{m}}_\text{v,cond}$ represents condensation on interior surfaces. These terms describe separate moisture-removal paths; water removed at the cooling coil is included only in ${\dot{m}}_\text{v,HVAC}$.
Relative humidity is obtained from $T$, $w$, and cabin pressure $p$. The water-vapour partial pressure is:
$p_{\mathrm{v}}=\frac{p w}{0.622+w}$
and relative humidity is calculated as $p_\text{v}/p_\text{sat}\left(T\right)$, where $p_\text{sat}$ is the saturation vapour pressure. A windscreen-condensation assessment additionally requires the interior glass temperature $T_\text{g}$. The condition $T_\text{g}$ < $T_\text{dp}$, with $T_\text{dp}$ denoting the cabin-air dew-point temperature, identifies the thermodynamic potential for condensation. Glass temperature is an additional prescribed quantity or submodel output; it is not one of the four cabin states.
Figure 2 summarizes how ventilation, occupant sources, conditioning, and filtration connect the thermal and environmental balances.

For CO$_2$, let $c$ and $c_\text{o}$ denote cabin and outdoor volume fractions. The balance is:
where, $g_{\mathrm{CO_2}}$ is the equivalent volumetric CO$_2$ generation rate per person, expressed at the reference conditions used for the airflow calculation. Multiplication of $c$ by 10$^6$ gives the reported concentration in ppm. Recirculated flow does not appear as a dilution term because its CO$_2$ concentration is that of the returning cabin air. Eq. (2) therefore links occupant-generated accumulation directly to effective outdoor-air exchange, consistent with the distinct roles of recirculation and outdoor-air exchange [4], [5], [13].
The PM$_{2.5}$ mass balance distinguishes filtered outdoor-air entry from particle penetration through infiltration:
where, $p_\text{m}$ and $p_\text{m,o}$ are the cabin and outdoor PM$_{2.5}$ mass concentrations. The first two terms represent particle entry through filtered outdoor-air supply and uncontrolled infiltration, respectively. The infiltration penetration factor $p_\text{inf}$ is the fraction of outdoor particle concentration transmitted through the leakage path. The subsequent terms describe particle removal by exhaust or exfiltration, recirculation filtration, and deposition, followed by the internal source $S_\text{m}$. The deposition coefficient $k_\text{d}$ has units of s$^{-1}$, and $S_\text{m}$ is expressed in μg$\cdot$s$^{-1}$. This separation preserves the different entry and removal mechanisms associated with each airflow path [13].
HVAC electrical demand combines sensible conditioning, latent cooling, air circulation, and auxiliary loads:
The heating and cooling coefficients of performance are ${\rm COP}_\text{h}$ and ${\rm COP}_\text{c}$, respectively. The nonnegative latent-cooling load ${\dot{Q}}_\text{lat}$ represents moisture removal at the cooling coil and is consistent with the coil dehumidification rate in the moisture balance. Condensation on interior surfaces is treated separately.
The fan contribution is calculated from total supply flow, system pressure rise $\Delta p$, and overall fan–motor efficiency $\eta_\text{fan}$. A measured or prescribed fan-power value may replace this term. Filter pressure losses are included in the system pressure rise, and $P_\text{aux}$ accounts only for electrical loads excluded from the preceding terms. The contributions included in each numerical scenario are specified with its operating settings.
With state, control, and disturbance vectors defined as:
$\mathbf{x}=\left[\begin{array}{llll}T & w & c & p_{\mathrm{m}}\end{array}\right]^{\mathrm{T}}$,
$\mathbf{u}=\left[\begin{array}{llll}Q_{\mathrm{sup}} & r & \dot{Q}_{\mathrm{HVAC}, \mathrm{s}} & \dot{m}_{\mathrm{v}, \mathrm{HVAC}}\end{array}\right]^{\mathrm{T}}$,
$\mathbf{d}=\left[\begin{array}{llllll}T_{\mathrm{o}} & w_{\mathrm{o}} & c_{\mathrm{o}} & p_{\mathrm{m}, \mathrm{o}} & I_{\mathrm{sol}} & n\end{array}\right]^{\mathrm{T}}$
the coupled model is written as:
The parameter vector $\boldsymbol{\theta}$ contains the effective thermal properties, infiltration characteristics, solar-gain parameters, occupant source coefficients, filtration and deposition coefficients, and equipment-performance parameters. These quantities are distinguished from the operating inputs and environmental disturbances. The formulation therefore evaluates cabin conditions and electrical demand under the same prescribed ventilation policy, with each response linked to the corresponding heat and mass exchanges.
3. Numerical Implementation and Assessment Methods
The coupled balances in Section 2 define an initial-value problem,
where, $\mathbf{x}=\left[T,w,c,p_m\right]^\text{T}$, $\mathbf{u}$ contains the HVAC and ventilation inputs, and $\mathbf{d}$ contains the prescribed environmental and occupancy conditions. The parameter vector $\boldsymbol{\theta}$ retains the definitions introduced in Section 2. Separating parameters from operating inputs allows the effects of ventilation policies to be examined without changing the underlying cabin description.
The summer demonstration uses explicit time integration with a fixed step of $\Delta t=$ 1.2 s over a 30-min interval. At each step, the active ventilation mode determines the outdoor-air and recirculation flows, filtration terms, and fan power. The state derivatives are evaluated from the coupled balances, and HVAC electrical power is integrated over the operating interval to obtain cumulative electricity use. The reported initial conditions, airflow settings, thermal parameters, and equipment settings are summarized in Section 4.
For the rule-based policy, the ventilation mode is retained between switching events. The controller changes from high recirculation to fresh-air operation when cabin CO$_2$ exceeds 1,100 ppm and returns to recirculation when it falls below 850 ppm. The two thresholds introduce hysteresis, preventing immediate reversal at a single switching concentration. Switching is evaluated on the integration grid, so the recorded transition time has the temporal resolution of the numerical step. Comparisons between policies use a common output grid and the same environmental forcing.
Numerical verification examines whether the implementation solves the governing equations correctly, whereas physical validation examines agreement with independent measurements [23]. The numerical assessment considered here comprises step-size comparisons, integrated balance residuals, and analytical solutions for constant-input cases. These diagnostics are distinguished from differences in cabin conditions and electricity use between ventilation policies.
For a state $x_i$, the normalized difference between solutions obtained with steps $h$ and $h$/2 is defined as:
where, $\mathcal{T}$ is the common comparison grid, and $x_{i,\text{scale}}>$0 is a characteristic scale with the same units as the state variable. This scale prevents division by a near-zero reference magnitude. Under threshold-based operation, state differences are assessed alongside switching times and cumulative electricity use because small concentration differences near a threshold can change the transition time.
An integrated balance residual provides a complementary check on the numerical solution. For a stored quantity $M$, the residual is:
$R_M(t)=M(t)-M(0)-\int_0^t\left[\dot{M}_{\text {in }}(s)-\dot{M}_{\text {out }}(s)+\dot{M}_{\text {source }}(s)\right] \mathrm{d} s$
The residual measures the difference between the change in storage and the integrated net transfer. All terms must use consistent units, with ${\dot{M}}_\text{source}$ representing net internal production or removal. For the CO$_2$ balance, the stored quantity is the equivalent CO$_2$ volume Vc under the adopted reference conditions. For the PM$_{2.5}$ balance, it is the particle mass $V$$p_\text{m}$. Residuals are interpreted relative to an explicitly stated storage or transfer scale.
Under constant occupancy, outdoor concentration, and effective outdoor-air exchange, the CO$_2$ balance has the analytical solution:
for \(Q_{\text{f}}>0\). The equilibrium concentration is \(c_{\text{o}}+G_{\mathrm{CO_2}}/Q_{\text{f}}\), and the response time is \(V/Q_{\text{f}}\). These expressions connect the concentration trajectory directly to occupancy and ventilation. When \(Q_{\text{f}}=0\), the corresponding limit is \(c(t)=c(0)+G_{\mathrm{CO_2}}t/V\).
A similar reference is available for PM$_{2.5}$ under constant conditions. Define:
$a_{\mathrm{p}}=\frac{Q_{\mathrm{oa}}+Q_{\mathrm{inf}}+\eta_{\mathrm{r}} Q_{\mathrm{r}}}{V}+k_{\mathrm{d}}$,
$b_{\mathrm{p}}=\frac{\left[\left(1-\eta_{\mathrm{f}}\right) Q_{\mathrm{oa}}+p_{\mathrm{inf}} Q_{\mathrm{inf}}\right] p_{\mathrm{m}, \mathrm{o}}+S_{\mathrm{m}}}{V}$
Eq. (3) then gives:
for $a_\text{p}>$0. Here, $a_\text{p}$ is the total first-order removal coefficient and $b_\text{p}$ is the concentration source rate. This form distinguishes particle entry from removal by outdoor-air exchange, recirculation filtration, and deposition. Eqs. (8) and (9) provide exact reference trajectories when the corresponding source terms, boundary conditions, airflow rates, and removal coefficients remain constant. For piecewise-constant operation, each segment uses the terminal concentration of the preceding segment as its initial condition.
The policy comparison reports cabin temperature, relative humidity, CO$_2$ concentration, PM$_{2.5}$ concentration, and cumulative HVAC electricity use over the same operating interval. Temperature and relative humidity characterize the mean thermal and moisture response, while pollutant concentrations describe CO$_2$ accumulation and particle loading. Reporting these trajectories alongside electricity use reveals the environmental conditions associated with each policy’s energy demand.
For a trajectory $y_i$ and a corresponding reference $y_i^\text{ref}$, their difference at sampling instant $i$ is:
The reference may be an analytical solution for numerical assessment or another ventilation policy for an operating comparison. Its origin determines the meaning of the difference. Mean absolute difference, MAD; root-mean-square difference, RMSD; and mean signed difference, $\bar{e}$ are defined as:
$\mathrm{MAD}=\frac{1}{N} \sum_{i=1}^N\left|e_i\right|$
$\mathrm{RMSD}=\sqrt{\frac{1}{N} \sum_{i=1}^N e_i^2}$
$\bar{e}=\frac{1}{N} \sum_{i=1}^N e_i$
where, $N$ is the number of paired samples. Comparison with an analytical solution quantifies numerical discrepancy, whereas comparison between ventilation policies quantifies differences in their predicted responses. The perturbed points in figures are illustrative and are excluded from accuracy assessment.
For cumulative HVAC electricity use, the relative difference between a policy a and a reference policy is:
A negative value indicates lower electricity use than the reference policy. All policies are evaluated over the same duration and use the same definitions of included electrical loads. Airflow rates and fan-power settings may differ as prescribed by each policy. The resulting energy difference therefore characterizes the complete operating configuration. Final concentrations and transient responses are reported alongside energy use.
Parameter sensitivity examines how the calculated outcomes depend on the assumptions entering the model. For a selected scalar output $Y$,
where, $\mathcal{M}$ denotes the simulation and output calculation over duration $\tau$. Relevant outputs include final or peak pollutant concentrations, cabin temperature, and cumulative HVAC electricity use. The sensitivity illustration in Section 4 focuses on electricity use at the end of the 30-min summer case.
For a positive continuous parameter $\theta_j$ and nonzero baseline output $Y_0$, the normalized local sensitivity is:
where, $\mathbf{e}_j$ is the unit vector selecting parameter $j$, and $\delta=$ 0.10 specifies the nominal ±10% perturbation for continuous parameters. Each parameter is varied separately while all other inputs retain their baseline values. Both perturbed values must remain within the parameter’s admissible range. Positive sensitivity indicates that increasing the parameter raises the selected output; negative sensitivity indicates the opposite. The coefficient describes the local response around the stated operating point.
Occupant count is treated as a discrete input. Comparisons between integer occupant counts change sensible heat release, moisture production, and CO$_2$ generation consistently. A fractional perturbation of a per-person source rate instead represents variation in occupant activity or source strength and is identified by the specific parameter being varied.
Trajectory comparisons quantify the differences between ventilation policies, while local sensitivity coefficients describe their dependence on selected model inputs. The combined assessment links electricity use and cabin conditions to outdoor-air exchange, filtration, deposition, and thermal loads under the prescribed scenario.
4. Simulated Scenarios and Coupled Trade-Offs
Three ventilation policies are compared over a prescribed 30-min summer scenario with four occupants, a warm outdoor environment, and elevated outdoor PM$_{2.5}$. The comparison examines how outdoor-air exchange affects CO$_2$ dilution, particle ingress, cooling demand, and cumulative HVAC electricity use.
The outdoor temperature is held at 35 ℃, with CO$_2$ and PM$_{2.5}$ concentrations of 430 ppm and 80 μg/m$^3$, respectively. The effective cabin volume is 3 m$^3$. All policies start from a cabin temperature of 35 ℃, CO$_2$ concentration of 430 ppm, PM$_{2.5}$ concentration of 8.0 μg/m$^3$, and relative humidity of 35.3%. The effective thermal capacitance is 130 kJ$\cdot$K$^{-1}$, the envelope conductance is 40 W$\cdot$K$^{-1}$, and the solar heat gain is 450 W. Table 2 summarizes the reported scenario settings.
Quantity | Value or Specification |
|---|---|
Simulation duration | 30 min |
Integration step | 1.2 s |
Cabin volume, $V$ | 3 m$^3$ |
Occupancy, $n$ | Four occupants |
Outdoor temperature, $T_\text{o}$ | 35 ℃ |
Outdoor CO$_2$ concentration | 430 ppm |
Outdoor PM$_{2.5}$ concentration | 80 μg/m$^3$ |
Initial cabin temperature | 35 ℃ |
Initial cabin CO$_2$ concentration | 430 ppm |
Initial cabin PM$_{2.5}$ concentration | 8.0 μg/m$^3$ |
Initial cabin relative humidity | 35.3% |
Effective thermal capacitance, $C_{th}$ | 130 kJ$\cdot$K$^{-1}$ |
Envelope conductance, $UA$ | 40 W$\cdot$K$^{-1}$ |
Solar heat gain, ${\dot{Q}}_\text{sol}$ | 450 W |
Effective outdoor-air exchange in fresh-air mode | 120 m$^3 \cdot$h$^{-1}$ |
Effective outdoor-air exchange in recirculation mode | 8 m$^3$$\cdot$h$^{-1}$, attributed to leakage |
Filtered recirculation flow | 45 m$^3$$\cdot$h$^{-1}$ |
Prescribed PM inlet penetration coefficient | 0.45 |
Recirculation filter efficiency | 0.70 |
Particle deposition rate | 1.2 h$^{-1}$ |
Maximum sensible cooling rate | 2,500 W |
Cooling coefficient of performance | 2.8 |
Fan electrical power | Fresh air: 95 W; recirculation: 65 W |
CO$_2$ switching thresholds | Fresh air above 1,100 ppm; recirculation below 850 ppm |
Continuous fresh-air operation, denoted F, maintains a prescribed effective outdoor-air exchange of 120 m$^3\cdot$h$^{-1}$. High-recirculation operation, denoted R, combines a filtered recirculation flow of 45 m$^3\cdot$h$^{-1}$ with an effective outdoor-air exchange of 8 m$^3\cdot$h$^{-1}$ attributed to leakage. Fan electrical power is 95 W in mode F and 65 W in mode R. The comparison therefore evaluates complete operating configurations with different airflow and fan-power settings.
The rule-based policy, denoted B, starts in mode R and switches to mode F when cabin CO$_2$ exceeds 1,100 ppm. It returns to mode R when CO$_2$ falls below 850 ppm and retains the current mode between these thresholds. The thresholds define the hysteresis rule adopted for this scenario.
Sensible cooling is prescribed by:
$\dot{Q}_{\text {cool }}(t)=\min \left\{\dot{Q}_{\text {max }}, K_T \max \left[T_\text{℃}(t)-T_{\text {set }}, 0\right]\right\}$
The cooling-capacity limit is ${\dot{Q}}_\text{max}=$ 2500 W, the proportional gain is $K_T=$ 500 W$\cdot$K$^{-1}$, and the temperature reference is $T_\text{set}=$ 24 ℃. Here, $T_\text{℃}$ is cabin temperature expressed in degrees Celsius, with $T_\text{℃}$ = $T$$-$273.15 when the temperature state is expressed in kelvin. The sensible-conditioning input in Eq. (1) is ${\dot{Q}}_\text{HVAC,s}=-{\dot{Q}}_\text{cool}$. The sensible-cooling electrical contribution is ${\dot{Q}}_\text{cool}$/2.8, and the prescribed fan power replaces the pressure–flow term in Eq. (4). The 24 ℃ reference determines the proportional cooling response; continuing thermal loads can produce an equilibrium cabin temperature above this value.
Cabin conditions are compared through their transient trajectories and the values reported at five-minute intervals. For a concentration $C$($t$), the time average over the operating duration $\tau$ is defined as:
The time average incorporates the concentration trajectory over the full interval and differs from the final concentration. It characterizes the mean cabin concentration rather than an inhaled dose.
With power expressed in watts and time in seconds, cumulative HVAC electricity use is:
Applying Eq. (11) with continuous fresh-air operation as the reference gives:
Negative values indicate lower electricity use. This comparison is interpreted alongside pollutant concentrations and thermal conditions, since the ventilation policies do not produce equivalent cabin environments.
Figure 3 shows the simulated cabin-temperature response together with reference points generated by perturbing model output. These points provide a graphical illustration of the trajectory and are not independent measurements. The temperature response is reported alongside pollutant concentrations, relative humidity, and cumulative electricity use in Table 3, allowing the consequences of each ventilation policy to be compared over the same interval.

Mode | Time (min) | Temperature (°C) | CO2 (ppm) | PM2.5 (µg/m3) | Relative Humidity (%) | Electricity Use (kWh) |
|---|---|---|---|---|---|---|
Fresh air | 0 | 35.0 | 430 | 8.0 | 35.3 | 0.000 |
Fresh air | 5 | 31.3 | 1,009 | 34.1 | 43.5 | 0.082 |
Fresh air | 10 | 28.3 | 1,029 | 34.9 | 51.7 | 0.163 |
Fresh air | 15 | 27.2 | 1,030 | 35.0 | 55.2 | 0.226 |
Fresh air | 20 | 26.9 | 1,030 | 35.0 | 56.2 | 0.279 |
Fresh air | 25 | 26.8 | 1,030 | 35.0 | 56.5 | 0.329 |
Fresh air | 30 | 26.8 | 1,030 | 35.0 | 56.5 | 0.379 |
Recirculation | 0 | 35.0 | 430 | 8.0 | 35.3 | 0.000 |
Recirculation | 5 | 31.2 | 2,224 | 7.1 | 44.0 | 0.080 |
Recirculation | 10 | 27.9 | 3,661 | 6.8 | 53.2 | 0.156 |
Recirculation | 15 | 26.7 | 4,811 | 6.7 | 57.2 | 0.209 |
Recirculation | 20 | 26.4 | 5,731 | 6.7 | 58.6 | 0.252 |
Recirculation | 25 | 26.3 | 6,469 | 6.7 | 59.2 | 0.292 |
Recirculation | 30 | 26.3 | 7,059 | 6.7 | 59.4 | 0.332 |
Rule-based | 0 | 35.0 | 430 | 8.0 | 35.3 | 0.000 |
Rule-based | 5 | 31.3 | 1,038 | 32.1 | 43.5 | 0.081 |
Rule-based | 10 | 28.3 | 1,030 | 34.9 | 51.7 | 0.162 |
Rule-based | 15 | 27.2 | 1,030 | 34.9 | 55.2 | 0.225 |
Rule-based | 20 | 26.9 | 1,030 | 35.0 | 56.2 | 0.278 |
Rule-based | 25 | 26.8 | 1,030 | 35.0 | 56.5 | 0.328 |
Rule-based | 30 | 26.8 | 1,030 | 35.0 | 56.5 | 0.378 |
Cabin temperature decreased from 35.0 ℃ under all three policies. After five minutes, the fresh-air and recirculation configurations reached 31.3 and 31.2 ℃, respectively. Their final temperatures were 26.8 and 26.3 ℃. Both remained above the 24 ℃ reference of the proportional cooling law because a positive temperature difference was required to sustain cooling against the continuing thermal loads.
The CO$_2$ responses differed substantially. Under fresh-air operation, the concentration reached 1,009 ppm after five minutes and approached approximately 1,030 ppm thereafter. Under high recirculation, it increased to 2,224 ppm after five minutes and 7,059 ppm after 30 min. The lower outdoor-air exchange in mode R restricted dilution of the occupant-generated source, producing continued accumulation over the simulated interval.
PM$_{2.5}$ followed the opposite pattern. Fresh-air operation increased cabin concentration from 8.0 to 34.1 μg/m$^3$ within five minutes and to approximately 35.0 μg/m$^3$ at 30 min. Under recirculation, concentration decreased to approximately 6.7 μg/m$^3$ as filtration and deposition removed particles while limited outdoor-air exchange restricted their entry. Based on the rounded values in Table 3, the final recirculation concentration was approximately 80.9% lower than the fresh-air value. This percentage compares endpoint concentrations, rather than time-averaged concentrations or inhaled doses.
Figure 4 presents these contrasting pollutant trajectories. Greater outdoor-air exchange diluted internally generated CO$_2$ but increased particle ingress from the polluted surroundings, consistent with the ventilation-dependent behavior reported in previous vehicle studies [4], [5]. The magnitude of the response here depends on the prescribed airflow, penetration, filtration, and deposition parameters.

Relative humidity increased as the cabin cooled, reaching 56.5% under fresh-air operation and 59.4% under recirculation. Relative humidity depends on both temperature and water-vapour content; its higher value under recirculation alone does not establish a larger increase in humidity ratio.
Cumulative electricity use reached 0.379 kWh under fresh-air operation and 0.332 kWh under recirculation at 30 min. Based on the rounded table values, the difference was 0.047 kWh, equivalent to approximately 12.4% of the fresh-air total. This difference reflects both the lower outdoor-air conditioning load and the lower prescribed fan power in mode R.
Over 30 min, the prescribed fan powers contributed 0.0475 kWh in mode F and 0.0325 kWh in mode R. Their difference of 0.015 kWh accounted for approximately 32% of the reported total energy gap. The remaining difference arose from the conditioning contribution included in the energy calculation. The comparison therefore concerns the complete operating configurations, rather than recirculation fraction alone. Figure 5 presents the temperature and cumulative electricity-use trajectories.

Policy B approached the response of continuous fresh-air operation after the initial switching period. CO$_2$ was 1,038 ppm after five minutes and approximately 1,030 ppm at 30 min. Final temperature, PM$_{2.5}$ concentration, and relative humidity matched mode F at the precision reported in Table 3. Electricity use was 0.378 kWh, approximately 0.3% below mode F based on the rounded totals.
This behavior followed from the position of the switching thresholds relative to the fresh-air CO$_2$ equilibrium. After switching to mode F, the concentration approached approximately 1,030 ppm, above the 850 ppm threshold required to restore recirculation. The controller consequently retained fresh-air operation. The selected rule prevented the continued CO$_2$ accumulation observed in mode R, while its return threshold limited subsequent use of recirculation.
Figure 6 presents the local sensitivity illustration for HVAC electricity use at 30 min around the fresh-air baseline. Eq. (13) defines the normalized response to individual continuous-parameter perturbations. Unlike the ventilation-policy comparison, which changes several operating settings together, this analysis examines the response to individual input changes around a common baseline.

5. Discussion and Engineering Implications
The summer comparison demonstrates that the energy performance of a ventilation configuration must be interpreted alongside the cabin conditions it produces. Recirculation reduced electricity use and particle concentration but restricted the removal of occupant-generated CO$_2$. Fresh-air operation produced the opposite balance. The engineering question is therefore which configuration satisfies the required environmental conditions at an acceptable electrical demand.
The reported 12.4% energy difference includes changes in both outdoor-air exchange and fan power. The prescribed fan-power difference contributed 0.015 kWh over 30 min, accounting for approximately 32% of the total gap. This decomposition shows why conditioning and air-circulation power should be reported separately, consistent with research examining the contribution of blower operation to vehicle HVAC energy consumption [22].
The configurations also ended at different temperatures and pollutant concentrations. Their energy difference consequently describes the response to two sets of operating inputs, rather than the energy required to maintain equivalent cabin conditions. A comparison at equivalent environmental performance would require common thermal and air-quality targets, with each configuration adjusted to meet them. Reporting the state trajectories together with electricity use makes this distinction explicit.
The direction of the response is consistent with previous studies linking recirculation to changes in particle concentration, CO$_2$ accumulation, and conditioning demand [4], [5], [6], [7]. The contribution here is the separation of these effects into identifiable airflow, filtration, thermal-load, and electrical-demand terms. This structure explains the calculated energy difference without assigning it entirely to recirculation fraction.
The opposing pollutant responses arise from different source and removal mechanisms. CO$_2$ is generated inside the cabin and removed through outdoor-air exchange. Recirculation filtration provides no corresponding CO$_2$ removal mechanism. For particles, recirculation provides repeated filtration while reducing the intake of polluted outdoor air. These mechanisms explain why a policy selected solely to reduce PM$_{2.5}$ can permit substantial CO$_2$ accumulation, as also observed in experimental vehicle-cabin research [5].
The rule-based result illustrates how switching thresholds interact with ventilation capacity. Once policy B entered fresh-air mode, CO$_2$ approached an equilibrium above the threshold required to restore recirculation. Fresh-air operation therefore persisted, and the thermal, particle, and energy responses approached those of policy F.
From Eq. (8), the fresh-air equilibrium under constant occupancy and boundary conditions is:
$c_{\infty, \mathrm{F}}=c_{\mathrm{o}}+\frac{n g_{\mathrm{CO}_2}}{Q_{\mathrm{f}, \mathrm{~F}}}$
When the concentration at entry to fresh-air mode exceeds this equilibrium, it approaches the equilibrium from above. A return threshold below that value cannot be reached through continued fresh-air operation under the same conditions. In this comparison, the equilibrium concentration and switching thresholds are expressed in ppm.
For the selected threshold pair, the controller retained fresh-air operation after the initial transition. This result connects threshold selection directly to occupant generation, outdoor CO$_2$ concentration, and available outdoor-air exchange. Related studies have incorporated CO$_2$-responsive ventilation and humidity-dependent recirculation into vehicle HVAC control [19], [20]. The present balance model provides a compact explanation of the equilibrium condition governing the switching response.
The four-state formulation connects outdoor-air exchange with cabin heat, moisture, and CO$_2$ transport, while the particle balance additionally represents penetration, filtration, and deposition. The electrical-demand calculation then relates conditioning and air circulation to energy use. Keeping these contributions separate allows an operating change to be followed through the model without counting the same physical exchange twice.
The single-zone representation assumes that cabin mixing occurs sufficiently rapidly relative to the environmental response being examined. A timescale ratio provides a useful criterion:
where, $\tau_\text{mix}$ is the characteristic mixing time and $\tau_\text{env}$ is the shortest environmental response time relevant to the analysis. A small ratio is consistent with representing the cabin through well-mixed states. Here, the ratio defines an applicability criterion rather than a measured characteristic of the simulated cabin.
Mean cabin conditions and local occupant conditions require different spatial descriptions. Supply jets, asymmetric radiation, surface temperatures, and breathing-zone gradients can produce local conditions that differ from the cabin mean. Two-zone thermal models [9] and coupled computational fluid dynamics–thermoregulation models [19] provide additional spatial information. Thermal-comfort models and equivalent-temperature methods also incorporate quantities beyond mean air temperature [10], [11], [12]. The present temperature and humidity states characterize bulk cabin conditions and conditioning loads.
Humidity interpretation requires similar care. Relative humidity changes with both temperature and water-vapour content, so its increase during cooling does not by itself quantify moisture accumulation. Windscreen condensation further depends on the interior glass temperature relative to the cabin-air dew point. A mean-air calculation therefore requires a glass-temperature input or an additional thermal submodel to assess condensation potential.
Numerical verification and physical validation provide different forms of evidence [23]. Analytical solutions, step-size comparisons, and conservation residuals provide numerical verification diagnostics, whereas independent measurements provide evidence for physical validation [23]. Independent measurements assess agreement with a physical cabin. The perturbed model outputs in Figure 3 serve as a graphical illustration and provide neither an independent numerical check nor experimental validation. The values reported in Section 4 are model outputs for the prescribed scenario.
The framework supports HVAC assessment by connecting electricity use with the environmental consequences of each ventilation policy. Separating conditioning and fan power identifies the sources of an energy difference, while the concentration trajectories show how that difference accompanies changes in CO$_2$ dilution and particle removal. The switching analysis further explains how controller settings determine the time spent in each operating mode.
Applying the formulation to a specific vehicle requires a consistent set of cabin, airflow, filtration, and equipment parameters. Cabin volume, effective thermal capacitance, envelope conductance, leakage, filter performance, and HVAC efficiency define the physical system. Outdoor conditions and occupancy define the operating case. Keeping these quantities distinct from the conservation equations allows ventilation policies to be compared using the same underlying vehicle description.
The relevance to sustainable vehicle energy use lies in coordinating conditioning demand with cabin environmental requirements. Reduced outdoor-air intake can lower electrical demand while increasing CO$_2$ accumulation; greater outdoor-air exchange can improve dilution while increasing cooling loads and particle entry under polluted conditions. Evaluating these effects together provides a basis for identifying energy-efficient operating choices within specified thermal and air-quality constraints.
6. Conclusions
A reduced-order dynamic model was developed to couple passenger-car cabin temperature, humidity ratio, CO$_2$ concentration, and PM$_{2.5}$ concentration with HVAC electrical demand. The formulation distinguishes controlled outdoor-air supply, infiltration, and recirculation filtration within a common set of conservation balances. This separation provides a traceable basis for comparing ventilation configurations and explaining their thermal, air-quality, and energy consequences.
In the prescribed 30-min summer scenario with four occupants, high recirculation consumed 0.332 kWh, compared with 0.379 kWh under continuous fresh-air operation. The difference was approximately 12.4%, based on the rounded results. At 30 min, PM$_{2.5}$ concentrations were 6.7 μg/m$^3$ under recirculation and 35.0 μg/m$^3$ under fresh air, whereas CO$_2$ concentrations were approximately 7,059 and 1,030 ppm, respectively. Final cabin temperatures were 26.3 and 26.8 °C. The lower electricity use therefore accompanied different cabin conditions. The prescribed fan-power difference accounted for approximately 32% of the total energy gap, demonstrating the importance of separating air-circulation power from conditioning demand.
The CO$_2$-based switching policy consumed 0.378 kWh and approached the thermal and pollutant outcomes of continuous fresh-air operation. After the initial transition, the fresh-air CO$_2$ equilibrium remained above the return threshold, so the controller retained fresh-air operation. This result identifies the relationship between threshold selection, occupant generation, outdoor concentration, and available ventilation capacity.
The framework connects vehicle HVAC energy efficiency with the cabin conditions produced by each operating policy. Its contribution is the consistent accounting of airflow pathways, pollutant removal, conditioning loads, and electrical demand within a compact dynamic formulation. For the stated scenario, the comparison shows how energy reductions redistribute thermal and pollutant loads, providing a basis for assessing ventilation choices against defined cabin environmental requirements.
Conceptualization, A.C. and A.M.; methodology, A.C.; software, A.M.; validation, A.C., A.M., and R.R.; formal analysis, A.M.; investigation, A.M. and R.R.; resources, A.C.; data curation, R.R.; writing—original draft preparation, A.M.; writing—review and editing, A.C., A.M., and R.R.; visualization, R.R.; supervision, A.C.; project administration, A.C.; funding acquisition, A.C. All authors have read and agreed to the published version of the manuscript.
The data used to support the research findings are available from the corresponding author upon request.
The authors declare no conflicts of interest.
