I focus on the temperature–energy consumption synergy of an electric vehicle battery pack because the thermal state of the pack directly controls safety, available power, charging capability, and driving range. In my work, I treat the electric vehicle battery pack not as a single isolated component but as a coupled electrochemical, thermal, fluid, and control system. The electric vehicle battery pack must remain within a favorable temperature window, usually near 25–40 °C, while temperature non-uniformity should remain below about 5 °C. When the electric vehicle battery pack operates above this window, degradation and thermal runaway risks increase. When it operates below this window, internal resistance rises, available discharge power falls, and heating energy consumption increases. Therefore, I design a layered indirect liquid-cooling thermal management system and evaluate it under low-, normal-, and high-temperature environments. The main objective is to keep the electric vehicle battery pack within safe thermal limits while reducing the energy consumed by pumps, fans, compressors, and heaters.
Motivation and scope. I begin from the observation that conventional threshold-based thermal management often over-cools or under-cools the electric vehicle battery pack. A simple on–off rule may keep the maximum temperature below a limit, but it frequently causes high compressor power, frequent actuator switching, and unnecessary fan or pump operation. My research therefore integrates theoretical modeling, experimental calibration, design of experiments, nonlinear model predictive control, and a grey wolf optimizer. I evaluate the proposed strategy by comparing it with threshold control and PID control. The comparison considers both battery temperature tracking and total thermal management energy consumption. I also quantify which actuators dominate the behavior of the electric vehicle battery pack through optimal Latin hypercube sampling and analysis of variance.
Theoretical modeling of the electric vehicle battery pack. A lithium-ion electric vehicle battery pack mainly consists of positive electrodes, negative electrodes, electrolyte, separator, current collectors, and packaging. In my model, I use a lithium iron phosphate electric vehicle battery pack because of its thermal stability and cycle life. During charging, lithium ions move from the positive electrode to the negative electrode through the electrolyte and separator, while electrons move through the external circuit. During discharging, the process reverses. The reversible electrochemical reactions are represented in general form as:
$$LiFePO_4 – xLi^+ – xe^- \rightleftharpoons xFePO_4 + (1-x)LiFePO_4$$
$$6C + xLi^+ + xe^- \rightleftharpoons Li_xC_6$$
$$LiFePO_4 + 6C \rightleftharpoons xFePO_4 + (1-x)LiFePO_4 + Li_xC_6$$
Heat generation in the electric vehicle battery pack includes reaction heat, polarization heat, Joule heat, and side-reaction heat. I neglect side-reaction heat because it is small under normal operating conditions. Thus, the total heat generation rate is expressed as:
$$Q=Q_r+Q_p+Q_j+Q_s$$
For control-oriented simulation, I use the Bernardi equation to calculate the heat generation power of the electric vehicle battery pack:
$$\dot Q_{gen}=I\left(IR_e-T_b\frac{dU_{op}}{dT_b}\right)$$
where \(I\) is current, \(R_e\) is equivalent internal resistance, \(T_b\) is battery temperature, \(U_{op}\) is open-circuit voltage, and \(dU_{op}/dT_b\) is the entropic coefficient. This equation captures both irreversible Joule heating and reversible entropic heating or cooling. It is particularly useful because I can update \(R_e\) and \(U_{op}\) from experimental data for the electric vehicle battery pack.
Heat transfer inside and around the electric vehicle battery pack includes conduction, convection, and radiation. I use Fourier’s law for conduction:
$$Q_f=-\lambda\frac{\partial T}{\partial x}$$
I use Newton’s law of cooling for convection between the electric vehicle battery pack and the coolant:
$$Q_a=h(T_b-T_c)$$
I use the Stefan–Boltzmann law for radiation:
$$\Phi=\varepsilon A\sigma T^4$$
For the liquid-cooled plate, I calculate the equivalent convective heat transfer coefficient from the Nusselt number:
$$h=\frac{Nu\lambda}{d}$$
$$Nu=0.023Re^{0.8}Pr^{0.3}$$
$$Re=\frac{ud}{\mu}$$
Here, \(u\) is coolant velocity, \(d\) is hydraulic diameter, \(\lambda\) is coolant thermal conductivity, and \(\mu\) is kinematic viscosity. These relations connect pump speed to the thermal resistance between the coolant and the electric vehicle battery pack. I use a lumped-parameter thermal model because it is computationally efficient and suitable for real-time control. The lumped model assumes that the electric vehicle battery pack has a nearly uniform temperature, which is acceptable for system-level thermal management design.
Thermal load calculation. I calculate the cooling and heating requirements of the electric vehicle battery pack before sizing the thermal management system. For a high-temperature cruise case, I assume the vehicle travels at 120 km/h for 1000 s, starting from 40 °C and cooling toward 30 °C. The total heat generation of the electric vehicle battery pack is estimated as:
$$\dot Q_{gen}=n_p n_s \left(\frac{P_{driv}}{n_p\eta_b U}\right)^2 R$$
Using the parameters in my model, the generated heat is approximately 2474 W, and the required cooling power is approximately 3457 W. For a low-temperature case, I assume the electric vehicle battery pack must be heated from -10 °C to 20 °C within 1500 s. The required heating power is:
$$Q_h=\frac{c_{p,b}(T_u-T_d)}{t}$$
This yields a heating demand of about 1886 W. These values define the boundary conditions for component selection and controller design.
| Parameter | Value | Unit |
|---|---|---|
| Vehicle mass | 1445 | kg |
| Rolling resistance coefficient | 0.01 | – |
| Aerodynamic drag coefficient | 0.31 | – |
| Frontal area | 2.09 | m² |
| Rotational inertia factor | 1.4 | – |
| Battery nominal voltage | 400 | V |
| Battery equivalent resistance | 0.03 | Ω |
| Battery specific heat capacity | 1250 | J/(kg·°C) |
Powertrain and thermal management modeling. I build the electric vehicle powertrain and the thermal management system in a one-dimensional simulation environment. The vehicle model includes longitudinal dynamics, driver behavior, motor, motor controller, inverter, and the electric vehicle battery pack. The longitudinal equation is:
$$m_v\dot v=F_t-F_r-F_a-F_g$$
where \(F_t\) is traction force, \(F_r\) is rolling resistance, \(F_a\) is aerodynamic drag, and \(F_g\) is grade resistance. The driver model tracks a target speed by generating accelerator and brake commands. The motor controller distributes torque requests and regenerative braking commands. The electric vehicle battery pack supplies power to both the drivetrain and the thermal management system. This coupling is important because the thermal management energy consumption reduces the available range of the electric vehicle battery pack.
My thermal management architecture has three main loops. The coolant loop contains a pump, a liquid-cooled plate, a chiller, a three-way valve, and a positive temperature coefficient heater. The battery heat dissipation loop contains a radiator and a fan. The air-conditioning loop contains a compressor, condenser, expansion valve, and chiller. The electric vehicle battery pack rejects heat to the coolant through the cold plate. The coolant then rejects heat either to ambient air through the radiator or to the refrigerant through the chiller. In cold conditions, the positive temperature coefficient heater warms the coolant, and the pump delivers heat to the electric vehicle battery pack.
| Loop | Main components | Primary function |
|---|---|---|
| Coolant loop | Pump, cold plate, chiller, three-way valve, PTC heater | Transport heat between electric vehicle battery pack and heat exchangers |
| Battery heat dissipation loop | Radiator, fan | Reject heat to ambient air when possible |
| Air-conditioning loop | Compressor, condenser, expansion valve, chiller | Provide active cooling for electric vehicle battery pack |
Model validation and parameter calibration. I use an equivalent circuit model with a single internal resistance, known as the Rint model, to describe the electrical behavior of the electric vehicle battery pack. The terminal voltage is:
$$U_b=U_{op}-I_bR_e$$
The battery power is:
$$P_b=I_bU_b$$
The state of charge is updated by current integration:
$$SOC=SOC_0-\frac{\int_0^t I\,dt}{C_b}$$
I calibrate open-circuit voltage, internal resistance, and thermal properties through experiments. I perform open-circuit voltage tests at low current, hybrid pulse power characterization tests for internal resistance, and heat-flux-based measurements for equivalent heat capacity. The equivalent heat capacity is computed as:
$$C_b=\frac{q\Delta t}{T_1-T_0}$$
From the thermal property test, the electric vehicle battery pack cell equivalent heat capacity is about 1142 J/K. The internal resistance surface shows that resistance increases significantly at low temperature and low state of charge. This behavior means that the electric vehicle battery pack generates more heat during cold low-SOC operation, which increases the demand on the thermal management system.
| Test | Purpose | Key result |
|---|---|---|
| Open-circuit voltage test | Map \(U_{op}\) versus SOC | Voltage decreases as SOC decreases |
| HPPC test | Identify equivalent internal resistance | Resistance rises at low temperature and low SOC |
| Thermal property test | Measure equivalent heat capacity | About 1142 J/K per cell |
| Battery temperature validation | Compare simulation and experiment | Temperature error below 3 °C |
I also calibrate the condenser and chiller. The condenser calibration uses airflow speed, air temperature, refrigerant mass flow, inlet temperature, and inlet pressure. The chiller calibration uses coolant flow, coolant temperature, refrigerant mass flow, inlet temperature, and inlet pressure. I import the experimental data into the calibration module and modify the Nusselt correlation coefficients until the simulated heat exchange rate matches the measurements. The relative errors of both the condenser and chiller are below 5%. This confirms that the heat exchanger models are reliable enough for control-oriented simulation of the electric vehicle battery pack.
| Component | Calibration variable | Validation metric | Error |
|---|---|---|---|
| Condenser | Air-side Nusselt correlation | Heat exchange rate | < 5% |
| Chiller | Refrigerant-side and coolant-side coefficients | Heat exchange rate | < 5% |
| Battery cell | Equivalent circuit and lumped thermal parameters | Temperature and voltage response | < 3 °C temperature error |
Key factor screening with optimal Latin hypercube sampling. The electric vehicle battery pack thermal management system has multiple strongly coupled actuators: compressor speed, pump speed, and fan speed. I first perform single-factor studies. For the compressor, I vary speed from 1000 rpm to 5000 rpm. Higher compressor speed increases cooling capacity, but compressor power rises rapidly, especially above 3000 rpm. The cooling benefit begins to saturate while the energy penalty grows. For the pump, I vary speed from 500 rpm to 2500 rpm. The battery temperature decreases with higher pump speed, but the benefit becomes marginal above 2000 rpm. Pump power increases roughly with the cube of speed. For the fan, I vary speed from 750 rpm to 4500 rpm. The fan has only a small effect on the electric vehicle battery pack temperature, often about 0.1–0.2 °C, while fan power rises from about 3 W to more than 670 W. Therefore, I treat the fan as an auxiliary actuator rather than a primary temperature-control actuator.
| Actuator | Range studied | Temperature effect | Energy effect |
|---|---|---|---|
| Compressor | 1000–5000 rpm | Strong cooling effect | Dominant energy consumer |
| Pump | 500–2500 rpm | Strong coolant-side effect | Moderate to high energy consumption |
| Fan | 750–4500 rpm | Small temperature effect | Rapid nonlinear power growth |
To quantify these effects more rigorously, I use optimal Latin hypercube sampling. Standard Latin hypercube sampling divides each factor range into equal-probability intervals and samples each interval once. The sample coordinate is:
$$X_i^j=a_i+\frac{\epsilon_i^j+U_i^j}{n}(b_i-a_i)$$
Optimal Latin hypercube sampling then optimizes the spatial distribution using a maximum–minimum distance criterion. I define the factor space as:
$$\Omega=\{(N_{comp},N_{pump},N_{fan})\mid 2000\le N_{comp}\le 4000,\;1000\le N_{pump}\le 2000,\;1500\le N_{fan}\le 2250\}$$
I normalize each factor to the unit interval:
$$X_1=\frac{N_{comp}-2000}{2000}$$
$$X_2=\frac{N_{pump}-1000}{1000}$$
$$X_3=\frac{N_{fan}-1500}{750}$$
I generate 30 samples and verify their distribution. The correlation coefficients remain below 0.2, and the minimum Euclidean distance is about 0.195. This gives a statistically independent and spatially uniform design, which is suitable for variance analysis.
| Factor | Lower bound | Upper bound | Unit |
|---|---|---|---|
| Compressor speed | 2000 | 4000 | rpm |
| Pump speed | 1000 | 2000 | rpm |
| Fan speed | 1500 | 2250 | rpm |
I perform analysis of variance on the simulation results. For battery temperature, the model F-value is 152.10 with p < 0.0001. The pump speed has the largest F-value of 237.13, followed by compressor speed with 121.01. The fan speed is not significant, with p = 0.5137. The coefficient of determination is 0.9461, and the adjusted coefficient is 0.9399. For system energy consumption, the model F-value is 1994.38 with p < 0.0001. The compressor speed has an F-value of 5105.98, the pump speed has 93.95, and the fan speed has 47.61. This shows that compressor speed dominates energy consumption, while pump speed strongly affects the temperature of the electric vehicle battery pack. Fan speed mainly adds energy consumption with limited thermal benefit. Therefore, I prioritize compressor and pump coordination in high-temperature control and use the fan as an auxiliary variable in normal-temperature control.
| Response | Factor | F-value | p-value | Conclusion |
|---|---|---|---|---|
| Battery temperature | Compressor speed | 121.01 | < 0.0001 | Highly significant |
| Battery temperature | Pump speed | 237.13 | < 0.0001 | Highly significant |
| Battery temperature | Fan speed | 0.4384 | 0.5137 | Not significant |
| System energy | Compressor speed | 5105.98 | < 0.0001 | Dominant |
| System energy | Pump speed | 93.95 | < 0.0001 | Significant |
| System energy | Fan speed | 47.61 | < 0.0001 | Significant but secondary |
Temperature–energy collaborative control. I divide the thermal management problem into low-temperature, normal-temperature, and high-temperature modes. The mode boundaries are summarized below. In the low-temperature mode, I use a positive temperature coefficient heater and a pump. In the normal-temperature mode, I use the pump and fan. In the high-temperature mode, I use the compressor and pump, with the fan set to a fixed auxiliary speed. The target of my controller is to keep the electric vehicle battery pack within a safe range while minimizing total energy consumption.
| Mode | Ambient temperature range | Main actuators | Control objective |
|---|---|---|---|
| Low-temperature | -20 to 5 °C | PTC heater, pump | Warm electric vehicle battery pack with low heating energy |
| Normal-temperature | 6 to 30 °C | Pump, fan | Prevent overheating and reduce cooling energy |
| High-temperature | 31 to 40 °C | Compressor, pump | Track target temperature precisely and reduce energy |
For the low-temperature mode, I compare threshold control and PID control. The threshold controller turns the positive temperature coefficient heater on when the electric vehicle battery pack temperature falls below a lower threshold and turns it off above an upper threshold. The PID controller adjusts heater power continuously. The objective is to reach and maintain about 5 °C. The threshold controller uses a fixed 4000 W heater power and consumes about 8837 kJ. The PID controller consumes about 7921 kJ. This is a reduction of about 10.37%. Therefore, continuous PID heating is more energy-efficient for the electric vehicle battery pack under the tested low-temperature cycle.
| Low-temperature strategy | Heater behavior | Energy consumption | Comparison |
|---|---|---|---|
| Threshold control | On–off at fixed power | 8837 kJ | Baseline |
| PID control | Continuous power modulation | 7921 kJ | 10.37% lower |
For the normal-temperature mode, I design a nonlinear model predictive controller. The state-space model is derived from energy conservation. The battery temperature dynamics are:
$$c_bm_b\dot T_b=\dot Q_g-\dot Q_d$$
The heat removed by the coolant is:
$$\dot Q_d=h_bA_b(T_b-T_{cb})$$
The coolant temperature near the electric vehicle battery pack evolves as:
$$c_lm_{c,all}\dot T_{cb}=h_bA_b(T_b-T_{cb})-\dot m_c c_l(T_{cb}-T_{cr})$$
The radiator coolant temperature evolves as:
$$c_lm_{cr}\dot T_{cr}=\dot m_c c_l(T_{cb}-T_{cr})-Q_r$$
The radiator heat rejection depends on fan-induced air velocity:
$$Q_r=V_{fan}A_r\rho_{air}(T_{cr}-T_{amb})$$
I express the system in nonlinear state-space form:
$$\dot x(t)=f(x(t),u(t),d(t))$$
$$y(t)=h(x(t),u(t),d(t))$$
where \(x=[T_b,T_{cb},T_{cr}]^T\), \(u=[\dot m_c,V_{air}]^T\), and \(d=[\dot Q_{gen}]\). I verify this predictive model against the high-fidelity simulation model. Under a driving cycle at 25 °C, the maximum temperature difference between the predictive model and the detailed model is below 0.5 °C. This validates the model for real-time control of the electric vehicle battery pack.
I then define the nonlinear model predictive control objective as:
$$J=\min\left\{k_1\sum_{i=1}^{N}(T_b(k+i)-T_{ref}(k+i))^2+k_2P_{comp}(k+i)+k_3P_{pump}(k+i)\right\}$$
subject to battery thermal dynamics and actuator limits. In the normal-temperature mode, the controller coordinates pump and fan speed. I compare threshold control, PID control, and NMPC. Without cooling, the electric vehicle battery pack reaches about 34.54 °C. Threshold control limits the peak to about 28.67 °C but causes overshoot and frequent switching. PID control reaches about 28.72 °C with a larger overshoot and slower recovery after load changes. NMPC keeps the peak near 28.37 °C, tracks the 27 °C target more smoothly, and reduces the maximum temperature error by about 20.35% compared with PID control. In terms of energy, threshold control consumes about 10000.87 kJ, PID control consumes about 8585.67 kJ, and NMPC consumes about 7613.71 kJ. Thus, NMPC reduces energy by about 23.87% compared with threshold control and about 11.32% compared with PID control. This confirms the advantage of using a predictive controller for the electric vehicle battery pack.
| Normal-temperature strategy | Peak temperature | Overshoot behavior | Energy consumption | Energy reduction |
|---|---|---|---|---|
| Threshold control | 28.67 °C | 0.67 °C overshoot | 10000.87 kJ | Baseline |
| PID control | 28.72 °C | 1.72 °C overshoot | 8585.67 kJ | 14.15% vs. threshold |
| NMPC | 28.37 °C | 1.37 °C overshoot | 7613.71 kJ | 23.87% vs. threshold; 11.32% vs. PID |
For the high-temperature mode, I introduce a grey wolf optimizer to solve the nonlinear model predictive control problem. The electric vehicle battery pack temperature and coolant inlet temperature are the states. The compressor speed and pump speed are the control variables. The heat exchange in the chiller is represented as:
$$\dot Q_d=c_{p,c}\dot m_c(T_{c,in}-T_{c,out})$$
The coolant outlet temperature is modeled with a uniform wall temperature assumption:
$$T_{c,out}=(T_{c,in}-T_b)\exp\left(\frac{-h_bA_b}{c_{p,c}\dot m_c}\right)+T_b$$
The battery temperature dynamics become:
$$c_{p,b}m_b\frac{dT_b}{dt}=\dot Q_{gen}+c_{p,c}\dot m_c\left(T_{c,in}-(T_{c,in}-T_b)\exp\left(\frac{-h_bA_b}{c_{p,c}\dot m_c}\right)-T_b\right)$$
The coolant inlet temperature is updated by the chiller cooling capacity:
$$c_{p,c}m_{c,all}\frac{dT_{c,in}}{dt}=c_{p,c}\dot m_c(T_{c,out}-T_{c,in})-Q_c^{bat}(N_{comp})$$
I fit the compressor cooling capacity as a cubic function of compressor speed:
$$Q_c^{bat}=43N_{comp}^3-241N_{comp}^2+414N_{comp}+3939$$
The coefficient of determination for this fit is 0.9978. I then write the high-temperature system in state-space form:
$$\dot x(t)=f(x(t),u(t),d(t))$$
$$y(t)=h(x(t),u(t),d(t))$$
where \(x=[T_b,T_{c,in}]^T\), \(u=[N_{comp},N_{pump}]^T\), and \(d=[\dot Q_{gen}]\). I validate this predictive model under a 35 °C environment. The maximum temperature difference between the predictive model and the detailed simulation is about 0.6 °C. This is acceptable for model predictive control of the electric vehicle battery pack.
I use the grey wolf optimizer to solve the nonlinear optimization problem. The social hierarchy and position update are mathematically represented as:
$$D_\alpha=|C_1X_\alpha(\tau)-X(\tau)|,\quad D_\beta=|C_2X_\beta(\tau)-X(\tau)|,\quad D_\delta=|C_3X_\delta(\tau)-X(\tau)|$$
$$X_1=X_\alpha(\tau)-A_1D_\alpha,\quad X_2=X_\beta(\tau)-A_2D_\beta,\quad X_3=X_\delta(\tau)-A_3D_\delta$$
$$X(\tau+1)=\frac{X_1+X_2+X_3}{3}$$
I initialize the wolf population with 20 individuals and run 20 iterations at each control step. The previous actuator commands are used as initial candidates to improve convergence. The optimizer returns the compressor and pump speeds. Only the first element of the optimal sequence is applied to the electric vehicle battery pack thermal management system, and the optimization repeats at the next sampling instant.
Under high-temperature conditions, I compare threshold control, PID control, and GWO-NMPC. All three strategies reduce the electric vehicle battery pack from 35 °C toward the 27 °C target within about 300 s. Threshold control causes mild overshoot. PID control suppresses low-speed fluctuations but shows oscillatory compressor behavior. GWO-NMPC produces smooth and continuous compressor and pump commands. The total energy consumption of GWO-NMPC is about 51.31% lower than threshold control and about 12.62% lower than PID control. Therefore, the GWO-NMPC strategy provides the best combination of temperature tracking and energy saving for the electric vehicle battery pack in high-temperature operation.
| High-temperature strategy | Temperature behavior | Actuator behavior | Energy comparison |
|---|---|---|---|
| Threshold control | Mild overshoot and oscillation | Frequent on–off switching | Baseline |
| PID control | Stable but oscillatory under load changes | Large speed fluctuations | Higher than GWO-NMPC |
| GWO-NMPC | Smooth tracking near 27 °C | Continuous coordinated control | 51.31% lower than threshold; 12.62% lower than PID |
Integrated interpretation. My results show that temperature and energy consumption cannot be optimized by a single fixed rule. The electric vehicle battery pack has strong nonlinearity, time delay, and coupling among actuators. In low-temperature conditions, the heater power should be modulated continuously rather than switched abruptly. In normal-temperature conditions, the pump and fan should be coordinated by a predictive controller because the fan alone has a weak effect on the electric vehicle battery pack temperature but a strong effect on energy consumption. In high-temperature conditions, the compressor is the dominant energy consumer and the primary cooling actuator. The pump must be adjusted simultaneously because coolant flow changes the heat transfer coefficient and the temperature difference across the chiller. The grey wolf optimizer improves the nonlinear model predictive controller by finding better compressor and pump sequences than a purely gradient-based or rule-based method.
From a design perspective, I conclude that a layered architecture is necessary for the electric vehicle battery pack. The lower layer contains physical components: cold plate, pump, radiator, fan, chiller, compressor, condenser, and heater. The middle layer contains mode selection and actuator constraints. The upper layer contains the predictive controller and optimizer. This structure allows the electric vehicle battery pack to operate safely at low, normal, and high ambient temperatures while reducing unnecessary energy use. It also provides a practical path for implementing advanced control in an electric vehicle battery pack thermal management system.
| Layer | Function | Key variables | Benefit for electric vehicle battery pack |
|---|---|---|---|
| Physical layer | Heat transfer and actuation | Coolant flow, refrigerant flow, air flow | Provides cooling and heating capacity |
| Mode layer | Ambient-temperature mode selection | Low, normal, high temperature | Avoids inappropriate actuator use |
| Optimization layer | Predictive control and energy minimization | Temperature, power, actuator limits | Balances safety and energy consumption |
Experimental and simulation evidence. I validate the electric vehicle battery pack model using open-circuit voltage tests, hybrid pulse power characterization, and thermal property measurements. The temperature response error remains below 3 °C. I validate the condenser and chiller using heat exchanger experiments, and the heat exchange rate error remains below 5%. I validate the predictive model against the detailed simulation model, and the maximum temperature difference remains below 0.5 °C in normal-temperature operation and below 0.6 °C in high-temperature operation. These validation results give confidence that the control-oriented model is accurate enough for the electric vehicle battery pack.
| Validation item | Condition | Maximum error | Interpretation |
|---|---|---|---|
| Battery temperature response | Cell discharge test | < 3 °C | Thermal model is reliable |
| Condenser heat exchange | Air-side and refrigerant-side tests | < 5% | Heat exchanger model is reliable |
| Chiller heat exchange | Coolant-side and refrigerant-side tests | < 5% | Chiller model is reliable |
| Normal-temperature predictive model | 25 °C driving cycle | < 0.5 °C | Suitable for NMPC |
| High-temperature predictive model | 35 °C driving cycle | < 0.6 °C | Suitable for GWO-NMPC |
Concluding remarks. I have developed a temperature–energy collaborative thermal management framework for an electric vehicle battery pack. I modeled the electrochemical and thermal behavior of the electric vehicle battery pack, built a one-dimensional powertrain and thermal system simulation, calibrated key parameters through experiments, screened dominant control factors using optimal Latin hypercube sampling, and designed layered control strategies for low-, normal-, and high-temperature environments. The low-temperature PID strategy reduces heating energy by about 10.37% compared with threshold control. The normal-temperature NMPC reduces cooling energy by about 23.87% compared with threshold control and about 11.32% compared with PID control. The high-temperature GWO-NMPC reduces energy by about 51.31% compared with threshold control and about 12.62% compared with PID control while maintaining the electric vehicle battery pack near 27 °C. These results demonstrate that predictive and optimized control can achieve both thermal safety and energy efficiency for the electric vehicle battery pack under complex operating conditions.
For future work, I plan to extend the model to include internal temperature gradients, thermal contact resistance, and non-uniform coolant flow in the electric vehicle battery pack. I also intend to study parameter uncertainty and ambient disturbances, because real vehicle operation rarely matches nominal conditions. Hardware-in-the-loop testing and reduced-order model predictive control will be important for real-time deployment. By combining high-fidelity modeling, experimental calibration, and optimization-based control, I believe the electric vehicle battery pack can be managed more safely, more efficiently, and with less energy waste across a wide range of climates and driving cycles.