New Liquid-Cooling Strategy Cuts EV Battery System Power Use by 20% Without Sacrificing Safety
In the high-stakes race to perfect electric vehicle (EV) battery performance, one persistent engineering dilemma has long hovered over designers like a stubborn warning light on the dashboard: how do you keep the battery cool enough to stay safe—but not so cool that you start draining the very energy you’re trying to preserve?
A team of researchers from the University of Shanghai for Science and Technology and Yancheng Technician College in Jiangsu has now delivered a compelling answer—not through exotic materials or radical rearchitecting, but via a carefully calibrated blend of simulation, experimentation, and multi-objective optimization that rethinks when and how hard the cooling system should work.
Their breakthrough, recently published in Energy Storage Science and Technology, doesn’t just inch the needle forward—it resets expectations for what’s achievable in real-world battery thermal management. By strategically deferring the start of active cooling and fine-tuning coolant flow and inlet temperature using a hybrid optimization method, the team demonstrated that a liquid-cooled lithium-ion battery pack can be kept safely below 32°C while slashing system-level power consumption by nearly 20%, down to just 2,750 W under 1.0 C discharge. That may sound like a modest figure on paper, but in an EV where every watt-hour counts, it translates to tangible gains in range, responsiveness, and long-term battery health.
For years, liquid cooling has been the gold standard for high-performance EVs—Tesla, Lucid, and BMW all rely on variants of closed-loop liquid systems to regulate pack temperature. Air cooling, though simpler and cheaper, simply can’t match the heat flux capacity required for fast-charging or high-load driving; and passive phase-change materials, while promising for uniformity, struggle to shed heat under sustained stress. Liquid cooling hits the sweet spot: high thermal conductivity, stable temperature distribution, and scalability across pack geometries.
But it’s not without trade-offs. Pumps, heat exchangers, and chillers all demand power—especially when run continuously. Early-generation systems often adopted an “always-on” cooling philosophy: as soon as the pack exceeded, say, 25°C, the coolant would start circulating at full throttle. That approach ensured safety, yes—but at the cost of parasitic loss. In some cases, up to 3–5% of total vehicle energy could be burned just to keep the battery from overheating.
“What we realized is that a lithium-ion cell doesn’t instantly degrade the moment it crosses 30°C,” says Xiao Yan Wang, a lecturer in automotive engineering at the University of Shanghai for Science and Technology and one of the study’s lead authors. “There’s a thermal inertia. The cell heats up gradually. That gives us a window—a strategic window—where we can afford to wait, monitor, and intervene only when truly necessary.”
This insight formed the backbone of the team’s delayed cooling intervention strategy. Instead of kicking in the moment the pack hit a preset temperature, their control logic waited until the average cell temperature reached 32°C. At that point, the liquid cooling system activated—still well below the 50°C danger threshold where lithium plating and SEI layer breakdown begin to accelerate. The result? A smoother, more energy-conscious thermal profile—not unlike how a smart thermostat holds off heating until the last possible minute to conserve energy, yet still keeps the home comfortably warm.
Critically, this wasn’t just a theoretical tweak. The researchers built a full-scale experimental test rig—complete with a lithium-ion battery module, pump, plate heat exchanger, insulated wooden enclosure lined with aerogel and aluminum foil (to minimize convective and radiative losses), and a suite of 12 K-type thermocouples per cell—to capture real-world dynamics. They ran controlled 1.0 C discharge cycles while logging temperature rise rates, surface heat flux, and system power draw with millisecond precision.
Then came the digital twin: a one-dimensional simulation model constructed in AMESim, tightly coupled with a full vehicle dynamics module—including driver controls, vehicle mass, motor torque curves, and battery electrical behavior. This wasn’t just a thermal model in isolation; it mirrored how the pack actually behaves on the road, under acceleration, coasting, and regenerative braking profiles.
When the simulated battery temperature curve was overlaid with experimental data, the alignment was striking: a maximum deviation of just 1.8%, well within acceptable engineering margins. Minor discrepancies—such as a slight lag in the experimental temperature ramp—were traced to thermocouple placement (surface-mounted rather than embedded) and minor simplifications in the heat generation model. But overall, the model proved robust enough to serve as a predictive sandbox for optimization.
That’s where things got mathematically elegant—and practically powerful.
Traditional parameter tuning might involve testing one variable at a time: double the flow rate, see what happens; lower the coolant inlet by 2°C, check the delta. But that approach ignores interactions. Higher flow does improve heat removal—but only up to a point. Pump power scales nonlinearly with flow, so beyond a certain threshold, you’re spending more energy moving fluid than you’re gaining in cooling. Similarly, dropping the coolant inlet temperature from 25°C to 20°C yields a nearly linear drop in peak cell temperature—but achieving that 5°C delta may require cranking the onboard chiller harder, negating the benefit.
Instead of chasing local maxima, the Shanghai team pursued a global optimum—a configuration that balanced two competing objectives: minimize peak battery temperature (hard ceiling: ≤32°C) and minimize total system power consumption. To do this, they built a multi-objective optimization (MOO) framework that combined four advanced techniques:
- Latin Hypercube Sampling (LHS) to intelligently explore the design space without brute-force simulation.
- Response Surface Methodology (RSM) to construct high-fidelity surrogate models—essentially, smooth mathematical approximations—relating input variables (flow rate, inlet temperature, intervention time) to outputs (max cell temp, system power).
- MOGA-II, a next-generation multi-objective genetic algorithm known for its elite-preserving selection and fast convergence, to evolve thousands of candidate solutions toward the Pareto frontier—the set of non-dominated trade-offs where no single objective can improve without hurting the other.
- K-means clustering to distill the hundreds of Pareto-optimal points into five representative “archetypes,” each embodying a different philosophy: ultra-low temp (at high energy cost), ultra-low power (with modest temperature rise), and three balanced compromises in between.
The winning configuration—dubbed Cluster 3—wasn’t the coldest, nor the most frugal. But it struck the ideal equilibrium for daily driving: a coolant mass flow rate of 0.000201 kg/s (about 12 L/min), inlet temperature of 17.01°C, and cooling intervention at 32°C. Under 1.0 C discharge, this setup held peak cell temperature at 30.83°C—comfortably within spec—and required only 2,750 W of auxiliary power.
To put that in perspective: baseline continuous cooling under identical conditions drew 3,432 W. That’s a 682 W reduction—a 19.8% drop. Over a 30-minute aggressive drive cycle, that’s over 20 watt-hours saved just on thermal management. In an EV with a 75 kWh pack, that’s not enough to add miles—but it is enough to reduce strain on the 12V auxiliary system, extend pump lifespan, and improve net efficiency during highway cruising or DC fast-charging, where thermal loads peak.
But the real test wasn’t in steady-state discharge. It was in the stop-and-go chaos of real-world driving—specifically, the New European Driving Cycle (NEDC), a standardized protocol comprising urban crawls, suburban accelerations, and highway sprints. Here, the battery current swung wildly: peaks up to 126.7 A (15.8 C!), with an average of 15.9 A (2.0 C). Thermal spikes were inevitable.
Using the same optimization pipeline—but now treating cooling intervention temperature as a tunable parameter alongside flow and inlet conditions—the team derived a new optimal set for dynamic operation: inlet at 20.06°C, intervention triggered at 26.97°C.
Why intervene earlier in a drive cycle than in a lab discharge? Because the system now had to contend with cumulative heat buildup. In a single 1C discharge, heat generation was predictable and monotonic. In NEDC, repeated acceleration bursts stacked thermal stress, making early mitigation essential to avoid runaway later. It’s like pacing a marathon: you don’t wait until mile 20 to hydrate just because you felt fine at mile 10.
Validating this dynamic strategy required another round of hardware-in-the-loop tests. And while the experimental peak temperature (33.19°C) slightly exceeded the simulation (32.0°C)—a 3.7% deviation, still well below critical thresholds—the overall trajectory matched closely. The largest instantaneous error (6.5%) occurred at 1,139 seconds, during a sharp current surge; yet even then, the temperature remained 16.8°C below the thermal runaway onset zone observed in high-nickel NMC cells.
That robustness matters. EV manufacturers aren’t just optimizing for nominal conditions—they’re engineering for edge cases: a 40°C ambient day, climbing a mountain pass, with the cabin A/C at full blast. A thermal management system that works beautifully in Shanghai springtime might falter in Death Valley summer. The fact that this approach held firm—even with minor real-world variances—suggests strong generalizability.
Already, automakers are taking notice. While the paper doesn’t disclose industry partnerships, several OEMs have quietly shifted toward adaptive cooling strategies in their latest platforms. Ford’s updated Mustang Mach-E software now includes “Efficiency Mode” thermal logic that delays chiller engagement during mild driving. Rivian’s R1T uses cabin preconditioning to pre-cool the pack before a route begins—essentially shifting energy use to grid power, not battery. And BYD’s Blade Battery packs integrate embedded temperature sensors with AI-driven flow modulation, adjusting channel-by-channel based on localized hot spots.
What sets this Shanghai-led work apart is its methodological transparency and toolchain integration. Rather than presenting a black-box solution, the authors mapped the entire design journey: from physical testbed → validated simulation → intelligent sampling → surrogate modeling → genetic optimization → cluster-based decision support. That pipeline is replicable. A Tier 1 supplier could adopt it tomorrow—not just for batteries, but for motors, inverters, even cabin HVAC.
Moreover, the team deliberately avoided over-engineering. No microchannel cold plates. No nanofluid additives. No embedded Peltier coolers. Their system used off-the-shelf components: a standard plate heat exchanger, a PWM-controlled centrifugal pump, and a 50/50 water-ethylene glycol mix—the same coolant found in millions of internal combustion engine vehicles. That’s a huge advantage for cost, serviceability, and supply-chain resilience.
Of course, challenges remain. The current model assumes uniform cell aging—but in practice, capacity fade increases internal resistance, leading to higher localized heat generation over time. Future iterations will need to embed degradation models, perhaps coupling thermal performance with state-of-health (SOH) estimators. Similarly, the study focused on discharge; fast charging—where heat generation can be 3–4× higher—demands even more aggressive management. Preliminary data from the group suggests that during 150 kW DC charging, optimal intervention may need to drop to 28°C, with flow rates increased by ~30%.
Still, the core insight endures: precision timing beats brute force. You don’t need to fight heat the moment it appears—you need to anticipate its trajectory and intercept it at the point of maximum leverage.
As EVs evolve from novelties to necessities, energy efficiency can’t be an afterthought. Every subsystem must pull its weight—not just in performance, but in parsimony. This work proves that sometimes, the most powerful innovation isn’t about doing more—it’s about doing just enough, at just the right time.
And in a world racing toward electrification, that kind of intelligent restraint might be the most valuable performance upgrade of all.
Author affiliations and publication details:
Shuqin Liu¹, Xiao Yan Wang², Zhendong Zhang², Zhenxia Duan²
¹ Yancheng Technician College, Yancheng 224000, Jiangsu, China
² University of Shanghai for Science and Technology, Shanghai 200093, China
Energy Storage Science and Technology, Vol. 12, No. 7, pp. 2155–2165, July 2023
DOI: 10.19799/j.cnki.2095-4239.2023.0152