Overview
Analysis
Solutions
Complete
·Feb 17, 2026
The Core Insight

The hardware is the bottleneck, not the software — 60-80% of the efficiency gap is caused by subsystems designed for the wrong duty cycle

  • The five-whys analysis reveals that the vehicle's parasitic drives, DC-DC converter, and belt drive were each independently optimised for highway cruise.
  • At urban operating points, these three subsystems multiply destructively: 0.75 × 0.85 × 0.85 = 54% urban drivetrain efficiency versus 0.95 × 0.92 × 0.95 = 81% at cruise.
  • This 27-point swing in drivetrain efficiency is the dominant source of the 18% vs 27% gap.
  • No energy management algorithm can make bad hardware efficient — it can only choose among bad options.
  • The correct development priority is to flatten the hardware efficiency curves first, after which any simple EMS achieves near-optimal results.
Viability
Solvable
  • Every intervention draws on proven physics from adjacent domains; the challenge is integration and calibration, not invention.
Key Decision

If you prioritise speed and certainty, deploy Layers 1+2 of the primary stack immediately (0-95 GBP, 6-10 weeks). If you're willing to invest 2,000-3,000 GBP in a university aging test for a potentially transformative result, commission the thermal-idle membrane validation in parallel.

Solution Paths
01NEEDS VALIDATION

Layered Hardware-Software Co-Optimization Stack

Four sequenced interventions (notch EMS → VSD parasitics → PCM thermal jacket → waste heat recovery) targeting each dominant loss mechanism in priority order, at 225-295 GBP and 1.75-2.25 kg total

02NEEDS VALIDATION

Thermal Decoupling: FC as Always-Hot CHP System

Deliberately run FC in thermal-idle during stops (3 g H₂/hour) to eliminate thermal cycling entirely — 22-83x ROI versus stack replacement, but PBI membrane durability under this mode is uncharacterised

Recommendation
  1. If this were my fleet, I'd start two parallel tracks on Monday morning.
  2. Track A is the no-regrets deployment: open the FC enclosure, photograph the blower motor, and export a week of fleet data.
  3. Those two actions — taking maybe 3 hours total — determine whether the VSD retrofit is a 25 GBP add-on or a 100 GBP motor swap, and whether the EMS simulation can proceed immediately or needs 2 weeks of instrumented driving first.
  4. Either way, I'd have the VSD blower running on one vehicle within 3 weeks and the notch EMS firmware deployed within 6 weeks.
  5. Together, those should recover 3-5 efficiency points for under 100 quid per vehicle.
  6. Track B is the higher-ceiling play: I'd email DTU Energy or Jülich and propose the 200-hour thermal-idle aging test.
  7. Budget 2,500 GBP.
  8. This runs in the background while Track A delivers immediate results.
  9. If the aging test comes back clean (degradation ≤40 µV/hour), you've solved HTPEM thermal management for the price of 6 kg of hydrogen over the vehicle's lifetime.
  10. That's a publishable result with real IP value.
  11. What I would NOT do is invest in the custom DC-DC converter (200-400 GBP, 3-6 months) until after measuring the combined effect of VSD + notch EMS.
  12. The DC-DC upgrade is the least cost-effective layer — if Layers 1+2 get you to 22-23%, the remaining gap to 24% might be cheaper to close with the PCM jacket and waste heat recovery than with a custom converter.
  13. I'd also resist the temptation to pursue the EIS plating detection (Concept 10) — the waste heat recovery approach solves the same problem more simply, and the EMI filter design is a rabbit hole that could consume weeks of analog engineering for marginal incremental benefit.
  14. The one thing that genuinely worries me is the Toyota patent.
  15. Before deploying the notch EMS commercially, get an IP attorney to review the specific claims of US10,096,853 B2 against your implementation.
  16. The combination with VSD, PCM, and thermal-idle is likely sufficiently different, but 'likely' isn't good enough for a fleet deployment.

By continuing, you agree to our use of cookies to improve your experience.