How a battery and a capacitor share one load

A battery sized to survive a load's current peaks is oversized for its energy. A battery sized for the energy is derated by the peaks. Either way something is being paid for twice.

Look at what the transients actually demand. A starter motor wants several hundred amps for about a second, then nothing for an hour. An industrial press draws three times its rated current for two seconds in twelve. A megawatt-class solar plant needs to hold its output steady for thirty seconds while a cloud edge crosses the array. In each case the transient's energy is trivial and its current is not.

That is the opportunity: a store that is small in energy and large in power can take the part of the load that the energy store handles worst.

The battery is Faradaic and bulk. Discharge moves ions into and out of a host lattice, throughout the volume of the electrode. Bulk means diffusion, and diffusion means a rate limit. High specific energy follows, and so does a power ceiling, and so does a life that depends on the mechanical and chemical consequences of all that bulk participation.

The capacitor is non-Faradaic and superficial. Charge separates electrostatically in a layer a few molecules thick at a very high-area carbon surface. No bulk reaction, no phase change, no diffusion limit. Very high power follows, and cycle life in the hundreds of thousands, and low specific energy — twenty to forty times below lithium-ion at cell level, because only a surface is taking part.

Neither device is better. They are answers to different questions, and the useful question is which part of your load belongs to which.

Each figure is the excess power multiplied by the duration it has to be supplied for. They are arithmetic from the stated conditions, not measurements of any product.

This ratio is why capacitor mass and cost are tolerable in a power role and prohibitive in an energy role. It is also, on its own, the commercial argument for the architecture. How the sizing is done.

The ratio is a property of the duty cycle, not of the architecture. Across a wider set of applications the absolute requirement spans four orders of magnitude — from under two watt-hours for a portable imaging bus to several kilowatt-hours for rail braking recovery — and in the largest cases the mass of the bank becomes the deciding term rather than a footnote. The four rows above are the cases where the arithmetic is favourable. They are not all the cases there are.

Peak current and terminal voltage. On a 48 V nominal, 100 Ah pack with 15 mΩ of internal resistance, 500 A of load pulls the terminals down to 43.7 V — inside the window where many inverters and DC-DC inputs drop out. Clamping the battery to 150 A leaves about 49 V. The system fails on voltage well before it fails on energy or temperature, and this is the effect that is largest, most immediate and easiest to verify.

Heat. Resistive dissipation rises with the square of current, so clamping to 30 % of demand reduces the peak instantaneous heating term by about 91 %. Over a complete duty cycle the governing quantity is root-mean-square current rather than peak, and the capacitor's charge returns through the battery — so the duty-cycle reduction is about 40 %, not 91 %.

Both figures are correct and they measure different things. Why the two differ by a factor of 2.3, worked in full.

• Calendar ageing is untouched. The architecture acts on current and excursion. A pack that spends its life sitting at high state of charge in a warm room needs a charge policy, not an architecture.
• Conversion loss is paid twice. Every joule routed through the buffer pays on the way in and on the way out.
• There is a break-even. Above a transient duty fraction of roughly 43 % — with a 90 % round-trip path — a buffer raises battery heating instead of lowering it, because the current has merely moved from one part of the cycle to another. In the lossless limit that break-even sits at exactly one half, independently of how peaky the load is.
• A faster cell is a real alternative. High-power chemistries buy rate capability from the chemistry, with no second store, no converter and no added control law. Where the whole duty cycle is high-rate, that is likely the better answer.
• It does not add range or reduce mass. It reduces the current the battery delivers. What it can reduce is the derating of everything else.

A battery and a capacitor bank on a common port, with separately switched charge and discharge paths and a controller that supplies sustained DC power from either store or from both at once. Charge and discharge of an electrochemical store are governed by different constraints — charge acceptance is the tighter one, especially cold — so separating the paths is what lets the controller treat them differently.

Granted and in force in India, the United States, Europe (validated in Germany, the United Kingdom and Switzerland), Japan, Canada, Mexico, Vietnam, South Africa and the Eurasian region (in force in Russia). Priority 10 July 2015; terms running to 2035. Twenty United States claims, two of them independent.

A patent recites structure and function, so it cannot support a performance claim, and none is made from it here. Every grant is listed with its number and grant date on the patent register.

Two independent Indian government laboratory reports exist. The Central Institute of Road Transport, Pune recorded 10,000 cycles at 60 °C on Model HPP35A (report B34000, 20 March 2020). DRDO's R&DE (Engineers) establishment tested environmental discharge at +70 °C under a sustained 450 A (report 020623ESG/ETF, 21 June 2023). Both are test reports covering a single configuration, and neither was run against a battery-only control, so neither is cited in support of a percentage improvement.

The comparative measurement — both arms, one duty cycle, current logged at a kilohertz and above, an end-of-life criterion declared in advance, and the decision thresholds written before the run — is specified in full. How to read a test report, including these two.