Internal resistance

What it is (precise)

Internal resistance is not one physical resistor but the sum of every obstacle current meets inside the cell. The electronic path runs through the aluminium and copper current collectors, the tab welds and the terminals; the ionic path runs through the electrolyte filling the separator and electrode pores; and at each electrode interface, polarization effects — charge transfer and diffusion — add an apparent resistance that builds over time rather than appearing instantly.

That last part is why the measured value depends on the timescale of the measurement: a very fast measurement sees mostly the ohmic paths, while a longer current pulse sweeps in more and more polarization, so the same cell reports a larger resistance the longer you hold the pulse.

Practice has settled into distinct measurement families that do not produce the same number. DCIR — DC internal resistance — applies a current step and divides the voltage change by the current change, at a stated state of charge, temperature and pulse duration; vendor protocols run from fractions of a second to tens of seconds, and they do not all use the same combination. ACIR injects a small AC signal, conventionally at 1 kHz, and captures mostly the ohmic contribution — a fast measurement suited to production lines.

Electrochemical impedance spectroscopy sweeps frequency in the lab and resolves the individual contributions. Because the DC pulse includes polarization that the 1 kHz measurement excludes, DCIR for a given cell is systematically larger than its 1 kHz ACIR — quoting one against the other is not a comparison. For modern 280-314 Ah LFP cells, datasheet DCIR is typically a few tenths of a milliohm.

The value is also strongly conditional on the cell's state. It rises toward the ends of the state-of-charge window, most sharply near empty. It rises steeply in the cold — electrolyte ionic conductivity falls by roughly an order of magnitude between +25 C and -20 C, and the electrolyte is on the current path. It can differ between charge and discharge direction at the same state of charge. And it grows through life as the cell ages. Treat any single quoted figure as one point on a surface, not a property of the cell.

Why it matters in a real grid-scale project

The first consequence is heat, and the law is quadratic: ohmic heat per cell is I²R. Take 0.25 mΩ as a representative figure inside the canonical few-tenths range — an assumption for arithmetic, not a datasheet claim. A 314 Ah cell discharging at 0.5C carries about 157 A and dissipates roughly 6 W.

A 5 MWh-class container holds roughly 5,000 such cells (5 MWh divided by 3.2 V × 314 Ah per cell), so the fleet inside one enclosure generates on the order of 30 kW of ohmic heat at 0.5C — and because the law is quadratic, running the same container at 1C quadruples that per-cell heat, not doubles it. This is the load the liquid cooling loop is sized to remove, and every kilowatt the thermal system draws to remove it lands in the plant's auxiliary consumption.

The second consequence is efficiency. The ohmic loss fraction per direction is roughly I × R divided by cell voltage — about 1% each way at 0.5C in the same arithmetic — and it is paid twice, once charging and once discharging.

That loss sits inside the DC slice of the plant's AC-to-AC round-trip efficiency, which for modern LFP systems is typically on the order of 85-92% at the POI alongside PCS, transformer and auxiliary losses. Because resistance is current-dependent, a system dispatched hard at high C-rate is measurably less efficient than the same system run gently, which is why RTE guarantees are always tied to a stated duty.

The third consequence is voltage sag: the terminal voltage drops by I × R the instant load is applied — about 40 mV at 0.5C in the worked example. On LFP that is not small: the open-circuit voltage moves only tens of millivolts across the whole 20-80% state-of-charge mid-range, so the sag under load is the same order as the entire plateau, which is one reason the BMS cannot read state of charge off a loaded voltage.

At system level, a cold rack has higher internal resistance, and the resulting sag can push the DC bus toward the bottom of the PCS VDC window during hard discharge — winter deliverable power is a real study case, not a footnote. Near-empty operation compounds it, which is part of why discharge power derates at low state of charge.

How it ages and how the BMS uses it

Degradation expresses itself as two measurable trends: stored energy goes down, and internal resistance goes up. The mechanisms overlap with capacity fade but are not identical — the solid-electrolyte interphase thickens on the anode and adds impedance as it consumes lithium, the electrolyte slowly degrades and dries out over years of cycling, and mechanical aging works on the welds and interfaces the current must cross.

The practical result is that a year-15 cell is both smaller and more resistive than it was at beginning of life, and every consequence in the previous section — heat, efficiency, sag — is computed on the aged value, not the datasheet one.

That is why State of Health has a second, resistance- or power-based definition alongside the capacity-based one: rising impedance limits how hard the system can charge or discharge before capacity runs out. Round-trip efficiency typically drifts down a point or two from beginning of life to end of life as resistance grows, and for demanding short-duration, high C-rate duty a system can hit a practical power or duration end-of-life before its energy end-of-life. A degradation model that tracks only retained MWh misses the axis that decides whether the plant can still deliver its contracted MW.

Operationally, resistance is one of the BMS's most useful diagnostics. The BMS watches each cell group's impedance signature as part of its State of Health estimation, and a cell whose resistance climbs ahead of its siblings runs hotter under the same current — an early flag worth investigating long before it becomes a thermal event.

Step changes tell their own story: a sudden resistance jump in one group points at a connection or weld, not chemistry. In series strings the highest-resistance cell dissipates the most heat and sags first, so resistance spread across a rack is a direct input to balancing and to which cells limit the discharge.

How it shows up in specs, studies and contracts

On a cell datasheet, internal resistance appears as a DCIR or ACIR line with — if the datasheet is any good — the state of charge, temperature and pulse duration behind it. Those conditions are the first thing to check and the first thing to normalize before comparing vendors.

Upstream of the datasheet, manufacturers grade cells by capacity and internal resistance at the end of the production line and match them into modules, because a series string performs like its most resistive member. At factory acceptance testing, capacity and internal-resistance sampling on cells is typical scope — the project's first independent look at the distribution it actually bought.

At system level the phrase almost never appears at the point of interconnection, but its fingerprints are on most of the guaranteed numbers. It is inside the round-trip efficiency guarantee and the capacity-test results that verify it; it is the physical content of the heat-generation tables the vendor supplies for thermal modelling and cooling design; and the commissioning baseline quietly encodes it, because the beginning-of-life test is the reference every later dispute is measured against.

When efficiency or deliverable power drifts over the years, resistance growth is usually the mechanism even though no contract clause names it.

Questions worth asking on any project: what conditions sit behind the quoted DCIR, and is the number beginning-of-life or a warranted later-year value? Does the vendor provide heat-generation data at the project's actual duty and site temperatures, or only at the lab reference? Does the warranty track any resistance or power-capability metric, or only retained energy? And what per-cell resistance data does the BMS log — and can the owner's engineer access it — since that history is the earliest available evidence in both a warranty claim and a safety investigation.

Common pitfalls

The classic tender-stage error is ranking vendors on mismatched DCIR figures. A value at 50% state of charge, 25 C and a short pulse is not comparable to one at a different combination, and a 1 kHz ACIR is not comparable to any DCIR at all — the two methods measure different physics and the AC figure is systematically the smaller. Normalize the conditions or ask the vendors to re-quote at matched ones; a bid that looks 20% better on resistance may just have been measured faster.

The second error is treating internal resistance as a constant. It varies with state of charge, temperature, current direction and age, so a plant's winter, near-empty, late-life behavior is set by a value that can be a multiple of the datasheet number — while the energy model quietly assumed the beginning-of-life lab figure. Cold commissioning tests and hot-summer capacity tests on the same plant are not measuring the same machine.

The last error is milliohm complacency. Fractions of a milliohm per cell read as negligible until multiplied across roughly 5,000 cells per container and the square of the dispatch current, at which point they become tens of kilowatts of heat per enclosure, a slice of the RTE guarantee, and the sag that decides whether the DC bus stays inside the PCS window on a cold hard discharge.

The same complacency shows up in degradation reviews that track only capacity: two fleets at identical retained energy can differ materially in deliverable power and efficiency once resistance growth is accounted for.