There is no simple window into the battery showing a liquid level. A battery gauge works with electrical measurements and a model of how the battery behaves. Useful inputs include current, voltage, temperature and time. Current tells the system about charge moving into or out of the battery. Voltage gives another clue, but its meaning changes with conditions. The model connects those clues with the quantity the gauge is trying to report. Component manufacturers distinguish the measurement circuitry from the algorithm that turns measurements into an estimate. [2][3] Understanding that division explains why a precise-looking number can still involve uncertainty.

Voltage is especially easy to misread. Drawing current changes a battery's terminal voltage, and temperature and the battery's recent history also matter. After the load falls, the voltage can gradually recover without a charger adding energy. A flat part of the voltage curve can also make a small measurement error important when someone tries to infer charge from voltage alone. [2][3] These facts do not make voltage useless; they make context essential. A battery-management system needs an appropriate interpretation of its signals. The component references explain that principle, while the particular sensors and algorithms in an individual car remain manufacturer-specific.

A battery-management system estimates an internal state from accessible electrical and thermal signals. Current integration constrains charge transfer; terminal voltage contains information about electrochemical state together with load-dependent and time-dependent effects; temperature changes the interpretation of those observations. TI explicitly separates measurement hardware from the battery model and gauging algorithm, and Analog Devices describes monitors that expose measurements to a host as well as gauges that perform estimation internally. [2][3] These are complementary implementation examples. They do not establish that a named gauge IC or a particular algorithm is installed in the EV a reader owns.

The distinction between terminal voltage under load and an appropriate open-circuit-voltage relationship matters. Load current, temperature, relaxation and battery condition affect a voltage reading, while a comparatively flat voltage-versus-charge region weakens the information available from one voltage observation. [2][3] A highly repeatable voltmeter can therefore provide an accurate measurement that remains insufficient for an accurate state estimate. Measurement precision and state observability are different properties. A practical estimator may combine the recent charge-transfer history with a model-informed voltage interpretation; evaluating that design requires its assumptions and operating conditions, rather than a single voltage threshold borrowed from another chemistry or application.

Keeping a perfect running total does not fix a wrong opening balance. Suppose the same hypothetical ledger starts from an assumed 65% when the defined starting value is actually 70%. Subtracting the same 12.5 Ah gives 52.5% instead of 57.5%. The five-percentage-point error remains. A current counter can accurately describe what changed while still being wrong about what is left. That follows directly from the starting-state term in the counting method. [1] It explains why a gauge benefits from information that can check its estimate, rather than only adding more digits to its current reading.

Now keep the correct starting balance but invent a constant sensor offset: every current reading is 0.2 ampere too high in the outward direction. Across the 1.75-hour example, the counter records an extra 0.35 Ah leaving the battery. It reports 57.15% instead of 57.5%, an error of 0.35 percentage point. This is arithmetic for the stated teaching model, not a claim about typical EV sensor accuracy. The lesson is that a small persistent bias accumulates with time. It is different from readings that happen to vary around the correct value. Neither the percentage displayed nor its number of decimal places tells an owner the size or cause of any actual error.

The initial-condition error is preserved by ideal open-loop integration when the same current and capacity are used. In the hypothetical ledger, replacing the correct z0 = 0.70 with 0.65 shifts the final estimate from 0.575 to 0.525. No subsequent exact current sample supplies the missing five percentage points by itself. This limitation follows from the estimator equation and is consistent with MathWorks' separate examples of correcting a state estimate using model-based information. [1] It should not be translated into a claim that every vehicle requires an owner-performed reset or that a particular vehicle's gauge has accumulated such an error.

A constant current offset produces another identifiable error term. With outward-positive measured current Im = I + b, the SOC error caused by that bias is −bT/Qref under the same ideal assumptions. Set the invented bias b = 0.2 A and total duration T = 1.75 h: the excess counted discharge is 0.35 Ah, and the state error is −0.0035, or −0.35 percentage point. The final estimate is 0.5715. A percentage-point error should not be confused with a relative percentage error. This example isolates bias; clock error, sampling behavior, reference-capacity error and model mismatch are different contributors and cannot be assigned numerical values from this toy calculation.

Cold conditions can affect how much stored energy is currently available for use, as well as how much energy the car needs. These effects should not all be called permanent battery damage. In the U.S. Model 3 manual, a blue snowflake indicates that some stored energy is unavailable because the battery is cold; the indication disappears when the battery is sufficiently warm. The same passage explains that charging, acceleration and regenerative braking may be limited. [6] This is a documented example of a temporary operating restriction. Other vehicles may communicate their limits differently, so the symbol and its meaning should come from the correct manual.

Aging is a different question: how the battery's longer-term condition has changed. A cold morning and a warm afternoon are poor conditions for comparing capacity unless the vehicle's diagnostic method accounts for temperature. The example of 90 Ah remaining full capacity in the earlier section assumed comparable conditions; without that assumption, the ratio would be harder to interpret. [3][5] Warming a battery also does not create electricity. The car may use energy while heating itself or its cabin. A change in access to stored energy and energy spent on warming can happen together. Reading both the operating conditions and the vehicle's indications is more useful than judging health from one changed number.

Temperature-dependent deliverability and irreversible capacity change should be analyzed separately. Terminal-voltage behavior and the charge deliverable before an operating cutoff depend on temperature and load; an SOC or available-energy interpretation needs an appropriate model of those conditions. [2][3] Tesla provides a concrete user-interface example: its U.S. Model 3 manual associates the blue snowflake with stored energy temporarily unavailable because the battery is cold, and states that the icon disappears once the battery is sufficiently warm. It also documents possible limits on regenerative braking, acceleration and charging. [6] These observations do not quantify another vehicle's temperature sensitivity or identify its internal control thresholds.

An assessment of long-term capacity retention requires a comparable reference, rather than comparing an arbitrary cold observation with an arbitrary warm one. The hypothetical 90/100 Ah retention ratio assumes matched relevant test conditions. Tesla separately documents natural aging that can reduce the maximum range available when fully charged. [5] There is no contradiction between an improvement in short-term energy availability as temperature changes and simultaneous energy consumption by thermal management. They affect different terms of the operating budget. Consequently, a dashboard change after warming is insufficient evidence to calculate either degradation or recovered energy without further information about loads, definitions and the estimator's behavior.

A battery gauge may have more information after conditions settle than it had during a changing load. The voltage can relax after current falls, and an estimator may use an appropriate battery model to check its running charge count. The underlying references describe both current-based estimation and model-informed correction. [1][2][3] A revised estimate does not mean new charge appeared from nowhere. It means the reported understanding of the battery's state may have changed. That is the same distinction as correcting an account balance after finding a bookkeeping error, rather than receiving a new payment.

There is also a less abstract possibility: the car used energy while parked. Tesla documents periodic system activity and support for the low-voltage battery while Model 3 is idle and unplugged. [5] An overnight change can therefore involve real energy use, altered conditions or an estimate revision. The display alone does not separate those causes. A useful observation records the time, whether the car was plugged in, broad temperature conditions, active parked features and the exact screen being compared. That information can make the manufacturer's explanation or a service discussion more specific. It cannot establish that an unexplained large change is harmless, and it should not be used to dismiss a warning.

An estimator that combines charge integration with voltage or parameter information can revise its state when the observations and model support a correction. Analog Devices discusses relaxation and fuel-gauge learning, while MathWorks documents examples of model-based state estimation that correct an initially inaccurate estimate. [1][3] The correction mechanism and any smoothing or display mapping are implementation-dependent. The existence of such methods is not evidence that a particular observed jump came from that mechanism. A state-estimate update changes a numerical representation; a physical energy balance still requires an actual source or sink for any energy transferred.

Parked energy consumption is another possible contributor. Tesla's Model 3 battery documentation identifies periodic system checks and replenishment of the low-voltage battery when the vehicle is idle and unplugged. [5] Distinguishing consumption from an estimation revision requires more than two dashboard screenshots: comparable display definitions, elapsed time, charging connection, operating conditions and relevant background loads are needed. An owner can preserve those ordinary observations without accessing high-voltage hardware or engineering diagnostic buses. The appropriate conclusion from incomplete data is often a set of possible explanations rather than a diagnosis. Persistent unexpected behavior or vehicle alerts should be handled through the documented support process, not explained away by a generic claim about calibration.

The charge ledger used Ah because it counted current over time. Your electricity bill and many driving-efficiency screens use kWh because they describe energy. Voltage links the two, but the voltage can change during use. That is why an ampere-hour count is not automatically an energy reading, and why a percentage of charge cannot always be multiplied by an advertised kWh figure to produce an exact remaining-energy answer. Battery-gauge references discuss charge capacity, voltage and operating conditions as distinct inputs. [1][3] A calculation should say which quantity it uses before presenting a result with apparently precise units.

Power is different again: kW describes an energy-transfer rate at a moment. A high charging-power number does not tell you the battery's present charge percentage or how much energy an entire session will add. Likewise, electricity supplied at a charger is not automatically the change in stored battery energy; a practical comparison needs a stated boundary and treatment of the vehicle's other energy uses. For the examples in this article, the boundaries are deliberately separate: Ah for the charge ledger, a defined capacity ratio for the health illustration, and an assumed battery-energy budget for range. Keeping those columns separate prevents a neat-looking calculation from answering the wrong question.

The electrical quantities have different dimensions. Charge transfer is the time integral of current, ΔQ = ∫I dt. Electrical energy transfer is the time integral of terminal power, ΔE = ∫VI dt, with signs and the measurement boundary defined consistently. If voltage changes, multiplying an Ah total by an arbitrary nominal voltage is an approximation rather than an exact energy integral. The charge-based estimator documentation explicitly integrates current and normalizes by Ah capacity; the fuel-gauge references also identify voltage and operating conditions as relevant to interpretation. [1][3] A displayed energy-based percentage may involve additional modeling beyond that simple charge-normalized state.

Power in kW, energy in kWh and charge in Ah should therefore remain separate in a log or worked example. Battery-terminal electrical energy, usable stored energy and electricity measured outside the vehicle also represent different boundaries; conversion losses and operating loads cannot be assigned to the same quantity without explanation. This article's ledger assumes ideal charge accounting, its capacity-retention ratio assumes matched conditions, and its range calculation begins with a separately stipulated usable-energy allocation. None is a charger-energy audit. Matching dimensions is necessary, but the reference state and boundary are equally important when deciding whether two readings can legitimately be compared.