Battery autonomy: is the system obliged to tell you how much is left?

Eight hours of battery autonomy. The number is in the specification, the acceptance report is signed, the maintenance contract is in place. Then at two in the morning the mains supply drops, and neither the person standing in the room nor the one watching the remote console can say how much time is left: the system reports that the battery is discharging, not for how much longer. If the service falls over forty minutes before the design promised, the question that follows is not whether the battery was compliant. It is who tested those eight hours, when, and against which pass criterion.

A battery that watches itself moves the problem

ETSI ES 202 336-11 V1.2.1, published in September 2025 by the Environmental Engineering technical committee, is the technical twin of Recommendation ITU-T L.1397 (10/2025), edition 1.0 approved on 7 October 2025 and in force: two references, one body of content, both downloadable at no cost. It applies, the scope says, to batteries “with integrated control and monitoring for telecommunication and datacom (ICT) equipment”.

The 2025 revision is not cosmetic: the change history reads “Revised table A and B to align with actual technologies”. What has changed is where the intelligence sits. “In most existing installations, there is no specific control/monitoring provided as part of the battery system. It is provided by the power system.” With the recent chemistries the control moves inside the storage itself: a Battery Management System for every Integrated Battery Unit, and a Master Battery Management Module once several units make up an Integrated Battery System. Whoever accepts the site is no longer buying reserve alone: they are buying a device that makes statements about itself.

And it makes them inside a wide band: “Battery systems provide typical standby of 10 min to 48 hours on grid connected sites.” Ten minutes or two days: between those two extremes it is the specification that decides, not the standard.

The mandatory minimum is generous, and autonomy is not in it

Table A.1 sits in a normative annex: it is the minimum a compliant system has to expose. State of charge and state of health — SoC and SoH — are there among the recorded data; so is remaining capacity, with the calculation spelled out, “Data in Wh and Ah, establish from SoC x SoH”; so are the cumulated charge and discharge figures in ampere-hours and watt-hours; so is the terminal voltage, “Terminals battery DC voltage (±0,1 V)”, and the temperature, “Battery temperature (±2 °C)”, logged every five minutes. There is a cycle counter, with the cycle defined in plain words: “One cycle is defined with equal or more than 80 % from initial capacity of cumulated discharge”. And among the mandatory inventory items, two separate entries: “Battery design capacity” and “Battery usable capacity”.

Only one thing is missing, and it is the one that is needed at two in the morning. “Estimated remaining battery autonomy (time) during discharge” sits in Table B.1, which is not mandatory. In the ITU twin it is Appendix I, and the heading leaves no room: “This appendix does not form an integral part of this Recommendation.” The same table holds the length of the discharge test — “Duration of battery discharge time test”, in the L.1397 wording — the “Battery test parameters” among the configuration items and the “Battery test on demand” among the commands, explained in five words: “This is related to SoH”.

Put plainly: a fully compliant battery system may tell you how charged it is and how far it has aged, never tell you how many minutes are left, and never accept a discharge test commanded from the supervision system.

The test result is mandatory, the criterion is not

None of this is new in 2025. Part 2 of the same series, ETSI ES 202 336-2 V1.1.1 (2009-03), still published, lists among the mandatory alarms of a direct current plant the “Battery test failure”: “Battery failed to pass defined test criteria e.g. autonomy time, voltage threshold”. Defined by whom, the document does not say. And the detailed record of the test — “discharge chart, temperature, string current, cell voltages” — sits there too, among the non-mandatory items. An alarm that says failed without the curve that proves it does not hold up in a dispute, much like redundancy that is calculated and never tested.

Two capacities, two estimates, one measuring point

Three lines to settle before delivery.

The capacity. Is the design autonomy calculated on the rated figure or on the usable one? They are two separate entries in the mandatory inventory, and the second is always the smaller.

The estimate. Remaining capacity is the product of two estimates, SoC and SoH, which the BMS computes by its own methods: the document prescribes the datum, not the way of arriving at it. Two units from different suppliers, same rating and same history, may well declare different states of health.

The point. A note to the table fixes it: “Battery voltage need be taken by BMS before the DC/DC converter.” Downstream of the converter the voltage is a different one, and so is the end-of-discharge threshold — the same discipline that applies to the −48 V power supply interface.

What to write in the specification, what to check at acceptance

In the specification: an explicit call-up of the non-mandatory items you actually need — remaining autonomy during discharge, test parameters, test on demand, recorded test duration; the capacity on which contractual autonomy is calculated, with the reference load and the end-of-discharge voltage; the voltage measuring point; and the “Safe mode actuated” alarm exposed to the supervision system, the moment at which the contactor opens and the unit goes to restricted operation.

At acceptance, reading the rating plate is not enough. Check that the test command travels from the supervision system and not only from the local display; that the recorded test states load, end-of-discharge voltage, temperature and duration; and that the report says which of the two capacities the autonomy was calculated on.

What we do not know

We read ES 202 336-11 V1.2.1 and L.1397 (10/2025) in full, downloaded on 19 September 2026, together with the tables of ES 202 336-2 V1.1.1. We did not read ES 202 336-1, the generic interface, which carries the general accuracy requirements the notes point to: the figures above are the ones written in the tables themselves. We did not read ETSI TS 103 553-2 or ITU-T L.1221 on energy storage architecture. The IEC standards the text cites — 60050-482, 62620, 62619 — are paid-for: we have not read them and we report nothing from them. The figures in both documents are images inside the PDF and we extract no values from them. An ETSI Standard is not a harmonised standard: it binds if the contract calls it up, and we have not checked whether any Italian provision makes it mandatory in network procurement.

The two axes, applied to a declared autonomy

First axis: in the specification we review autonomy becomes a method — reference capacity, load, end-of-discharge voltage, test command, curve record — and at acceptance we check that those items exist on the equipment as delivered, with testing and documentation handed over together with the work.

Second axis: a discharge test is only worth running if it can be compared with the one before it. With CSIDIA, the other company in the group, discharge curves, SoC and SoH readings, the inventory of units and the alarms become a single map of the site, where an AI flags the drift — the state of health falling on one unit alone, the test repeated at a load different from last time — and the team steps in before the lights go out. It runs on-premises, without deep integration into the client network, or in a dedicated cloud with data centres in Italy, always under shared management. It holds for a telecom exchange as much as for a corporate data room.

A site visit, at no cost, produces the list of your energy reserves: for each one, whether the autonomy is declared or tested, whether the test can be commanded remotely and against which criteria, and where the last discharge curve sits today — empty boxes included. Talk it through with an engineer: a blackout gives no notice, and how many hours were really left should not be left as an open question.