What monitoring evidence an insurer can actually price on a battery site

An underwriter pricing a battery energy storage system has a small population of comparable losses, a technology whose failure modes are still being characterised, and, on most sites, almost no visibility of how the asset has actually been operated. In that position the reasonable thing to do is price against the class: cells of this chemistry, in this enclosure format, at this scale, in this fire jurisdiction.

The consequence is that a well-operated site and a poorly operated one, with the same nameplate and the same enclosure, look identical at renewal. Nothing about that is unfair. It is what happens when the only site-specific information available is the specification and the claims history.

The question worth asking is therefore not how to argue for a discount. It is what evidence would let this site be assessed on what it has been doing rather than on what sites like it have done.

Thermal runaway and fire spread dominate the conversation because they dominate the severity distribution. A single cell failure that propagates can take an enclosure, and in a dense installation the question of whether it stays inside that enclosure is a design and separation question the operator cannot change after commissioning.

Frequency, though, usually sits somewhere less dramatic. Cooling and HVAC failures, power conversion faults, connection and busbar heating, auxiliary supply problems and control system issues account for a large share of events that cost money without making the news. These are conventional industrial equipment problems, and they are the ones a monitoring programme is best placed to speak to.

Business interruption is the third component and is often the larger number. It depends less on the physics and more on spares availability, contractual response times and whether the site can demonstrate what happened, which determines how quickly a claim settles.

Continuous rather than sampled. A question about a specific enclosure on a specific afternoon eighteen months ago has to be answerable. Monthly summaries and screenshots do not survive that question, and an underwriter or a loss adjuster will ask it.

Timestamped and synchronised. When an incident sequence matters, the order of events matters, and an order assembled from three systems with unsynchronised clocks is not a sequence. Time synchronisation across the battery management system, the conversion equipment, the fire and gas detection and the plant historian is unglamorous and decides whether a record is usable.

Traceable. A figure in a report has to be followable back to the measurements behind it. This is the property that distinguishes evidence from assertion, and it is the one most often missing.

Retained. Retention set to the default of whatever historian was installed at commissioning is almost never long enough. The useful horizon is the life of the policy relationship and the warranty, not the last ninety days.

Cell and module level condition rather than only pack aggregates. A pack-level state of health aggregates a large population, and a small number of cells departing from their neighbours is exactly the pattern that precedes a problem while barely moving the reported figure. In a series string the usable capacity is set by the weakest cells, so the spread carries information the aggregate cannot. The reasoning is the same as for an electrolyser stack, where cell voltage distribution carries information that the aggregate hides.

Thermal behaviour with spatial resolution. Temperature spread across a module and across an enclosure, not just a maximum. A rising spread at constant duty is a different signal from a rising maximum, and it is one of the precursors covered in thermal runaway: what is predictable and what is only detectable.

Duty history. Depth of discharge distribution, C-rate distribution, time at high state of charge, ambient conditions and cycle count against the warranty definition of a cycle. This is also what a battery warranty claim turns on, so the record serves twice.

Auxiliary and containment systems. HVAC performance, door and enclosure status, gas detection history, and the maintenance record against them, which is also where a HAZOP and LOPA study should have set the instrumentation requirement. A cooling system that has been degrading for six months is both a claims risk and an underwriting fact.

Incident and alarm history with resolution. Not only what alarmed, but what was done and by whom. An alarm history with no resolution field is a list of unanswered questions.

It does not entitle a site to a lower premium. Underwriting is priced per risk, per market and per renewal cycle, and capacity, reinsurance conditions and the loss experience of the wider portfolio move rates far more than any individual site's data. A year in which the market hardens will overwhelm a good evidence pack.

It does not substitute for design. Separation distances, enclosure ratings, detection and suppression, and the fire authority's position are set at design and construction. Monitoring evidence speaks to operation, not to whether the installation was laid out well.

And it is not automatic. What continuous monitoring and a traceable incident history do is let an underwriter price against evidence rather than against a class assumption. Whether that produces a better outcome depends on the underwriter, the evidence and the year, which is why this is described here as a precondition for the conversation rather than a result of it.

Treat it as a document rather than a data feed. Underwriters and their engineers do not want access to a portal; they want a pack that states what is measured, at what resolution, how long it is kept, how it is protected from alteration, and who produced it.

A workable structure: the measurement inventory, with tag, resolution and retention for each; the commissioning baseline, including thermal and capacity references captured while the asset was new; the duty history over the period, summarised and available in full; the incident log with resolutions; the maintenance record against the plan; and a statement of provenance describing which party generated each part.

The provenance point matters more than it looks. Where the operations contractor produces the performance report on its own performance, the incentive is not neutral, and both the insurer and the owner know it. Data that is system-generated, timestamped and traceable does not remove that structural issue, but it does mean the numbers can be checked rather than taken.

Every question above is retrospective. The renewal conversation that makes an operator wish they had the record is not usually the first one, and no instrument installed in the week before renewal can describe the year behind it.

The cheap moment is commissioning, when a thermal baseline, a capacity reference and a documented measurement inventory cost very little and are impossible to reconstruct later. The expensive moment is after an incident, when the questions are specific and the answers are recollections.

Yunify holds that record and attributes movement in condition to a mechanism rather than to a suspicion, which is what allows a figure in a pack to be traced back to something physical. It supports the reporting; the underwriter prices the risk.