---
title: 'Battery Forecast'
description: When this node's battery will die — current charge, drain rate and a three-scenario forecast of the critical date, plus an abnormal-acceleration detector.
section: AI Analytics
weight: 2
related:
  - ai-analytics/node-reports/battery-comparison
  - ai-analytics/fleet-reports/fleet-battery-forecast
---

import Alert from '@/components/docs/Alert.astro';

## What it is

A report for **a single metering node**: it shows the corrector's current battery charge and forecasts how long it has left to run. Operations teams use it to plan battery replacement in advance, rather than learning that a node went silent after the fact.

The report takes the battery-voltage history over the past year, estimates the rate at which the battery is draining, and names the date when the charge will reach a critical level — in three variants: optimistic, realistic, and pessimistic.

<Alert type="tip">
For the whole fleet at once, use [Fleet Battery Forecast](/en/platform/v3/ai-analytics/fleet-reports/fleet-battery-forecast) — a snapshot and ranking of all nodes by battery condition.
</Alert>

## What it shows

| Block | What it means |
|---|---|
| Current charge | the last measured value: percent, volts, or millivolts — depends on the device type |
| Drain rate | how much the charge drops per month; the faster it drops, the closer replacement is |
| Date forecast | three expected dates at which the charge reaches the critical level (see below) |
| Acceleration flag | a warning if, over the last 30 days, the battery has started draining noticeably faster than usual |

Three forecast variants:

- **Optimistic** — if the battery keeps draining at its current rate.
- **Realistic** — accounting for the fact that drain accelerated over the last month.
- **Pessimistic** — a worst-case margin (sharp acceleration).

## How it works

The data source is the device's communication-session history (a log of the device coming online, from which battery voltage readings are taken). For some devices the charge comes from passport settings, where the charge is given in percent.

Different correctors measure the battery in different ways, so the critical thresholds are specific to each type:

| Device type | What is measured | Critical level |
|---|---|---|
| Type 2 (Flowgaz / ULTRAMAG) | charge in percent | 20 % |
| Type 13 (ELSTER EK270) | voltage in volts | 3.3 V |
| Type 23 (BBT-DDT) | voltage in millivolts | 3300 mV |

Calculation logic:

1. Collect the voltage history for the period.
2. Compute the drain rate (how much is lost per month).
3. Extend the trend line forward to the critical level — this is the optimistic date.
4. Compare the rate of the last 30 days with the previous 90 days. If the battery is draining faster in the last month, the realistic and pessimistic dates move closer.

A year of history is the minimum for a reliable forecast: a battery lasts 5–10 years, and over a short period there may not yet be any noticeable drop.

## Run parameters

| Parameter | Default | What it means |
|---|---|---|
| Metering node # | required | the metering node the report is built for |
| Period from | one year ago | start of the history period |
| Period to | yesterday | end of the history period |

## How to read the result

- Look first at the **drain rate**: up to 1.5 % per month — normal, just monitor; 1.5–3 % — plan replacement within a year or two; faster than 3 % or with an acceleration flag — replace within the coming months.
- The **realistic date** is the main planning reference. The pessimistic one provides a margin: if the field visit is arranged by then, the node will not go silent.
- **The shorter the history, the rougher the forecast.** Less than half a year of data — the report will honestly flag the forecast as unreliable; with very few points, no forecast is built at all.
- The forecast **does not account for frost**: in cold weather the battery drains faster, so in winter the real date may come earlier than calculated.
- The passport charge (in percent) is updated rarely — once a week or a month. Voltage from communication sessions is fresher, changing each time the device comes online.