DE Arrange a call

For municipal utilities and regional distribution operators

Know where your grid stands, from data you already have.

GridBeat brings together your network, asset and metering data, connects it, and shows you how loaded each substation and feeder is, down to low voltage. Nothing new to install, nobody to hire.

Arrange a call See the examples

Berlinworking across Germanyfirst call free

Example: disruption risk in the grid around Esch-sur-Alzette (Luxembourg), calculated from public data only. Red dots are real faults. How this was made

Where things stand

The requirements are growing faster than the data behind them.

You know this better than we do. Still, four figures describe well what almost every operator is working on at the same time right now.

24.67 %

Average digitalisation index across all German grid operators. The higher voltage levels are well ahead; medium and low voltage are well behind.

E-Bridge report for the BNetzA, cited in dena/BDEW/Bitkom 2025
over 30×

the size of the extra-high-voltage network: the low-voltage network. Instrumenting it end to end takes years. Observability is being asked for now.

dena, Data4Grid final report 2023
5 minutes

to determine grid state and send a dimming signal under § 14a EnWG.

Bundesnetzagentur position paper, 2023
every 2 years

a grid expansion plan under § 14d EnWG above 100,000 customers. AgNes adds time- and location-differentiated grid fees on top, and those depend on accurate grid-state forecasts.

EnWG; the BNetzA’s ongoing AgNes process
“Grid operators do have a great deal of data”
dena, Data4Grid

The sentence continues with the real problem: getting an overview of that data, and of the systems holding it. That is where we start.

The obstacle is rarely the technology. What is missing is staff time. You have an expansion plan to write and connection requests to process.

What we do

Three steps. The first one runs on exports you can already produce today.

01

Take stock and put it in order

We collect what exists: GIS and network documentation, substation lists, switch positions, meter readings, connection requests, fault logs. Spreadsheets are fine. Exports from four systems that know nothing about each other are fine. We check the data for completeness and plausibility, connect it, and list what is missing and which gaps are worth closing.

02

Work out the grid state, including where there are no sensors

For every substation and feeder we calculate how loaded it is likely to be across the day and across the year, from your topology, the asset register, weather and load profiles. Where measurements exist, we check the result against them. Where none exist, the calculation stands in for the sensor. dena’s own phrase for this is supporting metering with data-based analysis.

03

Monitoring you will actually open

You get an overview in the browser: which substations and feeders get tight and when, how that shifts with the season, where a connection request is uncritical and where it is not. Plus an export for your expansion planning. No training required, no licences for your team.

The monitoring

What it looks like once your grid is visible.

One page in the browser, no software on your machines. Substations sorted by loading, the peak with a time of day, and an answer to every connection request in seconds rather than weeks.

Grid state · Local substations
Week 37TodayYear
SubstationLoadingPeakTrend
Bahnhofstraße sub.
94 %
Tue 18–20 h↑ +6 %
Lindenweg sub.
87 %
Thu 17–19 h↑ +3 %
Am Anger sub.
71 %
Mon 12–14 h→ ±0
Gartenstadt sub.
58 %
Sat 12–13 h↓ −2 %
Mühlenfeld sub.
41 %
Fri 07–08 h→ ±0

Check a connection request

Asset
Heat pump, 11 kW
Address
Lindenweg 14
Substation
Lindenweg sub.
Loading today
87 %
with this asset
89 %
No problem today. At the current rate of additions the substation gets tight from around 2028: a candidate for the next expansion plan.

Example view with fictional values. The real view shows your substations and your names.

What we are not

So you know what you are not signing up for.

Not a replacement for your control system

GridBeat switches nothing, controls nothing and does not touch operations. It is just the complete state of your grid.

Not an integration project

We start with exports: CSV, shapefile, PDF, whatever you can get out. We build an interface only once the value is proven and you want one.

No hiring

You don’t staff up. We do the data work, and we document it so your people can follow it and take it over later.

No handing over your data

Your data stays yours. Processed in Germany, contract processing under GDPR, deleted on request. Whatever we build from your data belongs to you.

Examples

What public data alone can already show.

We are not yet working with a distribution operator. Everything here was built from open sources, without a single line of operator data. We show it so you can examine how we work before you trust us with anything.

Example 01 · Germany

Predicting congestion and outages, nationwide

We rebuilt the German power grid from public sources, mainly the Marktstammdatenregister and OpenStreetMap. Then we calculated how heavily loaded every node and every line is.

As a test, we checked whether the model finds the places where operators actually had to step in since 2024: published redispatch measures and power plant outages. It was always tested in regions it had not seen before.

What this means for you: We find the weak points of a grid we know only from public data. With your own network data the picture gets sharper and reaches down to your substations.

Said plainly: this is the transmission grid, not your distribution grid, because public event data only exists there. And the test is coarse: a line counts as affected if it lies within two kilometres of a documented measure. All details are on the example page.
Open the full example →
Drag, zoom, click an event. The ☰ button opens the list of events.
Example 02 · Luxembourg

The same method, down to low voltage

Luxembourg publishes its network data openly. So there we could go all the way down to low voltage: every substation, every feeder down to the street. Generation assets and consumption came from public registers and statistics.

As a test, we collected 98 publicly known power cuts and outages from 2022 to 2026 and checked whether the model would have flagged the affected segments as risky beforehand.

What this means for you: This is exactly the grid level nobody in Germany can see today. With your fault logs and substation data, the same method delivers this picture for your area.

Side finding: most faults could only be located to the commune. Your own logs with substation references would be far better.
Open the full example →
Drag, zoom, click an event. The ☰ button opens the list of faults.
Example 03 · Germany, public

Open grid stress map in progress

We are building a freely accessible map showing estimated grid stress nationwide at postcode and substation level, updated daily, entirely from public data.

It is meant as a test: look at what we show for your area and see whether it matches what you know from experience. If it does, it is worth a conversation about what the picture looks like once your own data is added.

Tell me when it is live →

Working together

A start that costs you nothing if it doesn’t hold up.

Week 0

First call · 45 minutes · free

You tell us which systems you run and what is pressing. We tell you whether we can do anything with that. If we can’t, we say so.

Weeks 1–6

Stocktake · fixed price

You send us exports. We deliver two things: a data report recording what you have, what is missing and what closing the gaps would cost. A first grid-state picture of your area. Both are yours, even if it goes no further.

From week 7

Ongoing monitoring

Only once the stocktake has convinced you. Monthly or quarterly refresh, fixed annual fee, cancellable yearly. No minimum volume, no pricing tiers by network size.

Common questions

What grid operators ask us in the first call.

Do we have to build an interface or involve our IT?

No. We start with exports from the systems you already have: CSV, Excel, shapefile, PDF. Your IT installs nothing and opens nothing. An interface only comes once the ongoing monitoring is running and you want to save the effort of refreshing by hand.

Do we need smart meters or sensors in the substations?

Not to start. The grid-state picture is built from topology, the asset register, weather and load profiles. Where measurements exist, we use them to check and sharpen the picture. Every measuring point added later improves the result, but none is a prerequisite.

Who sees our data?

Only we do, under a GDPR data-processing agreement. Processing takes place in Germany. We pass nothing on, we train no models for third parties with it, and we delete on request. Whatever we build from your data belongs to you.

How reliable is the result if we are missing data?

As reliable as the data. We tell you that beforehand. That is why the data report comes first: it names the gaps before we draw a picture. The examples on this page show what public sources alone can achieve; with your fault logs and readings it gets considerably more accurate.

Is this AI?

Partly. Loading is calculated with a load-flow calculation, classic grid physics. Which substations and feeders will cause trouble first is scored by a learned statistical model; you can call that machine learning. It is data-based analysis of the kind dena examined for distribution grids in its Data4Grid project. Nothing in it decides on its own or intervenes, and for every score you can see which inputs drive it.

What does it cost?

The first call, nothing. The six-week stocktake has a fixed price we name before starting. Ongoing monitoring is a fixed annual fee, independent of network size, cancellable yearly. We give concrete figures in the first call, once we know which systems you have.

Who is behind it

Founded by a network scientist who does the data work himself.

Vitor Louzada

Founder · models and grid analysis

Vitor is a network scientist with a background at ETH Zürich. He built the load-flow models and scoring methods behind the two examples on this page and tested them against public fault data.

GridBeat is based in Berlin. We have no sales department: whoever answers you will later do the work.

Who this is for

Municipal utilities and regional distribution operators

We are built for operators without their own development department, the large majority of the 800-plus electricity network operators listed in the Marktstammdatenregister. Large groups have in-house teams for this. You have us.

Metering and asset operators

If you run metering systems, inverter fleets or charging infrastructure: every additional source sharpens the picture. Get in touch.

Where our figures come from. The example figures come from our own analyses; the methods and data sources are laid out on the detail pages. The statements about the current situation come from Deutsche Energie-Agentur publications:

  • dena, BDEW, Bitkom (2025): Gemeinsam digitaler. Sechs Thesen und ein Branchenprozess zu datenbasierten Anwendungen und KI im Stromnetz.
  • dena (2023): Data4Grid. Datenanalysen und künstliche Intelligenz im Stromverteilnetz.
  • Bundesnetzagentur, position paper on § 14a EnWG (2023).

Write to us. We reply to every email.

Tell us who you are, which part of the grid you run, and what is on your mind. A paragraph is plenty.

hello@gridbeat.ai

GridBeat · Berlin · working across Germany