What is the status of the meter roll out in Germany? 

As of 31 March 2025 (from data provded by Bundesnetzagentur and FFE) there are roughly 55 million locations across Germany. About 4,5 million smart meters (iMSys) should be installed by 2032 representing a total national share of roughly 9%. As of 31 March 2025 only 1.5 million (mandatory and voluntary) smart meters (iMSys) had been installed. This leaves roughly 97% of all meters—nearly 53 million units—as either traditional analog Ferraris meters or digital meters that lack a communication gateway.

Why have the new non smart digital meters not solved the problems?

Critically, the majority of new non smart digital meters that have been installed do not have real-time data communications, leaving the industry "blind" to localized power quality issues. This reduces the operator's ability efficiently to manage, stabilize, and fully digitalize the low-voltage network, activities required for a successful and safe energy transition. The DNO still has to rely, for the majority of data points, on pseudo-measurements when calculating Static or Dynamic State Estimates.

How can Steelhouse help?

Steelhouse can help „fill in the gaps“ left by the absence of real time recorded data. By using a combination of low cost, special purpose network sensors and neural network/transformer time series technology, Steelhouse Data Networks is able to collect and collate data and analyse that data to provide time critical voltage forecasts.

Steelhouse technology supports the monitoring and forecasting of the condition of the low-voltage network, detecting frequency drift and voltage fluctuations and enabling the development of predictive maintenance strategies.

How can a DNO realistically use such a voltage forecast?

The forecast can be used by the DNO to guide a Voltage Regulating Distribution Transformer (VRDT) equipped with an On-Load Tap Changer (OLTC). The Tap Changer is the primary physical tool the DNO has to keep voltage within the legal limits. A good forecast can help to solve three specific problems:

Mechanical Wear: Every "tap" is a physical movement. If a grid is volatile (clouds passing over solar panels), the tap changer might switch 50–100 times a day. This leads to mechanical failure and expensive maintenance. Reactive vs. Predictive: Standard tap changers react to what is happening right now. If a massive spike is coming in 15 minutes, a reactive tap changer might wait too long, causing a brief voltage violation. The "Hunting" Problem: If two transformers are close together, they can "hunt" - one raises voltage, the other lowers it, creating a cycle of wasted mechanical movements.

Does a DNO need to provide GPU compute capacity and/or AI technical knowledge ?

No. Steelhouse has GPU compute capacity and neural network/transformer time series capability. The DNO will need to provide a liaison person who is familiar with the objectives of the project and the operations of the DNO low voltage grid. 

Where are the Steelhouse servers located and does Steelhouse use energy intensive hyperscaler cloud compute?

No. Steelhouse, is a european company and does not use hyperscaler cloud compute. Steelhouse servers and software are located in Europe and are therefore not subject to US Cloud Act (2018) legislation.

What level of frequency measurement accuracy is useful for grid balancing?

A frequency forecast using equipment with an an accuracy in the range of ±10 mHz (IEC 61000-4-30 Class A) could be extremely useful for EU grid balancing, as the ±10 mHz threshold matches the deadband for the crucial Frequency Containment Reserve (FCR). This accuracy is suitable for proactive and optimal dispatch of the aFRR (Automatic Frequency Restoration Reserve) and mFRR (Manual Frequency Restoration Reserve) markets, allowing TSOs to anticipate control actions, improve system stability, and reduce the overall cost of balancing the grid. 

What is unintentional islanding?

This is a critical safety and operational problem in modern electrical grids caused by high penetration of local distributed generation such as solar PV (Balkonkraftwerke). It occurs when a section of the utility grid, containing both customer loads and generation, becomes electrically isolated from the main grid but remains energized by the local generation. The most challenging scenario is the Non-Detection Zone (NDZ), which occurs when local generation perfectly matches local load, preventing the resulting voltage and frequency from drifting enough to trigger the distributed generation standard protective shutdown mechanisms. 

Are neural network or transformer time series forecasting models classified as AI and are they regulated in the EU?

Neural network/transformer time series forecasting models are likely to be classified as AI under the EU AI Act 2024. Steelhouse is aware of Annex III under the Act but due to the nature of the products being offered – Steelhouse is supporting human operator decision and is not agentic – Steelhouse believes its activities will fall in the limited risk category. Steelhouse is in the process of discussing and registering our technology with the relevant regulator(s) and if appropriate will sign any EU sponsored code of practice.

Are there any GDPR implications regarding voltage and frequency?

The Steelhouse neural network/transformer time series forecasting models use, where possible, aggregated and anonymised data. Any voltage or frequency data gathered and/or forecasted will be stored in a separate, identified database and Steelhouse operations will, at all times, conform to the GDPR rules. 

What about the regulations KRITIS DachG and NIS2?

Steelhouse itself is not regarded as an essential entity or important entity as defined under the regulations (expected to be in force in 2026). The technology that Steelhouse is providing is intended to be compliant for any essential entity or important entity that is operating under KRITIS or NIS2

Is data technology and digitalisation supported by the regulator?

After industry consultation the Bundesnetzagentur in December 2025 introduced a regulatory framework and methodology for incentive regulation for electricity distribution system operators (RAMEN Strom). The RAMEN Strom formulas are designed to exert financial pressure on grid operators who rely mainly on conventional grid expansion strategies ("Kupfer statt Köpfchen"). By the start of the fifth regulatory period on 1 January 2029, a DNO that has failed to implement digitalisation measures will likely face a high individual efficiency factor causing their profit margins to diminish significantly. Conversely, the system rewards innovative operators who utilize data-driven efficiency to lower their cost base while simultaneously generating additional revenue through the new Digital Quality Factor.