EXPLAINABLE AND AUDITABLE AI FOR THE POWER GRID

From chaos to clarity. From reaction to prediction.

We are not the next control system. We are the explainability and audit layer that every AI-assisted control system needs.

Human-in-the-LoopOn-premise readyCIM integration in preparation
audit log · entry #004127 · examplechain intact
TIME2026-10-14T09:42:17Z
FORECASTNode 7 · transformer loading 112% in 3 h
CAUSEWind feed-in +38 MW · share 61%
RECOMMENDATIONRedispatch: WF North −12 MW
DECISIONApproved by operator · 09:44:02Z
SHA-2569f3c 71ad 0e42 b8c5 … a21e
THE PROBLEM

The grid is getting more complex and more expensive. And it must be run with fewer skilled staff.

ABOUT
€3.1 bn
spent on congestion management in 2025, up on the previous year
Source: Bundesnetzagentur, congestion management 2025
ABOUT
9.4 TWh
of renewable power curtailed in 2024 because the grid could not take it
Source: Bundesnetzagentur, congestion management 2024
ABOUT
€651 bn
grid investment needed by 2045, €323 bn of it in distribution grids
Source: IMK (Hans Böckler Foundation), 2024
ABOUT
25%
of jobs in the energy sector must be refilled within ten years
Source: BDEW
HOW IT WORKS
  1. 01 · SITUATION

    The grid is getting hard to read.

    Wind and PV feed in with fluctuations. Lines run at their limit. Without a forecast, all you can do is react.

    Keep scrolling ↓
  2. 02 · PREDICTION

    Congestion at node 7 in 3 h.

    The forecasting engine predicts generation and consumption per grid node and detects the congestion before it occurs.

  3. 03 · EXPLANATION

    Main cause: wind feed-in.

    SHAP and LIME show which factors drive the forecast. Rule extraction turns this into rules you can check.

  4. 04 · DECISION

    A human decides.

    XAIGrid recommends an action and explains why. The operator approves or rejects it.

  5. 05 · AUDIT LOG

    Every decision is documented.

    Recommendation, reasoning, factors, decision and timestamp go into the audit log with a checksum. Any later change becomes detectable.

STATE · REACTIVE
3 lines above 90% loading. Cause unclear, no forecast.
THE SOLUTION

Four layers. One traceable path to a decision.

REGULATION

Grid operators use AI in critical infrastructure. We support them with the explanations and documentation they need as evidence.

since 1 Jan 2024
§14a EnWG
Grid-oriented control of controllable consumer devices under the German Energy Industry Act.
in force
since Dec 2025
German NIS2 Implementation Act
Reporting duties of 24 h / 72 h for operators of critical facilities.
in force
since 2 Aug 2026
EU AI Act · transparency
Transparency duties for certain AI systems under Art. 50.
in force
2 Dec 2027
EU AI Act · high risk
High-risk duties for AI used as a safety component in critical infrastructure (Annex III).
next
from 2029
Grid tariff reform AgNes
The new grid tariff system takes effect.
upcoming
SECURITY & DATA PROTECTION

Your grid data stays under your control.

01
On-premise ready
Runs in your own infrastructure, even without an external connection.
02
Data stays with the grid operator
In on-premise operation, grid model, measurements and audit log never leave your premises.
03
Hosting in Germany
If you choose a hosted option, it runs in German data centres.
04
Aligned with NIS2 and §5c EnWG
Architecture and processes follow the requirements for critical infrastructure.
TEAM

Who is behind XAIGrid.

A team from electrical engineering, applied AI and research.

Belal Abulabn
Belal Abulabn
FOUNDER · CEO
Software and AI developer. M.Sc. in Electrical Engineering and Information Technology.
LinkedIn ↗
Ali Abdullah
Ali Abdullah
FOUNDER · CFO · FINANCE
Consultant for AI and intelligent automation.
LinkedIn ↗
Mertcan Kaya
Mertcan Kaya
FOUNDER · MARKETING & COMMUNICATIONS
AI specialist.
LinkedIn ↗
PROFESSOR ON THE FOUNDING TEAM
Prof. Dr.-Ing. Kai Michels
Prof. Dr.-Ing. Kai Michels
FOUNDER · PROFESSOR, UNIVERSITY OF BREMEN
Has headed the Institute of Automation Technology at the University of Bremen since 2010. Researches AI-based control and island grids with renewable generation.
LinkedIn ↗
PARTNER
MozaicAI

AI consulting and development from Oldenburg. Our research partner for explainable AI.

mozaic-ai-solutions.com ↗
FAQ

Frequently asked questions

Does XAIGrid replace our control system?

No. XAIGrid is an explainability and audit layer next to your control system. Switching commands and grid operation stay where they are today.

How does XAIGrid connect?

Via CIM (IEC 61970/61968) and OpenInterface. We are building the integration with our first pilot partners. XAIGrid reads the grid model, measurements and forecasts. Recommendations and explanations appear in the XAIGrid interface. Switching commands stay in your control system.

Where is our data stored?

With you, in on-premise operation. Grid data and the audit log never leave your premises. With the managed service, they are stored on dedicated servers in Germany.

Can XAIGrid run on-premise?

Yes. XAIGrid is designed to run in your own infrastructure.

Who makes the decision?

Always a human. XAIGrid recommends and explains. Your staff approve or reject, and the decision is logged.

What does the audit log for AI decisions do?

For every recommendation it stores the reasoning, contributing factors, the human decision and a timestamp. Checksums chain the entries together. This makes later changes detectable.

How does XAIGrid help with the EU AI Act?

AI used as a safety component in grid operation counts as high-risk AI. The duties apply from 2 December 2027. They include logging, transparency and human oversight. XAIGrid supports you with explanations and a complete audit log. It does not replace a legal assessment of your system.

How does a pilot work?

An initial call, scoping of grid area and data access, a supported pilot phase and a joint review. We agree the details with you.

FOR INVESTORS AND FUNDERS

The EU AI Act requires logging, transparency and human oversight for high-risk AI. Talk to us about funding, grants and partnership.

Contact for investors
CONTACT

Request a pilot call

You talk directly to the founder. In 30 minutes we work out whether a pilot makes sense for your grid area.

Belal AbulabnFounder, your contact