Digitalization in the Automation Domain�Oslo, 22 October 2025

Åpen informasjon / Public information

• Master in Computer System (MEng) with European Studies from University of Bristol.
• Worked with software solution for the electrical power system since 1996:
• Customer Relationship Management (CRM), Enterprise Resource Planning (ERP), Energy Trading System, Energy Data Management (meter clearing and change of supplier), Energy Trading and Risk Management (ETRM), Energy Management System (EMS), Power System Analysis, Transient Dynamic Stability
• Member of IEC TC57 Power systems management and associated information exchange
• WG13 Software interfaces for operation and planning of the electric grid
• WG14 Enterprise business function interfaces for electric utility operations
• IEC TC57 Outgoing Liaison
• ISO/TC 59/SC 13 Development of building data related standards – Building Information Model (BIM)
• ISO/TC 184/SC 4 Industrial data – Industrial Data Ontology (IDO)
• Member of SyC Smart Energy/WG 6 Generic Smart Grid Requirements
• ENTSO-E Convenor for CIM WG:
• Common Grid Model Exchange Standard (CGMES) Subgroup
• RCC Network Code Data Exchange (NCP) Subgroup

Svein Olsen

• Enterprise Information Architect
• Statnett SF

• Norsk Elektroteknisk Komité (NEK)

​

• European Network of Transmission System Operators for Electricity (ENTSO-E)

​

• International Electrotechnical Commission (IEC)

Åpen informasjon / Public information

Åpen informasjon / Public information

Introduction & Objective

• Introduction to the R&D project Talk2PowerSystem
• Introduction to the IEC Common Information Model (CIM) ontology
• What kind of instance data do the power system model include and how do we want to managed them
• What type of challenges are the uses of the data having – can Knowledge Graph (KG) and Large Language Model (LLM) help
• Simple Architecture and agentic AI explanation
• Example of quires using reasoning
• What other type of question can we ask the agent AI

Åpen informasjon / Public information

Statnett and Graphwise R&D project

Background

Power system data are getting more and more complex; the average experience of power engineers is dropping due to retirement. Electrical data must be used not only by power engineers with deep domain knowledge, IT skills and Common Information Model (CIM) knowledge, but also by other stakeholders and decision makers. Transitioning to a data-driven approach necessitates high-quality, well-defined metadata to ensure data clarity, consistency, and usability.

Goal

Use state-of-the-art semantic web and knowledge graph technologies to train large-language models (LLMs) to empower power system engineers and stakeholders to intuitively interact with complex CIM-based data using natural language (NLQ) supported by advanced AI methodologies.

Outcomes

Openly accessible CIM and Q&A datasets, tools for interfacing with LLMs, SPARQL & GraphQL querying, evaluation/validation framework, harmonization with other relevant standards (BIM, IEC 61850, W3C).

Expected Benefits

Intuitive and simplified access to CIM-based power system data. Enhanced decision-making through accurate, explainable, and trustworthy insights. Improved semantic interoperability across European and international standards. Increased operational efficiency and faster decision making to support the renewable and digital transition. Mitigation of knowledge loss due to expert retirement through systematic knowledge capture. Leverage semantic reasoning to infer implicit data relationships, reducing query complexity and manual modeling efforts.

​

​

TRL

2-7

Project period

Q1/2025 – Q2/2026

Åpen informasjon / Public information

IEC CIM

• IEC – International Electrotechnical Commission
• CIM – Common Information Model

​

Common Information Model (CIM)

• Developed by IEC TC57 (WG13, WG14, WG16) for power system data exchange.
• Standardised under IEC 61970 (EMS), IEC 61968 (DMS, DER, AMI), IEC 62325 (Market).

Åpen informasjon / Public information

CIM Background

• Smart Grid Architecture Model (SGAM)

Åpen informasjon / Public information

CIM description of a Wind unit impact on the Power System

© https://www.electronicproducts.com/

CIM description a Wind unit dynamic stability

© https://www.jvejournals.com/article/20595

© https://www.electronicproducts.com/

CIM description a Wind turbine as asset

State Estimator (SE) and Power Flow (PF) as foundation for a Power System

Power Flow (PF)

State Estimator (SE)

Contingency (CO)

Remedial Action (RA)

Capacity Calculation (CC)

Transient Stability (TS)

Weather forecast

Operational Schedules

Network Model

Observation

SCADA

(Real-Time accusation)

ModelOps: Instance Data

• Named graph per model
• Implications for Reasoning
• Model versions and ontology versions
• ModelOps, SchemaOps, MLOps in project wiki

Datasets

• Ontologies (Profiles):
• CIM Comon Grid Model Exchange Standard (CGMES) v3.0
• ENTSO-E Network Code r2.3 (r2.4)
• CIM Asset (AS) and Asset Catalog (AC) from Common Distribution Power System Model (CDPSM) v2.0

​

• Instances (Model data):
• Statnett Nordic44 (hourly snapshot for 2015 will be added)
• Telemark-120 (old DIGIN10) from ElBits
• BIM-default substation (To be added)
• ENTSO-E ReliCapGrid – Reliability and Capacity Test Model (To be added)

​

• Timeseries in Cognite Data Fusion
• Public timeseries – price and flow

Åpen informasjon / Public information

Nordic44

EQ Content:

Instance file content:

Åpen informasjon / Public information

Åpen informasjon / Public information

CIM as an Early RDF Adopter

• Before CIM:
• Power system data exchange relied on proprietary formats.
• CIM’s RDF Adoption (2000s):
• CIM adopted RDF/XML to model grid components as a semantic graph.
• Used in ENTSO-E’s CGMES (Common Grid Model Exchange Standard) for cross-border grid operations.
• Graph-based modeling enabled semantic interoperability before linked data became mainstream.
• Challenges of RDF/XML:
• Verbose, not web-friendly, limited cross-domain adoption.
• Need for modernized JSON-LD, SHACL, and federated graph alignment.

Åpen informasjon / Public information

What challenge can Knowledge Graph (KG) and Large Language Model (LLM) help us with

Relevant skill-set

© https://electricalacademia.com

SPARQL

PREFIX cimr: <https://cim.ucaiug.io/rules#>

PREFIX cim: <https://cim.ucaiug.io/ns#>

PREFIX sesame: <http://www.openrdf.org/schema/sesame#>

select ?sub1Name ?lineName ?sub2Name {

values ?sub1Name {"ARENDAL"}

?sub1 a cim:Substation; cim:IdentifiedObject.name ?sub1Name;

cimr:connectedThroughPart ?line.

?line a cim:Line; cim:IdentifiedObject.name ?lineName.

?sub2 a cim:Substation; cim:IdentifiedObject.name ?sub2Name;

cimr:connectedThroughPart ?line.

filter(?sub1 != ?sub2)

}

Åpen informasjon / Public information

Simplified Architecture

Åpen informasjon / Public information

Agentic AI

• Fully realized GraphRAG approach
• AI Agent is an LLM with a selection of tools for knowledge retrieval
• No predetermined workflow, the agent uses tools as required and appropriate for a given question
• If a tool uses is unsuccessful, it can try to fix the issue or use an alternative tool
• SPARQL or other DB reads allow relevant aggregation, summarisation, averaging etc over the full dataset

Åpen informasjon / Public information

Q: List all substations that are connected via AC-line or DC-line to substation XYZ

Åpen informasjon / Public information

Detail Model

Åpen informasjon / Public information

Reasoning Helps LLM

Query is very complex, hard to generate

Query becomes easier to generate

Add reasoning (OWL-RL-optimized)

• cim:EquipmentContainer.Equipments|cim:Substation.VoltageLevels|cim:VoltageLevel.Bays → cimr:hasPart; inverse cimr:isPart
• cim:Terminal.ConductingEquipment|cim:Terminal.AuxiliaryEquipment → cimr:Terminal.Equipment; inverse cimr:Equipment.Terminals
• cimr:hasPart+ → cimr:hasPartTransitive; inverse cimr:isPartTransitive
• cim:ConductingEquipment.Terminals / cim:Terminal.ConnectivityNode / cim:ConnectivityNode.Terminals / cim:Terminal.ConductingEquipment → cimr:connectedTo (symmetric)
• cimr:hasPartTransitive / cimr:connectedTo / cimr:isPartTransitive → cimr:connectedThroughPart (symmetric)

List all substations that are connected via AC-line or DC-line to substation XYZ

PREFIX cimr: <https://cim.ucaiug.io/rules#>

PREFIX cim: <https://cim.ucaiug.io/ns#>

PREFIX sesame: <http://www.openrdf.org/schema/sesame#>

select ?sub1Name ?lineName ?sub2Name {

values ?sub1Name {"ARENDAL"}

?sub1 a cim:Substation; cim:IdentifiedObject.name ?sub1Name;

cimr:connectedThroughPart ?line.

?line a cim:Line; cim:IdentifiedObject.name ?lineName.

?sub2 a cim:Substation; cim:IdentifiedObject.name ?sub2Name;

cimr:connectedThroughPart ?line.

filter(?sub1 != ?sub2)

}

PREFIX cim: <https://cim.ucaiug.io/ns#>

PREFIX sesame: <http://www.openrdf.org/schema/sesame#>

select ?sub1Name ?lineName ?sub2Name {

{select distinct * {

values ?sub1Name {"ARENDAL"}

?sub1 a cim:Substation;

cim:IdentifiedObject.name ?sub1Name;

(cim:EquipmentContainer.Equipments|cim:Substation.VoltageLevels|

cim:VoltageLevel.Bays)+ / # equipment in ?sub1

cim:ConductingEquipment.Terminals / cim:Terminal.ConnectivityNode /

cim:ConnectivityNode.Terminals / cim:Terminal.ConductingEquipment /

cim:Equipment.EquipmentContainer ?line. # part of ?line

?line a cim:Line; cim:IdentifiedObject.name ?lineName}}

{select distinct * {

?sub2 a cim:Substation;

cim:IdentifiedObject.name ?sub2Name;

(cim:EquipmentContainer.Equipments|cim:Substation.VoltageLevels|

cim:VoltageLevel.Bays)+ / # equipment in ?sub2

cim:ConductingEquipment.Terminals / cim:Terminal.ConnectivityNode /

cim:ConnectivityNode.Terminals / cim:Terminal.ConductingEquipment /

cim:Equipment.EquipmentContainer ?line}}

filter(?sub1 != ?sub2)

}

Åpen informasjon / Public information

Class description

Åpen informasjon / Public information

Object (instance) description

Åpen informasjon / Public information

Description the query

Åpen informasjon / Public information

Conclusion & Key Takeaways

• LLM can ease the user interaction by having a natural langue input and output.
• LLM is very good on understanding the question and explain the result to the relevant user in their language and terms
• Knowledge Graph (KG) with well-defined ontology (e.g. CIM, BIM, 61850) can reduce hallucination
• Semantic reasoning is very useful for simplify the work for the LLM to address and explain complex dataset
• A minimum solution can give a lot of value

Åpen informasjon / Public information

More information

Important activities to ensure the impact and sustainability of development efforts:

​

• Working with real problem/challenges

​

• Everything accessible on Statnett Open GitHub account in the following repositories:
• statnett/Talk2PowerSystem: Talk to the Power System
• statnett/Talk2PowerSystem_LLM: Large-Language Model (LLM) part of Talk2PowerSystem
• statnett/Nordic44: Synthetic model of the Nordic high voltage transmission system

​

• More information will be published, including video recording:
• CIM Fundamentals (1h)
• Knowledge Graph and LLM Fundamentals (1h)
• Talk2PowerSystem Presentation, Demo and discussion (1.5 h)

Generated by ChatGPT

Åpen informasjon / Public information

THANK YOU FOR YOUR TIME!

Åpen informasjon / Public information