Glossary

Organisational psychologist at Wharton. 'Give and Take' shows that the people who give most end up both at the top and at the bottom, and the difference between them is worth understanding.

Development methodology where autonomous AI agents plan and execute multi-step tasks without manual intervention at each step.

Artificial Intelligence - technology that enables machines to perform tasks that normally require human intelligence, such as language understanding, pattern recognition, and decision-making.

An AI system that independently plans and executes multi-step tasks by using tools, making decisions and adapting along the way.

An infrastructure layer sitting between applications and AI models, handling authentication, logging, rate limiting and model selection.

Person learning a trade through practical work combined with theoretical training - a key part of vocational education in Norway.

Communication or programming pattern where processes do not need to wait for each other - used in message-based systems and event-driven architecture.

Power that Statnett buys and sells in real-time to maintain system frequency stability at 50 Hz.

Researcher at the University of Houston. Vulnerability is not weakness, but uncertainty, risk and exposure without a guaranteed outcome. Without it there is no trust.

Balance Responsible Party - actor responsible for balancing production and consumption in its portfolio against the system operator.

Elhub’s process specifications. Each BRS describes one market process step by step - who sends what, when, and what happens when something fails.

Systematic measurement and reporting of greenhouse gas emissions - divided into scope 1 (direct), scope 2 (purchased energy) and scope 3 (value chain).

EU directive on resilience of critical entities - sets requirements for protecting critical infrastructure against physical threats.

Anthropic's AI coding assistant - integrated into the terminal and IDEs for agentic software development.

Approach to development where applications are designed from the ground up for cloud infrastructure - containers, microservices, and automated scaling.

Domain knowledge and technical skills - a core value for Tvimenning and a prerequisite for delivering quality in the energy sector.

A maintenance tactic: continuously monitor a component's condition and intervene when the data shows a deviation - not on a fixed calendar. What FRAMSYN does in practice.

The maximum amount of text an AI model can "remember" and process in a single conversation or operation.

Microsoft's brand for AI assistants integrated into products such as Teams, Word, Excel and Azure - powered by underlying LLM technology.

Corporate Sustainability Reporting Directive - EU directive requiring detailed sustainability reporting from large companies starting 2024.

Software built specifically for one organisation - gives full control, but costs more to maintain than off-the-shelf solutions.

Overview and search system for a company's available datasets, including metadata, ownership and data quality.

Architecture paradigm where data responsibility is decentralised to domain teams and data is treated as a product with clear ownership.

Formal representation of data structure, relationships and constraints - foundation for database design and system messages.

Secure digital environment for sharing sensitive information during M&A processes, audits or due diligence.

Containerisation platform that packages applications and dependencies in isolated, reproducible units.

The third step in VEE - manual correction of metering values that cannot be resolved automatically, with full audit trail for all changes.

A Norwegian excise duty per kilowatt-hour consumed, collected by the grid company on the grid rent invoice and paid to the state.

Proactive role at DSOs (such as Lnett, Elvia and Tensio) acting as advisor and liaison between the grid company and customers - mapping future capacity needs and optimising grid capacity through dialogue with property developers and municipalities.

Environmental, Social and Governance - framework for measuring and reporting sustainability in business.

EU's classification system for sustainable economic activities - determines which investments can be labeled 'green'.

Forvaltning, drift og vedlikehold: asset management, operation and maintenance. The Norwegian collective term for everything that has to follow an installation through its life: what is there, how it is run and what has been done to it.

The process of taking a pre-trained AI model and training it further on a specific dataset to improve performance in a bounded domain or task type.

Tvimenning’s tool for condition-based maintenance: raises work orders from how the component is actually doing, not from how old it is.

Google's family of large language models (LLM) - integrated in Google Cloud and Workspace for AI-assisted work processes.

Google's AI-powered note-taking tool, called NotebookLM until July 2026 - analyses your own documents and answers questions based on them.

The transition from test to live operation against Elhub. The final step in the approval process, and the only one that cannot be undone without someone noticing.

Google's web-based development tool for experimenting with and building applications on Gemini models.

Google's suite of productivity tools - Gmail, Docs, Drive, Meet and more.

NVE's framework for grid tariff calculation - contains fixed, energy, and capacity components to fairly distribute grid costs and encourage balanced consumption.

When an AI model produces seemingly credible answers that are factually wrong, fabricated or without basis in reality.

The framework that wraps, drives, tests and constrains an AI agent. The model supplies the power; the harness decides where it goes.

Swedish software company, founded in Linköping in 1983. Known for ERP and maintenance solutions aimed at asset-heavy organisations, and widely used among Nordic energy and industrial companies.

Open source container orchestration platform - automates deployment, scaling and management of containerised applications.

Lightweight Directory Access Protocol - protocol for accessing and maintaining directory services such as Active Directory.

Large Language Model - AI model trained on large text data that can understand and generate human language, such as GPT-4 and Claude.

An agreed flow of events between the actors in the power market, such as a supplier switch or a move-in and move-out, with defined messages, deadlines and failure cases.

The authoritative source data in a System of Record (Master) - data owned, maintained and edited in one place, from which all other systems retrieve copies.

Open protocol from Anthropic enabling AI agents to communicate with external tools and data sources in a standardised way.

EDIFACT message format for exchange of metering data in the electricity market - used in Elhub to transmit hourly, daily and monthly values between market participants.

Elbits' openly available reference model of the Norwegian power grid - a simplified national grid model that lets data from different systems be tied together.

Software with openly available source code that can be used, modified, and distributed freely.

Notice that something in operations deviates - planned maintenance, a fault or reduced availability. Should reach those affected before they notice it themselves.

Oracle's low-code development platform for building web applications directly against Oracle Database - without a heavy application server.

Development practice where two developers work together at one machine - one writes code, one reviews and thinks - continuously and synchronously.

Concrete and detailed information about the power supply that can be used to harm it. Subject to a duty of confidentiality under the Norwegian power emergency preparedness regulation, section 6-2.

Settlement against an assumed consumption curve instead of actual hourly readings. The counterpart to hourly settlement.

The technique of formulating precise instructions to AI models to obtain relevant, reliable, and accurate responses.

Cloud infrastructure operated by third parties and available to everyone over the internet.

Retrieval Augmented Generation - AI architecture where the model is supplemented with documents retrieved in real-time to provide fact-based answers.