In May 2025, we joined our client Helen on stage at the Data Innovation Summit in Stockholm to present the work we've been doing together over the past few years. This blog is based on that presentation and describes the collaboration behind it.

Helen, an energy company with a long history in the Helsinki region, is currently going through a significant transition: shifting from combustion-based energy production to a carbon-neutral energy system. This change has required new capabilities both in terms of physical infrastructure and how data is used to support planning, operations, and optimisation. Brightly has been working with Helen to build and maintain the data platform that supports this shift.

Moving away from combustion-based production

Helen has operated in Helsinki for over a century. Today, the company manages electricity production and distribution, district heating, and a range of energy services. In early 2025, coal-based energy production was shut down permanently. The target is to be carbon neutral by 2023 and to end all combustion by 2040. With this change, the energy will come from sources such as wind, solar, bioenergy, hydro, and potentially small modular nuclear reactors and hydrogen.

This shift makes energy production more weather-dependent and variable, which increases the importance of being able to forecast demand and adjust production accordingly. Meeting these requirements without a centralised data infrastructure would be impossible.

The role of the data platform

Helen’s current data platform was built to support a wide range of use cases: from operational monitoring and analytics to optimisation models and real-time control systems. It brings together data from sensors, control systems, internal business platforms, and external data sources like electricity market pricing.

The platform was designed to be scalable, with the ability to support new use cases as the business and energy systems evolve. One of the main goals has been to ensure that data is accessible across the organisation, not only to technical teams, but also to analysts, product owners, and business units.

Together with Helen’s teams, Brightly has supported the development of this environment to make it robust and manageable for long-term use.

HeatFlex: A use case for demand-side flexibility

One concrete example of how the platform is used is HeatFlex, a solution that helps manage demand-side flexibility in the district heating network.

As heating moves toward electric boilers, production costs become more dependent on electricity prices. These prices can vary significantly throughout the day. At the same time, many buildings in the Helsinki area already have the ability to store heat for short periods essentially functioning as distributed thermal storage.

HeatFlex takes this storage capacity into account. When electricity is cheaper, buildings are heated more actively; when prices rise, heating can be briefly reduced without affecting indoor comfort. This allows Helen to reduce peak load and manage costs more effectively.

To enable this, the system must understand each building’s heating characteristics, weather conditions, and market signals and then use that information to calculate the right control actions.

Platform architecture in practice

The technical setup includes IoT devices installed in buildings, which send telemetry data to the cloud using Azure IoT Hub. Data is then routed through Helen Streams, a stream management layer developed for real-time processing.

Helen Streams is configured using DevOps practices. Data pipelines, access rules, and monitoring features are defined as code, which simplifies deployments and maintenance. This allows new use cases to be added in a controlled and repeatable way.

Data flows into Helen’s central platform, which uses Databricks and Azure. The architecture follows the medallion pattern and includes tools like Unity Catalog for governance and discoverability. These components help manage data quality and make data easier to work with at scale.

The setup supports both real-time and batch workloads and serves as the foundation for analytics, modelling, and operational solutions.

Modelling at the building level

HeatFlex uses building-level machine learning models to predict heating behaviour. Each building gets its own model, trained on historical data. This approach allows the system to take into account building-specific characteristics without needing manual configuration.

The modelling process includes three main stages:

• Daily model training: Updates building-level heat models using historical data
• Hourly flexibility assessment: Estimates how much flexibility is available in each building for over- or under heating
• 10-minute control loop: Applies the optimal radiator settings in real time and sends control signals to buildings

This continuous loop ensures that the system stays aligned with current conditions and reacts to changes in both demand and energy prices.

Supporting self-service and scalability

Another focus area has been enabling more users within Helen to work with data directly. Instead of relying only on central data teams, analysts and developers are being given tools to access, process, and use data independently.

Brightly has supported this by helping define shared data products and simplifying access to curated datasets. These changes help reduce bottlenecks and make it easier for different teams to explore data and develop their own solutions.

This self-service approach is seen as a key enabler for Helen’s long-term goals, particularly as energy markets and production methods continue to change.

Reflections from the journey

The key learnings from the past few years of platform development are:

Start small: Early results help build momentum. Moving to cloud and demonstrating small-scale use cases early on helped bring stakeholders on board.

Accept iteration: Not everything needs to be perfect from the start. It’s been more important to get started, test ideas, and improve the platform as needed than to aim for full coverage from day one.

Keep things practical: A platform is only useful if it supports actual decision-making. The value of the data platform is tied to its ability to help align daily work with broader strategic goals.

Next steps

The work around HeatFlex is ongoing, and Helen is expanding the same flexibility concept to other areas, such as electricity consumption, electric vehicle charging, and hydrogen-based heating.

For our part at Brightly, we’ve worked closely with Helen’s teams to make sure the data platform meets practical needs, is technically sound, and stays aligned with the company’s evolving goals. Brightly continues to support Helen as a strategic partner in building flexibility capabilities across different use-cases and building the shared control framework for all of the energy optimization areas. At the same time, self-service capabilities are being developed further to support more teams and use cases across the company.