Buyer's guide
Power prices in the Baltic and Nordic markets now move every fifteen minutes, swing below zero on windy days and spike when the wind drops. Any asset that can shift when it uses or delivers energy, an electric boiler, a heat pump with a tank, a battery, earns most of its value in a few hours a day. This guide explains who can turn that into money with NordShifts, and how we work together.
01
District heating companies are adding electric boilers, large heat pumps and storage tanks to replace fuel. Their economics depend on running them in the right hours and bidding spare capacity into reserve markets, without ever risking heat supply.
02
Since the Baltic grids synchronised with continental Europe in February 2025, the three states procure their own balancing reserves. Prices for that flexibility have been high and are now falling as batteries connect. Owners and lenders need to know where revenue will settle.
03
Optimisers and traders all claim strong results. Banks and boards increasingly ask for evidence that can be checked. NordShifts publishes every forecast before the market closes and seals it in a fingerprint chain, so its record cannot be rewritten.
These charts are computed from the platform's own models and public data and update as the data does. Each is scored on data the models never saw.
Error of each model forecasting Helsinki's real district heat the day before, with that morning's weather forecast, over more than two years out of sample. Lower is better; the dot shows how often the 80% range held (target 80%).
Results are being computed.
What 1 MW of a 2-hour battery could earn per year from day-ahead price swings with perfect foresight, last 12 months, in euros. An upper bound, measured the same way in every market.
Results are being computed.
Battery value of the Kalman nowcast on two unseen weeks per zone, euros per MW per year.
Results are being computed.
Seven kinds of organisation get value from the platform. Find yours to see what you would receive, what we need from you and the best first step.
Municipal and private heat utilities with electric boilers, heat pumps, CHP and hot-water storage.
Best first step
A replay of last year on your own data.
The problem
Deciding every day which unit makes heat, when to charge the tank and how much boiler capacity can be offered as reserve, with demand that depends on tomorrow's weather.
What you get
What we need
12 months or more of hourly heat output, the list of units with capacities and efficiencies, and your fuel and grid costs.
Food, paper, chemicals and other sites that use process heat and are adding electric boilers or heat pumps.
Best first step
A sizing study on one year of meter data.
The problem
Electrifying heat raises the electricity bill unless the boiler runs in cheap hours and the site's grid connection and tariff are respected.
What you get
What we need
Hourly heat or steam demand, the electricity meter export and the grid tariff.
Owners of grid-scale batteries, solar and wind parks considering storage.
Best first step
An audit of the last three months.
The problem
Knowing whether the optimiser captures the market, and how revenue will change as more batteries connect.
What you get
What we need
The asset's dispatch export (power per interval) and, if available, the operator's revenue statement.
Banks, infrastructure funds and family offices financing storage or heat electrification.
Best first step
One asset, one quarter, as a sample report.
The problem
Revenue cases come from the parties that benefit from them. Independent, checkable evidence is scarce.
What you get
What we need
The borrower's dispatch and revenue exports, shared with their consent.
Companies that pool flexible assets and bid them into reserve and balancing markets.
Best first step
A two-site trial feeding your bidding process.
The problem
Adding heat assets to a portfolio needs reliable forecasts of what each site can offer, hour by hour.
What you get
What we need
Site lists and meter data from the sites you represent.
Engineering consultancies, heat pump and battery installers, ESCOs.
Best first step
Run three customer files with us.
The problem
Every proposal needs a credible saving figure built on the customer's own data, quickly.
What you get
What we need
Your customer's meter file and tariff.
City energy departments, regulators, universities and public agencies planning heat electrification.
Best first step
A scoped study with a fixed deliverable.
The problem
Plans for electric heat need hourly evidence about prices, demand and the value of storage, not annual averages.
What you get
What we need
Depends on the question; often public data is enough.
The platform does not add a new revenue stream by itself; it makes existing assets earn more and cost less. These are the levers, each tied to a module you can test on your own data.
01
Electric boilers, heat pumps and storage tanks run when power is cheap and the tank carries heat to the dear hours. The whole-plant optimiser shows the saving on your past year.
02
Spare boiler capacity is offered to reserve markets only where the demand range shows heat supply is safe, so income does not come with supply risk.
03
Every MW of heat forecast wrongly is made up with the most expensive unit at the last minute. A better day-ahead forecast and an honest range cut that cost.
04
CHP and boilers are planned across the day with their minimum loads and start costs, so units start less often and run where they are most efficient.
05
Before buying a boiler, heat pump, tank or battery, the replay shows what each size would have earned on your own load and real prices.
06
Monthly reports compare what assets earned with frozen forecasts and hindsight, so financing and budgets rest on checked numbers.
There is no single best model for every plant. A small network with short history needs something different from a city system with ten years of data. Every model below takes the same inputs, is scored the same way, and can be tuned.
Strength
Every coefficient readable; trains in seconds; stable with little history.
Best for
First pilots, regulators and anyone who must explain each number.
Can be tuned: alpha
Strength
Strong on tabular data, handles holidays and non-linear behaviour, explainable by feature importance.
Best for
Plants with one to three years of hourly data.
Can be tuned: max_iter · learning_rate · max_leaf_nodes
Strength
Shapes all 24 hours together and gives the most reliable uncertainty range in the reference study.
Best for
Reserve bidding, where the range matters as much as the forecast.
Can be tuned: hidden · dropout · seeds · noise
Strength
Deeper network in which each block explains part of the forecast; often better on unusual days.
Best for
Large systems with long history and irregular days.
Can be tuned: hidden · blocks · dropout · seeds · noise
Strength
Reads weather and heat hour by hour like the network's own inertia; strong on cold snaps and thaws.
Best for
Systems with long pipes and slow response.
Can be tuned: hidden · dropout · seeds · noise
Strength
Sees several days at once with dilated convolutions; faster than the recurrent network.
Best for
Plants with strong daily routines, such as industry.
Can be tuned: channels · dropout · seeds · noise
Strength
Self-attention links any hour of the last three days to any hour of tomorrow, learning long-range patterns.
Best for
Large datasets with complex weekly and holiday patterns.
Can be tuned: d_model · layers · heads · dropout · seeds · noise
Strength
Averages several deep models' forecasts; usually the steadiest accuracy and range.
Best for
Daily operation, where reliability matters more than training time.
Can be tuned: use_mlp · use_residual · use_tcn · use_gru · use_transformer · use_boosting
All sources are public or come from the client. Public sources are free and their terms allow use with attribution.
01
We learn your assets and the decision you want to improve, and agree whether the data exists to support it.
02
You send an export under an NDA if you wish. Within a week we confirm in writing what the data can and cannot support.
03
We replay the past year (or audit the past months) on your data and show the result, including where the models are wrong.
04
If the numbers hold, the plan or report runs every day or month, with the AI agents drafting explanations for your team to approve.
The same problem, assets that must be scheduled against volatile power prices, is growing fastest in markets that are adding batteries, solar and electric heat or cooling. These are the markets we see as natural next steps.
District cooling is heat dispatch in reverse: chilled-water storage is charged when power is cheap and discharged at the afternoon peak. The same optimiser and forecasting models apply, which opens hot-climate markets as well as cold ones.
Market
A day-ahead exchange (JEPX) with nine price areas; prices fall to the floor in sunny hours in the south and west.
Why now
Long-term decarbonisation auctions are funding grid batteries, and solar curtailment creates daily swings.
Where NordShifts fits
Battery revenue audits, solar-hour scheduling, heat pumps in industry.
Data
JEPX prices are now in the platform's comparison.
Market
A cost-based single price (SMP) run by the power exchange KPX; a real-time market is being piloted on Jeju island.
Why now
Government tenders contract multi-gigawatt-hour battery storage for 15 years, and large district heating systems run on CHP.
Where NordShifts fits
Battery dispatch and audits, district heating with CHP and storage.
Data
Hourly SMP is published through the national open data portal (free key).
Market
A half-hourly wholesale market (USEP), mostly gas-fired, with an active national AI strategy.
Why now
One of the world's largest underground district cooling networks, at Marina Bay, uses thermal storage; data centres need efficient cooling.
Where NordShifts fits
District cooling and data-centre cooling dispatch against half-hourly prices.
Data
Monthly averages are public; half-hourly prices from the market operator.
Market
No wholesale market: the utility buys and dispatches centrally.
Why now
Solar is planned to reach about 4 GW by 2030, and Qatar runs some of the world's largest district cooling plants with thermal storage.
Where NordShifts fits
District cooling optimisation and solar forecasting; sovereign and strategic investors in energy technology.
Data
Project-based data from operators; no public spot price.
Market
A 5-minute real-time energy-only market across five regions, with open data.
Why now
The fastest battery build-out of any large grid: batteries now set the price in about a third of intervals.
Where NordShifts fits
Independent revenue audits for owners and lenders, real-time price nowcasting.
Data
AEMO prices are now in the platform's comparison.
The core models, data pipelines and public track record exist today. Investment turns them into a product that runs inside clients' operations and travels to new markets.
First 6 months
6 to 18 months
18 to 36 months
No. A pilot runs on exported files. Daily operation can use the API or a scheduled file exchange.
On servers in the EU. Files are used only for your analysis, are never shared, and are deleted on request. An NDA is available before you send anything.
See the heat forecast lab and the accuracy pages: every model is scored on data it never saw, against simple baselines, and the live record is public.
A pilot has a fixed fee agreed before work starts. Ongoing use is an annual subscription sized to your plant or portfolio.
The Baltic and Nordic bidding zones in depth, with data for Germany, France, Italy, Great Britain and other European zones.
If you run, finance or advise heat or storage assets, the fastest way to judge this is to see it on your own data. NordShifts is built by Girts Liepins, a data scientist working on Baltic and Nordic power markets. If you run heat or storage assets, finance them, or advise those who do, write and we will reply within two working days.
To make the first reply useful, please include: