Manufacturing
Shifts, presses and furnaces that peak at the worst tariff.
Shifts flexible loads into solar hours and pre-charges the battery before each peak.
−21% peak demandillustrative
AI energy agents for industrial sites
Your energy, running itself.
Fieldline forecasts your solar output and demand, then decides every 5 minutes whether to store, use or sell each kilowatt-hour.
Running 1.9 GWh a year across 40 pilot sites. Illustrative
Placeholder marks · fictional companies
The problem
Your site makes energy at the wrong time.
On-site solar should cut the bill. Without a plan that changes every few minutes, most of it is sold cheap and bought back dear.
12:00
Solar spills at noon.
Your panels peak while the line is on lunch break. The surplus leaves for the grid at €0.04 per kWh.
18:00
You buy it back at six.
Second shift starts as the peak tariff hits. The same energy comes back at €0.31, nearly eight times the price.
MON 07:30
The plan lives in a spreadsheet.
Someone rebuilds the week's schedule by hand from last month's bill. By Wednesday, the weather has made it wrong.
Tariffs illustrative
How it works
Sense. Predict. Act. Every 5 minutes.
01Sense
Every inverter, meter and battery reports in.
An edge agent reads your SCADA, inverters and meters every 5 seconds, so the model starts from what your site is doing right now, not last month's bill.
02Predict
A forecast for every 5 minutes of tomorrow.
Weather, production schedules and tariff feeds become a 24-hour forecast of solar output and demand, with a confidence band your team can check.
03Act
Store, use or sell, decided for you.
Agents route each kilowatt-hour to the battery, the line or the grid, inside limits you set. Every decision is logged, explained and one click from override.
The platform
Every decision, on one screen.
Your energy manager sees what the agents decided, why, and what it saved. Nothing happens that your team cannot read, question or reverse.
Halden Works · Site 04Tue 14 Oct · 12:05
Agents runningIllustrative data
Agent decision log
every 5 min- 12:05STORE38 kWh → Battery B2Solar 2.31 MW > load 1.74 MW; peak in 5 h 55 m
- 12:00USELine 2 on solarShift press warm-up 20 min early
- 11:55HOLDBattery at 61%Cloud cover 14:10, forecast −0.4 MW
- 11:50SELL120 kWh → gridExport price spike €0.19 for 10 min
- 11:45STORE52 kWh → Battery B1Cold room pre-cooled to −22 °C
Today's solar forecast
confidence 94%actualforecast
Saved so far
€1,284
today · €38,410 YTD
Battery
61%
Solar self-use
92%
was 61% before agents
Results
Measured on the meter, not in a model.
Illustrative figures for a fictional product
- Energy cost
- −18%Average cut in the first year across pilot sites.
- Solar self-use
- 92%Of on-site generation used on site or stored, up from 61%.
- Saved every week
- 11 hOf manual scheduling and spreadsheet work per site.
Use cases
Built for sites that run all day.
Integrations and security
Fits the equipment you already run.
No rip-and-replace. Fieldline reads the protocols your site speaks and writes back only through the controls you approve.
Solar inverters
SunSpec · Modbus TCP
Battery systems
BMS over Modbus · CAN
SCADA & PLC
OPC UA · MQTT
Utility tariffs
Day-ahead · time-of-use feeds
Smart meters
IEC 62056 · pulse
Weather
Satellite irradiance · NWP
ERP schedules
CSV · REST import
Your BI
Read-only API · webhooks
Data stays on site
The edge agent runs on your network. Only forecasts and decisions leave it, encrypted.
Audit log of every decision
Each action records its inputs, the reason and who could override it.
Limits your team sets
Battery floors, temperature bands and export caps the agents can never cross.
One-click override
Hand any asset back to manual control at any time, from the dashboard or the PLC.
“We stopped arguing about the spreadsheet. The agents move the battery and the shift loads, and I check the log over coffee. The six o'clock peak used to be my problem. Now it's a line in a report.”
Book a demo
See your site run itself.
Send a month of meter data and we will replay it through Fieldline: what the agents would have decided, every 5 minutes, and what it would have saved.