Who uses how much electricity?
The plant's own consumption split across consumers, even when not everything is measured.
To the pageEnergy and plants · machine learning
Your plant answers. With numbers that are right.
The machine learning system answers questions about your plant in plain sentences: from the measurement data of your control system, from your documents and from our expert knowledge. Every number comes from a fixed calculation, not from the language model. Example from practice: a 1.5-megawatt biogas power plant in Lastrup.
The demo of the example plant runs behind a login. Access is given to interested parties on request.
To the demo login
Your concerns
Five questions operators ask. Five pages on them.
Each page answers one concern from daily plant operation: with the learning system (RAG) that knows your documents and measurement data, and with the Pseudomizer if you also want to use a cloud AI.
What to feed, and when?
Feeding recommendation next to the electricity prices: substrates, amounts and times.
To the pageWhen should the CHP run?
Electricity price outlook as a heat map, to plan operation against the price.
To the pageMay this go to the cloud?
Original versus mirror: use cloud AI without originals leaving the house.
To the pageWhere does the heat go?
Heat network at a glance: flow, return, output, storage. One click opens the chart.
To the page
The control room answers
Questions about the plant in plain sentences, with numbers from fixed calculations and with a source.
The principle
Numbers from calculations, knowledge with sources.
A language model is good at phrasing but bad at calculating. That is why the machine learning system splits the work.
Fixed calculations
How much electricity was it yesterday? Did the plant run price-driven? What did the repair cost? For each of these questions there is a fixed calculation directly from your measurement data and receipts. The language model only picks the right calculation and phrases the answer. It does not calculate and adds no number.
Answers with a source
Plant knowledge, contracts, analyses, receipts and official mail sit in one searchable collection. Every answer states where it comes from, with a link to the document. So you can check the passage in the original.
A guardian checks along
A second system continuously looks at the new answers and reports anomalies, for example swapped numbers. It only reports and changes nothing, so errors do not disappear unnoticed.
Videos
See how the machine learning system works.
Short films of about one minute. They show sample data and an example plant, no customer data.
The overview: four plant films
How the system works: nine episodes from the series “Machine learning biogas”
Examples
Questions as you would ask a colleague.
These are example questions, not answers. Which questions are possible depends on the data connected to your plant.
How much electricity did the engines generate yesterday?
Did the plant run price-driven in recent weeks?
Which faults were there on the agitators?
When is the next inspection deadline due?
What does the contract say about direct marketing?
Make me a report with charts on heat use.
From the meter to the heat network
Measurement data, documents and expert knowledge queryable in one place.
What belongs to it
More than a chat.
Reports with charts
On request a separate document with charts and tables is created, to pass on or file.
Questions by voice
You can also speak your question to the system and listen to the answer.
Learning at night
At night the system asks itself new questions, checks the answers and reports gaps in knowledge. This makes visible what is still missing.
Every role sees its area
Access is by login. Which areas and functions someone sees depends on their role.
Digital twin
Tanks, engines and heat at a glance. The machine learning system can be queried about it.
Guardian and deadlines
Gas values, engines, inspection deadlines and messages are monitored, anomalies reported early.
Data protection and language model
You decide where the language model runs.
The system works with a language model in the house or, if you wish, with a cloud service. If the cloud service fails, the model in the house keeps answering, just a little slower.
In the house
The language model runs on your own device. Questions and excerpts of your data do not leave the business.
With a cloud service
With a cloud model the question and the excerpt needed for the answer are transmitted to the service. We discuss this with you beforehand and put it in a contract.
More tools for energy producers
Calculate, plan, optimise.
Besides the machine learning system there are separate tools for biogas and energy plants, farms and industry, with the local AI as the core.
Schedule optimisation
Run flexible plants by exchange electricity price and compare your own schedule with that of the direct marketer.
Flex assistant
What does flexibilisation bring? Calculates your plant with your numbers.
To the flex assistantSubstrate calculator
Plan feedstocks, quantities and gas yield and calculate alternatives.
Learning
How does the system get better?
At night the learning system asks itself questions and reports gaps. A person adds what is missing.
Learning overnight in 30 seconds00:00
Own questions–
Gaps reported–
Gaps closed–
Example with invented values
The way to your plant
From the first question to a running system.
01
Conversation
We look at your plant and your most important questions.
02
Connect data
Measurement data from control system, meters and sensors is connected.
03
Read in documents
Contracts, analyses, receipts and inspection reports go into the knowledge base.
04
Test and operation
We check with your questions whether the answers are right, then it goes into daily use.
What this costs for your plant and how long it takes we will tell you after the conversation.
Become a partner
You know your industry. We bring the local AI.
AI only helps if it fits the daily work of an industry. That is why we do not build industry solutions alone, but with partners who know that daily work.
We are looking for: plant operators, direct marketers, metering and energy consultants.
What you bring
- You know the daily work of the industry and where time is lost.
- You have the trust of your customers or colleagues.
- You tell us what is really needed in practice.
What we bring
- The local AI: your documents stay in the house.
- Setup, connecting the documents and training.
- Support and joint further development with you.
01
Get to know each other
A short conversation: what does your industry need?
02
Joint pilot
We test with a real business and its own documents.
03
Offer it together
We agree the terms in conversation.
Bring us your biggest annoyance.
One conversation, one real task from your daily work. We use it to show what AI and automation can do for you.