Planning production with AI and the price of energy
What order to produce in, which shifts to run, and what times to switch on the equipment that consumes the most. A model of the factory, with its real cycle times, its constraints and the hourly electricity tariff, calculates the sequence that meets deliveries at the lowest cost, and explains it to the planner.
What a spreadsheet can't recalculate every morning
The planner knows the factory better than anyone. What they can't do by hand is cross-reference, every day, committed deliveries, each machine's real cycle times, format changeovers, the maintenance schedule and the hour-by-hour electricity price. So the sequence gets built once and tweaked whenever something goes wrong: an urgent order, a stopped machine, a batch running late.
The usual result is some energy-intensive production landing in the most expensive hours, format changeovers that could have been grouped together, and replanning hours that nobody counts.
What the model does
A model of the factory
Real process and changeover times, capacities, shifts, maintenance schedules and each piece of equipment's hourly consumption. It's built from ERP and MES data and checked against the planner before it's used.
Energy as a constraint
The hourly tariff goes into the calculation. The operations that consume the most get moved to the cheapest hours within what deliveries allow, and the sequence carries its calculated energy cost.
A sequence, explained
Every morning, the proposed order comes with the reason behind each decision: why this batch first, why this piece of equipment at this hour. The planner accepts, changes or rejects it.
Recalculated in minutes
When an urgent order comes in or a machine stops, the model recalculates the entire sequence under the new constraints and shows which deliveries shift.
Where it shows
Energy cost per tonne
The energy cost of the actual sequence is compared with that of the proposed sequence, at each hour's price. It's the number measured in the pilot.
On-time delivery
Orders delivered on time and in full, before and after. Deliveries are a constraint of the model, not a consequence.
Replanning hours
How many hours a week the planner spends reworking the sequence. They're counted at the diagnosis stage and counted again after the pilot.
Furnace and equipment occupancy
In thermal processes, grouping loads avoids keeping a furnace hot and idle while it waits. GASAI's furnace-occupancy planner grew out of exactly this problem.
From the ERP to the first pilot
We pick the line or product family where it hurts most: the one that consumes the most energy or gets replanned the most.
Export of orders, times and shifts; an interview with the planner; a first sequence calculated over a past month to compare against what actually happened.
The model proposes every morning; the planner decides. Energy cost per tonne, deliveries and replanning hours are measured against the previous month.
To the rest of the lines, with the same model and the same measurements.
Related: Predictive models Energy efficiency Method Glossary
Common questions
What do you need to plan with AI?
Manufacturing orders and their delivery dates (from the ERP or MES), real process and changeover times, the shift and maintenance schedule, and each piece of equipment's hourly consumption. That builds a model of the factory that knows what can and can't be done. The hourly electricity tariff is read from the public source.
Does it replace the planner?
No. The planner still decides. Every morning the model gives them a sequence that meets deliveries at the lowest energy and changeover cost, and explains why; when an urgent order comes in or a machine stops, it recalculates in minutes instead of an afternoon.
Does it work with our ERP?
Yes, as long as orders and times can be exported. We've worked with both flat-file exports and direct connections; the diagnosis says which one fits best.
When do you start seeing results?
The first pilot runs on one line or product family, over two to three months. That's where we measure how much the energy cost per tonne changed and how many replanning hours were saved compared with the previous month.
How does energy factor into the sequence?
Every hour has a different price and every piece of equipment a different consumption level. The model moves the highest-consuming operations to the cheapest hours within what deliveries and shifts allow, and calculates the sequence's energy cost before it's approved.
45-minute exploratory meeting
We pick a process, review what signals exist, and decide whether it's worth a diagnosis. No cost, no obligation.
Request the meetingSee the method