A Dutch energy operator runs multiple biomass-fired power stations. Sourcing biomass (wood pellets, agricultural residue) from 200+ suppliers across Northern Europe involves a daily optimisation problem: which suppliers to purchase from, how much to hold in which storage facility, and how to route deliveries to minimise cost while maintaining feedstock continuity.
The existing approach was a spreadsheet run weekly by a small procurement team — manually balancing price, moisture content (which affects energy yield), transport distance and supplier reliability. The cost of sub-optimal decisions was running to millions of euros annually.
The challenge: the problem is genuinely hard combinatorially. 200+ supply nodes, 5 storage depots, 3 power stations, daily price changes and minimum stock requirements at each plant. No off-the-shelf tool could handle the full constraint set.
Worked with procurement and engineering to enumerate all constraints: minimum plant stock, maximum supplier allocations, transport capacity, moisture thresholds. Formulated as a multi-period MIP.
Built and tuned a Gurobi-based MIP solver that runs a 7-day rolling optimisation daily. Warm-starts from prior solution to keep solve times under 3 minutes on standard hardware.
Discrete-event simulator that stress-tests optimised plans against realistic uncertainty — supplier delays, price spikes, weather events. Gives procurement team a risk distribution, not just a single plan.
Daily optimisation runs automatically at 06:00. Results surfaced in a procurement dashboard with one-click PO generation. Deployed on-premise at client's data centre per their security requirements.