Show how steam balances, energy prices and capacity constraints can be translated into an optimization problem and compared with a reference dispatch policy.
3 · What it produces
Reference/optimized cost, steam dispatch, direct-combustion ΔCO₂, LP shadow prices and perturbation-based MILP marginal values.
The images come from demonstration runs included in the attached repository and are shown without dates to illustrate workflow structure, method and evidence.
4 · How it works
From data to output, with explicit controls.
1Synthetic HP/MP/LP demand
2Gas, power and carbon prices
3Utility balances and constraints
4HiGHS LP optimization
5MILP with min-load/start-up
6Perturb/re-opt marginal values
Transferable pattern: Utility balances + LP/MILP + frozen reference policy.
5 · Data, AI/ML/RPA and method
Data and features
synthetic steam demand, capacities/efficiencies, EEX TTF, licence-checked Energy-Charts Italian bidding-zone day-ahead power, EUA
Calculation / inference
HiGHS LP and MILP; dispatch and perturb/re-opt
Validation
feasibility, balances, optimality, solve time, stability and reference policy
Stack
Python, SciPy HiGHS, NumPy, pandas, SQLite, Matplotlib
6 · Technical dossier
Operating conditions and declared limitations.
| Compute cadence | Daily 11:50 Europe/Rome |
|---|---|
| Planned publication | IT Sat 13:30; EN Sat 19:30 |
| Possible applications | energy management, steam balancing and capacity debottlenecking |
| Research use | LP duality, unit commitment and sensitivity |
| Required condition | synthetic utility demand; modelled opportunity, not measured plant savings; electrical balance not modelled |
| Failure mode | calling MILP marginal values duals or presenting simulated savings as measured savings |
Interpretation boundary
Utility demand is synthetic and the plant electrical balance is not modelled. Savings are modelled opportunities, not measured savings; MILP marginal values are not duals.
7 · Verification and further reading