Unit commitment is the daily decision that determines which generation units run, at what output, at what hour — and it is the single largest source of avoidable cost in a portfolio of thermal, hydro, and renewable generators. Independent System Operators (MISO manages 146,000 MW across 13 states and Manitoba alone) solve unit commitment as a mixed-integer linear programming problem with hundreds of binary variables, thousands of ramp constraints, and security-constrained transmission limits. Portfolio operators solve the same problem at smaller scale but with the same math. A working UC discipline is not an ISO market feature — it is the tool that turns fuel cost, ramping capability, and reserve requirements into a defensible dispatch decision every day.
iFactory / Unit commitment for generation portfolios
Which Units Run, When, and Why — Defensibly, Every Day
A unit commitment framework built on MILP with security-constrained SCUC and SCED, ramp constraints, minimum up/down times, and start-up cost recovery — applied to your portfolio of thermal, hydro, and renewable units.
Unit A · CCGT
ON
ON
ON
ON
ON
ON
Unit B · Coal
off
ON
ON
ON
ON
off
Unit C · Peaker
off
off
ON
ON
off
off
MILP
mixed-integer linear program
SCUC + SCED
security-constrained UC + dispatch
The Problem in the Control Room
Unit commitment has two coupled parts: scheduling start-up, operation, and shutdown of each available generation unit over the horizon, and allocating total power demand across the running units to minimize overall cost. The first uses binary variables — is Unit B committed at hour 4? — while the second is the economic dispatch problem determining continuous output per unit. Together they form a MILP with objective minimizing total generation cost subject to load balance, spinning reserve, transmission thermal limits, ramp rates, minimum up/down times, and start-up costs. Solving it well saves money on every dispatch day; solving it poorly leaves cost on the table and creates ramping scarcity in real-time operation.
Where the Record Actually Breaks
Unit commitment done manually or in spreadsheets breaks in specific ways. Each of these gaps costs money on every dispatch day.
Spreadsheet dispatch
Hourly commitment set by a planner using rules of thumb and yesterday's load curve. Correct on average, suboptimal on any given day. No systematic reserve or ramping check.
Ramping constraints ignored
Units committed without full ramp-rate modeling. Real-time operation hits ramping scarcity when load moves faster than the committed fleet can follow. Emergency starts follow.
Start-up costs mis-priced
Cold-start, warm-start, and hot-start costs treated as a single number. The economics of decommitting a unit for four hours versus keeping it at minimum load get miscalculated.
Security constraints late
Transmission thermal limits and contingency constraints applied after commitment as an override, not as an integrated constraint. The resulting dispatch is feasible but not optimal.
What Good Looks Like at the Desk
A working UC framework holds four disciplines together — proper MILP formulation, security-constrained integration, ramping capability modeling, and rolling horizon with intra-day re-optimization.
MILP Formulation
Binary commitment variables and continuous dispatch variables solved jointly. Objective minimizes total cost including fuel, start-up (cold/warm/hot), no-load, and reserve procurement.
Standard MILP
SCUC + SCED
Security-constrained UC applies transmission thermal limits and contingency constraints inside the optimization, not after. SCED runs economic dispatch on the committed fleet with the same security envelope.
Security in the model
Ramping Capability
Ramp rates, minimum up/down times, and increasingly flexible ramping products modeled per unit — capturing the real cost of load-following vs baseload operation.
Ramp as first-class
Rolling Horizon
Day-ahead UC re-optimized intra-day as load forecast, renewable output, and unit availability update — the commitment plan reflects reality, not yesterday's forecast.
Update, don't hold
How iFactory AI Fits
iFactory AI provides the UC engine and integrates with the SCADA/EMS, meter data, forecast, and fuel-cost sources your portfolio already uses — supporting both ISO-market participants and vertically integrated utility operations.
MILP Solver
UC Engine
Commercial-grade MILP solver (Gurobi, CPLEX, or open-source) configured for your unit count and horizon length. Solve times under an hour for typical portfolio sizes.
Unit Model Library
UC Engine
Per-unit parameters for capacity, heat rate curve, ramp rates, min up/down times, start-up costs by state, and fuel type — the input that makes the MILP solve match reality.
Security Constraints
UC + Grid Model
Transmission thermal limits, contingency scenarios, and reserve zones integrated into the SCUC formulation — commitment plans that are simultaneously optimal and secure.
Rolling Re-Solve
UC Engine
Intra-day re-optimization on a defined schedule (hourly, 15-minute, or on-trigger) with warm-start from the current commitment — updated plans in minutes, not hours.
If your generation portfolio is currently dispatched from a spreadsheet or a legacy tool from 2010, the annual cost of leaving MILP optimization on the table is measured in millions on a fleet above 1 GW. Book a UC modeling session — we'll benchmark your current dispatch against an SCUC-optimized plan.
12-Week Rollout on One Unit
One portfolio segment (thermal fleet, or thermal + one hydro asset), one operational horizon, twelve weeks. The pilot proves the MILP solve completes reliably and the commitment cost improves versus the current dispatch method.
Weeks 1–2
Unit Model Build
Build the per-unit parameter set for the pilot portfolio — heat rate, ramp, min up/down, start-up cost by state, fuel cost. Validate against operator experience.
Weeks 3–4
Formulation & Solver
Configure the MILP formulation with objective, constraints, and security envelope. Test-solve on historic days. Compare committed plans against actual dispatch.
Weeks 5–8
Parallel Dispatch
Run SCUC in parallel with your current dispatch for 4-6 weeks. Track committed plan vs actual dispatch cost. Refine unit parameters based on discrepancies.
Weeks 9–12
Cost Improvement
Twelve-week window closes. Compare total generation cost, start-up cost, and reserve procurement against the baseline. Rollout decision based on measured savings.
Who Owns the KPI
Unit commitment crosses generation planning, real-time operations, trading, and finance. Each function needs a specific number they own or the framework becomes a planning tool nobody dispatches from.
Generation Planner
Day-ahead commitment cost $/MWh
Owns the day-ahead plan quality — the cost per MWh committed for the next day versus the actual delivered cost when the day runs.
Real-Time Operator
Ramping scarcity events per week
Owns the real-time follow-through — the count of hours where the committed fleet couldn't follow load without emergency starts or reserve draw.
Trader / Market Ops
Day-ahead vs real-time price capture
Owns the market-participation outcome — the delta between day-ahead prices offered against and real-time outcomes achieved, which UC quality directly drives.
Fuel / Finance
Fuel cost per MWh vs plan
Owns the P&L consequence — the total fuel and start-up cost per MWh generated relative to the UC plan. This is where all the optimization work shows up on the ledger.
FAQ
Is MILP overkill for a small generation portfolio?
It depends where 'small' cuts off. For a single-unit operator or a portfolio under 100 MW with simple duty cycle, a rules-based dispatch may be adequate. Above 500 MW with mixed thermal, hydro, or renewables — where start-up decisions have four-figure economics per event and ramping constraints interact with reserve requirements — MILP typically pays back on the first quarter of use. The break-even isn't fleet size alone; it's the value of a marginal 1% cost improvement on your energy volume, which for most operators above 500 MW is measured in millions per year.
How does UC handle renewable variability and forecast uncertainty?
Two approaches, often combined. Deterministic UC treats the point forecast as the load and generation input and re-solves the UC as forecasts update intra-day — the rolling-horizon approach. Stochastic or robust UC includes multiple scenarios of renewable output in the formulation, producing a commitment plan that is feasible across a range of realized outcomes at some cost premium. The right choice depends on your renewable penetration, forecast accuracy, and appetite for real-time re-optimization. Increasingly, ISOs (CAISO, MISO, SPP) also procure flexible ramping products explicitly to manage the residual variability — an active area of formulation research.
Book a demo to see UC with your renewable forecast integrated.
What about hydro units with reservoir constraints — does UC handle those?
Yes, but with additional constraints. Hydro units add water-balance constraints (reservoir volume, inflow, outflow), often with intertemporal coupling across the horizon (water used today is water unavailable tomorrow) and cascade constraints if multiple plants share a river system. These are integrated into the MILP as additional linear constraints. Pumped-storage hydro adds an additional set of constraints for pump vs generate modes. Portfolios with significant hydro benefit from longer commitment horizons (a week or more) to properly value the intertemporal water opportunity cost.
Stop dispatching from a spreadsheet built in 2010.
Benchmark One Week of Your Dispatch Against SCUC-Optimized Commitment
Bring one week of your generation portfolio's actual dispatch, load, and fuel costs. We'll re-solve the same week with a full SCUC formulation, quantify the cost delta, and identify where ramping or start-up cost is being left on the table.
