Real-Time Optimization — RTO, DRTO & Economic MPC
Close the gap between feasible operation and economically optimal operation — in real time, every 30 minutes, continuously.
Book a Discovery CallReal-Time Optimization (RTO)
Real-Time Optimization (RTO) runs on a 30-minute to 4-hour cycle, solving an economic optimization problem — maximize throughput, maximize margin, minimize energy cost — using a rigorous steady-state process model. The result is an optimal target setpoint for the APC/MPC layer below it.
Without RTO, operators set targets manually, typically conservative and sub-optimal. With RTO, the plant continuously moves toward the economic optimum as prices, feed quality, and demand shift.
Problems RTO Solves
Conservative manual setpoints leave money on the table
Plants operated manually drift to setpoints that feel safe — typically 3–8% below the true optimal throughput or energy efficiency point. Every shift at a non-optimal setpoint is lost margin. RTO eliminates this conservatism systematically.
Optimal setpoints change every few hours
Feed composition, utility prices, ambient temperature, downstream demand — all change continuously. Operators cannot track these interactions manually. RTO recalculates optimal targets on every cycle.
APC holds setpoints but does not set them economically
MPC/APC is excellent at tracking a setpoint and handling disturbances. But the setpoint itself may not be economically optimal. RTO provides those economically optimal targets for APC to execute.
Step tests and steady-state models drift over time
A rigorous steady-state model initialized at commissioning drifts as equipment ages, catalysts deactivate, and heat exchangers foul. RTO includes model parameter estimation (data reconciliation) to keep models calibrated.
Classic RTO vs. DRTO vs. Economic MPC
| Aspect | Classic RTO | DRTO / Economic MPC |
|---|---|---|
| Model type | Steady-state (LP/NLP) | Dynamic model (DAE/ODE) |
| Cycle time | 30 min – 4 h | 5–30 min |
| Handles transients | Waits for steady state | Optimizes through transients |
| Economic objective | Explicit (separate layer) | Embedded in MPC horizon |
| Computational load | Moderate | High (NLP/QP at each step) |
| Best for | Continuous steady processes | Batch, grade changes, fast dynamics |
Industrial Applications
Refinery CDU / Crude Blend Optimization
Optimizes crude selection and cutpoint temperatures every shift to maximize distillate yield and minimize energy per barrel. Reference: $0.20–$1.00/barrel incremental margin improvement documented at large refineries.
Ethylene Cracker Severity Optimization
Cracker furnace severity (coil outlet temperature vs. feedstock) is the primary lever for ethylene yield vs. byproduct mix. RTO continuously finds the optimal severity given current feedstock prices and product demand — a calculation too complex for operators to track manually.
Ammonia Synthesis Loop
Ammonia synthesis loop pressure, purge rate, and recycle ratio have a multi-dimensional optimum that shifts with natural gas price, power cost, and ammonia demand. RTO updates the operating point continuously.
Building HVAC / District Energy
Pre-cool thermal mass during low-price electricity hours; dispatch storage to cover peak demand periods. ETH Zurich OptiControl-II documented ~17% reduction in non-renewable primary energy use on an occupied office building.
How We Deliver RTO
Rigorous Steady-State Model
First-principles model calibrated to plant data — heat exchanger duty, reactor conversion, distillation separation — the foundation every RTO layer requires.
Parameter Estimation & Reconciliation
Data reconciliation closes mass and energy balances; parameter estimation keeps the model accurate as equipment ages and feed changes.
Optimizer Integration
The optimization problem (LP or NLP) runs on a dedicated compute layer, reads live historian data, and writes optimal setpoints to the DCS/APC layer below.
APC as Setpoint Executor
RTO optimal targets flow down to MPC/APC controllers that track them precisely, with constraint handling and disturbance rejection. RTO and APC form a tightly coupled two-layer system.
Important: RTO Requires a Working APC Layer
RTO requires a functioning APC/MPC layer below it. Without APC, the plant cannot hold the optimal setpoints RTO computes — operators drift back to manual targets within minutes. If your plant does not yet have APC, the recommended sequence is: first deploy APC, then layer RTO on top.
Learn more: Economic MPC Optimization · Industrial Process NMPC / APC
Benefits
$1–15M/year per major process unit
Documented across refinery CDU, ethylene cracker, FCC, and ammonia plants. Payback typically 3–12 months for large units.
24/7 economically optimal operation
Humans cannot optimize a 30-variable process with shifting prices every 30 minutes. RTO does it automatically on every cycle.
No new hardware
RTO runs as software on a standard industrial PC or existing compute infrastructure, reading from the historian and writing setpoints to APC.
Captures transient value too (DRTO)
Dynamic RTO eliminates the "wait for steady state" penalty — optimizing through grade changes, feed transitions, and startups rather than pausing during them.
Relevant Design Patterns
Book a Discovery Call
If you want to close the gap between your current operation and the economic optimum — a 30-minute call is enough to assess whether RTO is feasible and what the payback looks like.
Book a Discovery Call →