The Challenge
A Gulf Coast petrochemical refinery operates a continuous reactor train producing ethylene derivatives. The process consists of three continuous stirred-tank reactors (CSTRs) in series, followed by separation and purification units.
The 24/7 operation processes over 500,000 metric tons of feedstock annually. With razor-thin margins in commodity chemicals, even small efficiency improvements translate to significant financial impact.
Key Challenges
- Feedstock variability: Raw material composition varies by supplier and season, requiring constant adjustment of operating conditions
- Energy-intensive heating: The reactors require significant steam input, representing 40% of total operating costs
- Coupled dynamics: Changes in one reactor affect downstream units, making optimization complex
- Environmental pressure: Growing regulatory and corporate focus on reducing carbon emissions
"Our operators are skilled, but optimizing three coupled reactors in real-time while handling feed changes is humanly impossible. We needed a system that could see the whole picture."
— Process Engineering ManagerThe Solution
The refinery deployed Acaysia across their CSTR train, implementing coordinated multi-unit optimization while maintaining individual reactor safeguards.
System Architecture
Unlike batch reactor applications, continuous processes require real-time coordination across multiple units. Acaysia's architecture addressed this by:
- Plant-wide model: A unified model capturing interactions between all three reactors and downstream units
- Hierarchical control: Plant-level optimization sets targets; unit-level controllers execute within constraints
- Real-time feed analysis: Integration with online analyzers for rapid response to feed composition changes
Implementation Timeline
| Phase | Duration | Activities |
|---|---|---|
| Data Collection | 6 weeks | Historical data analysis, model development |
| Shadow Mode | 4 weeks | Model validation across operating envelope |
| Advisory Mode | 3 weeks | Operator training, recommendation refinement |
| Closed-Loop | Ongoing | Full automated optimization |
Technical Approach
Feed-Forward Optimization
Traditional control reacts after disturbances propagate through the system. Acaysia uses online feed analyzers to predict impacts before they occur:
- Near-infrared spectroscopy for real-time feed composition
- Predictive models correlate feed properties to optimal conditions
- Proactive temperature and flow adjustments minimize disturbance impact
Multi-Unit Coordination
The three CSTRs have significant thermal and mass interactions. Optimizing one reactor in isolation often shifts problems to another. Acaysia's plant-wide approach considers:
- Conversion distribution across the reactor train
- Heat integration opportunities between units
- Downstream separation costs in overall optimization
Dynamic Constraint Management
Equipment constraints vary with ambient conditions and unit health. Acaysia continuously adapts:
- Heat exchanger fouling factors updated daily
- Cooling tower capacity adjusted for ambient temperature
- Compressor curves incorporated for accurate constraint handling
The Results
Energy Reduction
Total energy consumption decreased by 4.2%, primarily from:
- Optimized temperature profiles reducing steam demand (3.1%)
- Improved heat integration between reactors (0.7%)
- Reduced cooling water consumption (0.4%)
At $6/MMBTU for natural gas and continuous operation, this translates to approximately $2.1M in annual energy savings.
Throughput Improvement
By operating closer to true constraints rather than conservative estimates, effective capacity increased by 2.3%. This additional production contributed approximately $1.0M in incremental margin.
Environmental Impact
The energy reduction corresponds to approximately 15,000 metric tons of CO₂ emissions avoided annually—equivalent to taking 3,200 cars off the road. This helps meet corporate sustainability commitments and positions the facility favorably for future carbon regulations.
Operational Stability
Beyond direct savings, operations became more stable:
- Unplanned transitions reduced by 35%
- Product quality variability decreased by 28%
- Operator interventions for grade changes cut in half
Performance Summary
| Metric | Before | After | Change |
|---|---|---|---|
| Steam Consumption | baseline | -3.8% | -3.8% |
| Cooling Water | baseline | -5.1% | -5.1% |
| Throughput | baseline | +2.3% | +2.3% |
| Quality Variability | ±2.1% | ±1.5% | -28% |
| CO₂ Emissions | baseline | -15K tons/yr | -4.2% |
Safety Considerations
Petrochemical operations demand the highest safety standards. Key safeguards in this deployment:
- Independent SIS: Acaysia operates entirely separately from the Safety Instrumented System, which maintains full authority
- Conservative constraints: Operating limits set 5% inside actual equipment ratings
- Gradual changes: Rate limits on all control moves prevent rapid disturbances
- Automatic fallback: Any anomaly triggers immediate return to base PID control
"We had extensive discussions with our process safety team before deployment. The defense-in-depth architecture and automatic fallback gave us confidence to proceed. After a year of operation, we haven't had a single safety incident related to the system."
— HSE DirectorLessons Learned
- Model quality matters: Investing extra time in model development paid dividends in optimization performance
- Operator buy-in is essential: Early involvement of operators in system design led to faster adoption
- Start with advisory mode: The 3-week advisory period built trust and identified edge cases before automation
- Continuous improvement: Monthly reviews with operations identified additional optimization opportunities