Inspiration
Refinery crude oil operations are the backbone of modern petrochemical supply chains, involving complex, multi-stage scheduling across tanker unloading, storage blending, and continuous Crude Distillation Unit (CDU) feeding. Traditional scheduling heavily relies on manual experience or rigid heuristics, which often struggle with combinatorial state spaces and sudden operational disturbances. We were inspired to build CrudeScheduling—an intelligent, automated optimization platform designed to deliver cost-effective, conflict-free, and adaptive scheduling solutions for real-world industrial operations.
What it does
CrudeScheduling streamlines the end-to-end refinery scheduling lifecycle:
- Multi-Stage Coordination: Seamlessly couples logistics from berth offloading and storage tank transfers to charging tank blending and distillation unit operations.
- Dynamic Optimization: Formulates operational constraints (inventory capacity, flow-rate limits, and quality-component tracking) into an adaptive optimization model: $$\min \quad J = C_{\text{demurrage}} + C_{\text{inventory}} + C_{\text{switch}}$$
- Automated Conflict Resolution: Automatically detects pipeline contentions, non-overlapping operation constraints, and tank state violations to generate feasible execution plans.
How we built it
- Algorithm & Optimization Core: Developed in Python using modular problem formulation, combining cooperative co-evolutionary strategies and agentic coordination to tackle large-scale discrete-continuous decision variables.
- Workflow & Rapid Prototyping: Leveraged Kiro developer tools to scaffold backend services, automate testbench generation, and streamline algorithm integration.
- Backend & Service Layer: Implemented lightweight API endpoints with FastAPI and structured data pipelines using NumPy and Pandas for rapid scenario evaluation.
Challenges we ran into
- Combinatorial Complexity: Balancing non-linear blending dynamics with binary execution states without suffering from exponential search-space explosion.
- Temporal Synchronization: Accurately modeling strict transfer delays, settling times, and non-preemptive feeding windows across interconnected pipeline networks.
Accomplishments that we're proud of
- Successfully delivered an end-to-end scheduling engine that generates validated, constraint-satisfying operating schedules within minutes.
- Significantly minimized vessel demurrage overhead while maintaining smooth, fluctuation-minimized feed quality for continuous distillation units.
What we learned
- Bridging classical process systems engineering (PSE) with modern agentic workflows significantly accelerates the development of complex optimization pipelines.
- Designing modular constraint validation rules is essential for scalable industrial optimization systems.
What's next for CrudeScheduling
- Integrating real-time digital twin telemetry to enable instant disturbance recovery and dynamic rescheduling.
- Expanding multi-unit support for concurrent multi-refinery petrochemical clusters.
Log in or sign up for Devpost to join the conversation.