Production-Distribution Coordination for Optimal Operational Planning of an Industrial Gases supply-chain

Transcription

1 Production-Distribution Coordination for Optimal Operational Planning of an Industrial Gases supply-chain Vijay Gupta, Ignacio E. Grossmann Department of Chemical Engineering Carnegie Mellon University Pittsburgh PA Sujata Pathak, Jean André DRTC, American Air Liquide Inc. Newark, DE Enterprise-Wide Optimization (EWO) Meeting, Oct. 2011

2 Industrial Gases Supply-Chain C2 C1 C5 Plant 1 Plant 2 C4 C6 C3 Alternate Source Multiple Plants Multiple Products (LIN Liquid Nitrogen, LOX Liquid Oxygen etc.) Over-the-fence, call-in and distributed customers (some shared customers) Storage facilities at production sites and customer locations Delivery Modes and Routes 2

3 Sequential vs. Simultaneous Approach: Key Differences Sequential Strategy Simultaneous Strategy Production Model (based on given truck withdrawal or demand forecast) Optimum Production Schedule Integrated Model (truck withdrawal amounts and time are variables) Distribution Model (based on product availability) Optimum Distribution Schedule Goal Total Cost? Total Cost Optimum Production and Distribution Schedule To quantify and access the savings associated with the Production-Distribution Coordination at Operational Level using an approximate model 3

4 Production-Distribution Coordination: Multi-plant Level Multi-plant Sequential (Multiple Sourcing) Currently used Production Schedule Distribution Schedule Coordination among plants for production and for Distribution Individually Multi-plant Simultaneous (Multiple Sourcing) Production and Distribution Schedules Coordination among plants as well as Production-Distribution (Fully Integrated) 4

5 Production-Distribution Coordination Levels Sequential Based on Truck Withdrawal Forecast Based on Demand Forecast Simultaneous Single Plant (Single Sourcing) No Coordination b/w plants and production-distribution (production depends on truck forecast) No Coordination b/w plants and production-distribution (production depends on demand forecast) Coordination b/w production-distribution but No coordination b/w plants (partially integrated) Multi-plant (Multiple Sourcing) Coordination b/w plants but No Coordination b/w productiondistribution (production depends on truck forecast) Coordination b/w plants but No Coordination b/w productiondistribution (production depends on demand forecast) Coordination b/w production-distribution as well as plants (fully integrated) Truck withdrawal is a given parameter Freeze the product demand instead truck withdrawals Truck withdrawal is a variable in the optimization model 5

6 Toy Problem Statement Given 3 Plants, 2 main products (LIN, LOX) 3 modes for each plant and respective capacities (hi LOX, hi LIN, shut-down) 17 Customers (truck delivery), 42 Pick-up customers, 2 alternate sources 14 time periods (peak and off-peak), 20 trucks (10 for LIN, 10 for LOX) Demand, min/max inventory, distance, Electricity price etc. data Decisions in each time period t Production rates at each plant Inventory levels at customer locations and plants How much product to be delivered to each customer through which route Objective Function Minimize total production and distribution cost over planning horizon (1 week) 6

7 Results: Multi-plant Simultaneous vs. Sequential Models Comparison: Shared Customers Deliveries C1 C3 C10 C11 C6 P: Plant P1 C: Customer A: Alternate C12 C5 C9 LIN Customer LOX Customer Shared Customers C4 C7 C16 C8 C13 C2 C15 A2 P2 C14 P3 C17 A1 7

8 Results: Toy problem (Production-Distribution Coordination Levels) Single Plant Sequential (Truck Forecast) Single Plant Sequential (Demand Forecast) Single Plant Simultaneous Multiple Plant Sequential (Truck Forecast) Multi plant Sequential (Demand Forecast) Multi plant Simultaneous Production Cost Distribution Cost Total Cost 8

9 Conclusions Simultaneous and Sequential MILP Models are proposed for optimal operational planning of industrial gases. Multiproduct Multi-plant Variety of customers Routing decisions Numerical results on test case show significant potential savings (~10%) and different production/distribution schedules with the coordination due to switching sourcing/routing strategies, electricity price differences, better inventory management... Simultaneous Model will further be extended to include other details on production and distribution sides to be more realistic. This will allow to further confirm the promising results obtained till now. 9