An Agent-based Simulation-optimization Approach for Device Allocation in Offshore Oil Spill Recovery Pu Li, Zelin Li, Bing Chen*, and Liang Jing Northern Region Persistent Organic Pollution Control (NRPOP) Laboratory, Faculty of Engineering and Applied Science, Memorial University of Newfoundland, St. John’s, NL, Canada, A1B 3X5; *Email:
[email protected]
Initial conditions
Device conditions
Interaction of recovery devices
No
Optimal settings for stage n
To ensure a sound design and optimal operation in terms of time with expected recovery rate
2. Methodology
Objectives:
Perception Agent 1 (Behavioral specification 1)
Perception Interaction Agent 2 (Behavioral specification 2) Actions
...
Interaction Interaction Agent k (Behavioral specification k)
Environment
Perception
Simulation-based dynamic mixed integer nonlinear programming (DMINP): simulationoptimization coupling for decision support to offshore oil spill recovery
ASO approach: further considers the interaction of the recovery devices (reflected by ABM) with the DMINP to provide sound decisions
Optimize the routes of ships to achieve 90% of oil recovery with minimum time
1 2 3 4 5 6 7
Location X (km) Y (km) 24.03 5.80 19.97 14.07 27.49 16.61
10.03 18.46 20.99 3.43 5.42 29.39
Oil volume (m3) 132.44 219.37 146.69 137.82 81.07 79.86
ORRn of Ship A 40
ORRn of Ship C 30
20
10
0
3.27
13.84
202.76
10
20
30
Slick thickness (mm)
Optimized routes based on the ASO approach optimal time: 21 hours
40
50
0
Transport to Slick 7
2.9
Oil recovery on Slick 7
Oil recovery on Slick 4 6 6.3
Transport from Slicks 7 to 2
6 6.6
Oil recovery on Slick 2 9 9.8
15
Oil recovery on Slick 3 12 12.5
Transport from Slicks 3 to 6
10 Oil recovery on Slick 6
14 14.3
Transport from Slicks 4 to 1 Oil recovery on Slick 1
Transport from Slicks 5 to 3
9 9.3
Transport from Slicks 1 to 5
11 11.3 12 12.6
Oil recovery on Slick 5 Transport from Slicks 5 to 1 Oil recovery on Slick 1 Transport from Slicks 1 to 3
Transport from Slicks 2 to 7 Oil recovery on Slick 3 Oil recovery on Slick 7
17 17.8
3
5
7
9
11
13
15
19
17
19
21
17 17.5
Transport from Slicks 6 to 2 Oil recovery on Slick 2 Stop operation
21
Transport from Slicks 7 to 2 Oil recovery on Slick 2 Stop operation
19 20.1 21
Transport from Slicks 6 to 7 Oil recovery on Slick 7 Stop operation
1000
Ship C
250
90% of oil recovery
900
Ship B
200
150
100
800 700 600
Comparison of oil recovery by routes based on shortest distances and ASO optimization
500 400 300 Routes based on ASO optimization
200
50
3
5
7
9
11
13
15
17
19
Routes based on shortest distance without consideration of ship interactions
100
Cumulative oil recovery by ships 0 1
Transport from Slicks 3 to 6 Oil recovery on Slick 6
19 19.3
21
300
50
0
Oil recovery on Slick 1 Transport from Slicks 1 to 5
Ship A
ORRn of Ship B
Slick
5 5.3
350
Three ships (Ship A, B, C) with three different types of skimmers different oil recovery rate (ORRn) with slick thickness as follows:
ORRn (m3/hr)
Agent-based model (ABM): a class of computational approaches for micro-scale simulation of the actions and interactions of autonomous agents (e.g., competition of recovery devices in offshore oil spill events) with a view to assessing their effects on the system as a whole
An offshore oil spill with a release of 1,000 3 m crude oil initial thickness: 50 mm Due to advection and spreading, the spilled oil was separated to 7 slicks with different volumes as shown in the following table. Assume that no further weathering process occurs during the recovery operation
Transport from Slicks 4 to 1
Action
Transport to Slick 4
Time (hour)
3. A case Study
Time (hr)
Action
0 0.2
Transport to Slick 4
Oil recovery on Slick 5
0 1
Update the operational stage n=n+1
Time (hr)
Action
3 3.6
20
5
OBJECTIVES: To develop a novel agent-based simulationoptimization (ASO) approach and first apply to offshore oil spill recovery
Ship C
25
Ship C
Ship B
Oil recovery on Slick 4
Hourly oil recovery by ships
Goal; Rule; Sequence of activities
response device
0 0.2
Ship B
Predefined Plans Agent
Time (hr)
Ship A
30
Is the pre-set goal achieved
Simulation-based dynamic mixed integer nonlinear programming (DMINP)
Oil recovery simulation New environmental conditions
Efficiency; Time; Cost
Cumulative oil recovery (m )
Environmental conditions
Man power; Finance; Regulation
Ship A
35
Oil recovery by ship (m3)
Some trajectory models (e.g., GNOME and OSCAR) have been developed for offshore oil spill response some are integrated with oil spill simulation such as oil weathering and recovery none was found in further consideration of the interactions of response devices
Yes
3
Offshore oil spills: a release of a liquid petroleum hydrocarbon into the ocean or coastal waters due to human activities leads to serious impacts requires quick response
4. Discussion
Operation ends
Policies (Targets)
Cumulative oil recovery (m3)
1. Introduction
21
0 1
3
5
7
9
Time (hour)
11
13
15
17
19
21
23
Time (hour)
5. Conclusions The ABM can reflect the interactions of components in a offshore oil spill recovery system and integrate with the optimization The DMINP can integrate the simulation processes with the optimization of offshore oil spill recovery actions The ASO approach can provide sound decisions for oil recovery under highly interactive conditions and improve recovery efficiencies Weathering processes and hydrodynamic simulation will be considered in future study
Acknowledgement Sincere thanks go to Natural Sciences and Engineering Research Council of Canada (NSERC), Research & Development Corporation (RDC) of Newfoundland and Labrador, and Canada Foundation for Innovation (CFI) for funding support of our research, as well as Eastern Canada Response Corporation (ECRC) and Canadian Coast Guard (CCG) for technical advices in the case study.