1859
A geoenvironmental application of an optimisation model
Application d’un modèle d’optimisation à un problème geoenvironnemental
Azimi, K., Merrifield C., Gallagher E., Smith D.
Coffey Geotechnics, United Kingdom
ABSTRACT: A network of monitoring wells was installed in and around a refinery in mid 1990s as part of a research project aiming
to investigate the impact of local groundwater on corrosion of buried foundations and underground storage facilities. Oil contaminated
groundwater was evident in some of the monitoring wells. A second research project was started in 2000 to delineate the extent of the
oil contamination mound(s) beneath the refinery and devise appropriate remedial measures. Of 30 initial monitoring wells, 15 were
found operational inside the refinery in 2000. An optimisation technique is presented herein which assisted with augmentation of the
monitoring network, thereby the cost-effective delineation of the oil mounds beneath the refinery. The Maximal Covering Location
Problem (MCLP) was adapted and utilised to find the optimum number and locations of additional monitoring wells. The
contamination results obtained from the augmented and optimised network of monitoring wells were analysed using a geostatistical
tool and the oil contamination hot spots beneath the refinery were delineated cost-effectively.
RÉSUMÉ : Au milieu des années 1990, un réseau de puits de surveillance a été installé à l’intérieur et autour d’une raffinerie de
pétrole pour étudier l’action des eaux souterraines sur la corrosion des fondations enterrées et des structures de stockages souterrains
de la raffinerie. Une contamination par le pétrole a été détectée dans certains de ces puits. Un second projet de recherche a été lancé en
2000 pour suivre l’étendue de la contamination sous la raffinerie et concevoir des solutions appropriées pour y remédier. 15 des 30
puits fonctionnaient encore à l’intérieur de la raffinerie en 2000. Cet article présente, une technique d’optimisation du réseau de puits
de surveillance afin de cartographier l’évolution de la tache de pétrole sous la raffinerie pour un coût limité. Les auteurs ont modifié le
modèle d’optimisation ‘Maximal Covering Location Problem’ (MCLP) pour trouver le nombre optimal de puits de surveillance
supplémentaires et leurs emplacements. L’analyse de ces résultats en utilisant une méthode statistique a permis de confirmer le
contour de la contamination sous la raffinerie pour un coût bien défini.
KEYWORDS: Groundwater, contamination, monitoring, optimisation, MCLP, network augmentation
1 INTRODUCTION
An oil refinery constructed in the early 1970s and operated ever
since caused groundwater contamination. A research project
was conducted at the refinery in mid 1990s to investigate the
impact of local groundwater flow on corrosion of buried
foundations and underground storage facilities inside the
refinery. As part of that project, a network of 30 monitoring
wells was installed in and around the refinery. Oil
contamination of groundwater was evident in some of the
monitoring wells. A second research project was started in 2000
to delineate the extent of the oil contamination mound beneath
the refinery and devise appropriate remedial solution(s). Of 30
initial monitoring wells, 15 were found to be operational inside
the refinery at the beginning of the second research project.
Monitoring of these wells demonstrated that free phase of oil
contamination was present in groundwater at least at two
separate locations inside the refinery. However, the
contamination data obtained from the existing monitoring
network of 15 wells were too sparse for the purpose of oil
contamination delineation. Therefore, it was decided to add
monitoring wells to the network within the refinery. The two
major engineering challenges were identified as:
How many monitoring wells should be added to the existing
network?
In which locations should these wells be installed?
There is substantial evidence in the literature on the
application of Operations Research (OR)-based optimisation
methods in different civil and environmental engineering
practices (ReVelle et al. 1997). Fields of practice such as
transport engineering, urban planning and water resources
management are examples where successful applications of OR
methods including optimisation techniques have been
demonstrated. The Maximal Covering Location Problem
(MCLP) is an optimisation model proposed in the literature,
primarily devised to find the optimum locations for public
facilities, such as ambulance dispatch centres, on a network of
demand nodes (Church and ReVelle 1974). The model was
modified and applied to the groundwater contamination
problem in this study to assist with the cost-effective delineation
of the oil contamination mound beneath the refinery.
2 MAXIMAL COVERING LOCATION PROBLEM
(MCLP) – CONCEPT, THEORY AND APPLICATION
An example where the MCLP can be used for the optimum
usage of the resources is the requirement to add a certain
number of a public facility (e.g. ambulance dispatch centres) to
an existing network in a city. The city is discretized to a set of
demand nodes where additional dispatch centres can be situated.
Each demand node is assigned a weight representing its
population with the existing dispatch centres being attributed to
the nearest nodes. From the network operator’s point of view,
coverage of the demand nodes (i.e. population nodes in this
example) on the network is the key objective with demand
nodes not being located farther than a threshold distance (i.e.
maximal service distance,
S
) from an ambulance dispatch
