Optimizing Long Term Monitoring at a BP Site Using Multi-Objective Optimization Barbara Minsker (University of Illinois and RiverGlass Inc.), Peter Groves (RiverGlassInc.), and Dennis Beckmann (BP)AbstractBP (formerly British Petroleum) incurs significant costs associated with monitoring subsurface remediation sites. The purpose of this project is to evaluate whether these costs could be reduced by identifying and eliminating both spatial and temporal redundancies in the monitoring data at a BP site without significantly increasing monitoring errors. The project also aims to demonstrate the potential for multi-objective optimization approaches to improve monitoring decision making at the many sites at BP and elsewhere with long-term monitoring records.The first step in the optimization process is to identify monitoring objectives and constraints, and express them in mathematical form. In this case, the initial objectives were to minimize the number of samples collected and to minimize relative BTEX interpolation error. The BTEX interpolation error for trial sets of sampling plans are calculated by comparing the concentrations interpolated using all sampling locations and times with those interpolated using only reduced sampling frequencies or locations. Historical data from the wells that are currently being sampled are used to develop a suite of interpolation models, which are then tested using a cross-validation approach. Adaptive Environmental Monitoring System (AEMS) software, developed at the University of Illinois and RiverGlass Inc., is then used to search through the billions of sampling plans to identify the optimal tradeoffs between the number of samples collected and the relative error.IntroductionRoutine groundwater monitoring can account for a significant portion of the lifecycle spending for many remediation projects. Some estimates put long-term monitoring costs at 30% of an overall environmental restoration project budget [EWRI 2003]. Numerous monitoring optimization approaches exist (see EWRI 2003 for an overview), but most recent work has focused on reducing spatial and temporal redundancy in monitoring plans. Spatial redundancy focuses on eliminating wells whose sampling data can be estimated from surrounding wells without introducing significant errors. Temporal redundancy focuses on reducing sampling frequencies to eliminate redundant data at the same well over time. This paper proposes a new approach to optimizing for both spatial and temporal redundancy simultaneously using evolutionary multi-objective optimization. The approach is demonstrated at a BP site in New Jersey.Site DescriptionThe BP site is a 100-acre terminal built in the late 1920s with both dense and light nonaqueous phase liquid (DNAPL and LNAPL) from chlorinated solvents and BTEX. The geology of the site is complex, with a multi-layered sand and gravel aquifer and a
本文来源：https://www.2haoxitong.net/k/doc/51a7177e27284b73f2425094.html
