GWLAB - Multi-Objective Groundwater Management

Groundwater management refers to the problem of finding pumping schedules that minimize associated discounted present costs given resource and supply constraints. The costs have 2 components: facility installation costs and energy costs that result from pumping and conveying. Constraints exist in the form of drawdown restrictions, gradient criteria as well as quantitative supply constraints. The search space is highly dimensional since potentionally, pumps can be installed at any location to tap the groundwater system. For each of these pumps, the puming schedule is a time-series of pumping rates. Due to installation costs as well due to the complex nature of the resource, the problem is highly non-dimensional and standard requirements such as continuity and differentiability do not exist.

Various users from differing insititutional and economic background access groundwater to satisfy their demand each one pursuing his objective. Hence, optimal groundwater management is intrinsically a multi-objective task.

Here, we present a novel approach to the above problem. It has various advantages over previous approaches. First, the resource is not overly simplified but rather conceptualized as a finite-difference model. Second, installation, pumping as well as conveyance costa are taken into account. Third, an adaptive heuristics ensures constraint compliance and moves resp. switches on and off boreholes. Third, the optization problem is truely multi-objective and does not rely on the aggregation of the objective vector. With this, we can approximate the Pareto-optimal front of solutions that provides a base for further negotiations to be carried out between the users.

The variator part has been entirely coded in MATLAB. Generally, any aquifer geometry can be modelled. Different watershed partitioning schemes can be investigated. For example, Full cooperation in the form of permissive boundaries can be compared with territorial restricitve setups. Any optimization time horizon T can be modelled.

For further documentation, see paper.

• Author: Tobias Siegfried
• Documentation
• Paper
• License: PISA
• Download: (Matlab Package)