Author(s): Andrew R. Slaughter; James Tomlinson; Julien J. Harou
Keywords: Decision-making under uncertainty; Water resource simulation; Network optimisation; Waterstrategy. org; National water infrastructure planning
Abstract: Pywr is an open-source, Python-based generic dynamic modelling library for network-based resource allocation models. Pywr can flexibly simulate water resource systems at varying spatial and temporal scales by allocating available water resources using a linear program, user-defined penalties, and a mass-balance approach. Models are represented as nodelink systems in which the links are directional and the nodes contain all data required for resource allocation. Python scripting capabilities mean custom and complex operational rules can be represented, including reservoir release rules. A major advantage of Pywr is its computational efficiency, which allows linking Pywr simulation models with multiobjective evolutionary algorithms for optimisation. This functionality is available by default in the open-source software library. We present two case studies of using Pywr to represent national- and regional-scale water supply systems. First, a model for optimised design of cost-effective water transfers and supply systems for England and Wales is presented. Using this model, we briefly demonstrate the advantages of Pywr in identifying a range of cost-effective portfolios of national-scale infrastructure options which can be aligned to planner preferences. Second, a regional simulation model for Water Resources East (WRE) in England is presented. This case study demonstrates the use of the regional model to estimate safe yields of various proposed new planned reservoirs using 400 stochastic future climate change inflows. Finally, we introduce a web-based Pywr interface available at www. waterstrategy. org, which researchers and practitioners can use to rapidly create and share large-scale water models.