A python framework for multi-agent simulation of networked resource systems

• Published in: Environmental modelling & software : with environment data news Vol. 103; pp. 16 - 28
• Main Authors: Knox, Stephen, Meier, Philipp, Yoon, Jim, Harou, Julien J.
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 01.05.2018; Elsevier Science Ltd
• Subjects: Algorithms; case studies; Computer programs; Computer simulation; computer software; Electric power generation; Evolutionary algorithms; Hierarchies; Hydrologic models; Hydrology; Integration; Jordan; Modelling; Modelling framework; Modules; Multiagent systems; Multiple objective analysis; Open source; Open source software; Power plants; Python; Resource management; Rivers; Simulation; simulation models; Source code; Switzerland; Jordan; Switzerland; Modelling framework; Simulation; Open source; Python
• ISSN: 1364-8152; 1873-6726; 1873-6726
• DOI: 10.1016/j.envsoft.2018.01.019

Modelling managed resource systems can involve the integration of multiple software modules into a single codebase. These modules are often written by non-software specialists, using heterogeneous terminologies and modelling approaches. One approach to model integration is to use a central structure to which each external module connects. This common interface acts as an agreed mode of communication for all contributors. We propose the Python Network Simulation (Pynsim) Framework, an open-source library for building simulation models of networked systems. Pynsim's central structure is a network, but it also supports non-physical entities like organisational hierarchies. We present two case studies using Pynsim which demonstrate how its use can lead to flexible and maintainable simulation models. First is a multi-agent model simulating the hydrologic and human components of Jordan's water system. The second uses a multi-objective evolutionary algorithm to identify the best locations for new run-of-river power plants in Switzerland. •Python framework for building simulations of managed resource networks.•Open Source, component-based modelling framework.•Support for agent-based modelling.•Water resource system applications (Jordan, Switzerland) highlight Pynsim features.