swmm_wq_rl

This repository contains python codes and SWMM models to develop real-time control policies that seek to mitigate flooding and reduce pollution for the stormwater system in the Hague area of Norfolk, Virginia. Details and results of this research have been published in Environmental Science: Water Research and Technology available at https://doi.org/10.1039/D1EW00582K.

Required packages:

1. pyswmm
2. keras-rl (once installed, replace rl.core, rl.callbacks, and rl/agents.ddpg with modified files in the rl_wq/keras_rl_files folder)
3. openai-gym

Note that running RBC-DTN requires the pyswmm version in the rbc/pyswmm_hs.zip folder. This has a modified swmm5 dll.

Control Scenarios

1. Passive (uncontrolled) SWMM simulations use code in the passive folder.
2. Rule-based Control SWMM simulations use code in the rbc folder.
3. Reinforcement Learning SWMM simulations use code in the rl_wq folder.

References

A report on the RL agent training metrics is available as a pdf in the rl_wq folder.

This work builds on https://github.com/UVAdMIST/swmm_rl which has been published in the Journal of Hydroinformatics and is available via open access at https://iwaponline.com/jh/article/doi/10.2166/hydro.2020.080/77759/Flood-mitigation-in-coastal-urban-catchments-using.