Jiadalee / storm-swarm

Modular API for large scale RTC SWMM5 simulations.

⛈️ ⛈️ Storm Swarm ⛈️ ⛈️

Modular API for running large ensembles of real time control simulations in EPA-SWMM

Dependencies

This codebase heavily draws from the OWA's pyswmm and SWMM. Though this code can be used along pyswmm, I've provided the necessary components from pyswmm with this library, so that it can be used as a standalone package.

Additional Dependencies

1. Numpy
2. Multiprocessing

Installation

    git clone  https://github.com/kLabUM/storm-swarm.git
    pip/pip3 install .

Getting Started

This library is 3 primary components:

1. Config (dict)
2. Environment (class)
3. Swarm (function)

Config

This dictionary configures the storm water network being stimulated and identifies the state and action space.

        config = { "swmm_input" : "./networks/parallel.inp",
			       "state_elements"  : {"state_order" : ["P1", "P2"],
								        "P1" : "depth",
								        "P2" : "depth",
								        "state_len" : 2},
			       "control_elements" : ["1"],
                   "controller" : ctrl,
                   "reward_function" : reward,
                   "timesteps" : 1}

Environment

This class acts as a handle to control SWMM simulation and can also be used independently

    env = environment(config)

Swarm

This function can take single/mutiple config files and runs them in parallel.

    generate_swarm(config, workers=3, jobs=10)