Awesome

Status: Expect regular updates and bug fixes.

Gym environment for PV-DER

Solar photovoltaic distributed energy resources (PV-DER) are power electronic inverter based generation (IBG) connected to the electric power distribution system (eg. roof top solar PV systems). This environment consists of a single DER connected to a stiff voltage source as shown in the following schematic:

schematic of PV-DER

Basics

The dynamics of the DER are modelled using dynamic phasors. One step in the environment is equivalent to n simulation time steps and each simulation time step is one half-cycle (1/120 s).

Events in the environment

There are two types of perturbations in the environment: a) Change in grid voltage magnitude b) Change in solar insolation.

Installation

First install the Solar PVDER simulation utility. Then you can install the PVDER environment using following commands:

git clone https://github.com/sibyjackgrove/gym-SolarPVDER-environment.git
cd gym-SolarPVDER-environment
pip install -e .

Other dependencies: OpenAI Gym, Numpy, SciPy

Using the environment

The environment can be instantiated just like any other OpenAI Gym environment as show below:

import gym
import gym_PVDER
env = gym.make('PVDER-v0')

Available user options

DISCRETE_REWARD: If this is set to True discrete rewards (i.e. +1, -1, -5 etc) will be returned every time step. Otherwise reward will be the error between goals and actual values.

Try it out in Google Colab:

Create environment and make a random agent interact with it:

The environment can also be used with TensorFlow agents. Try out the demo below which trains a DQN agent using the environment:

Citation

If you use this code please cite it as:

@misc{gym-PVDER,
  title = {{gym-SolarPVDER-environment}: A environment for solar photovoltaic distributed energy resources},
  author = "{Siby Jose Plathottam}",
  howpublished = {\url{https://github.com/sibyjackgrove/gym-SolarPVDER-environment}},
  url = "https://github.com/sibyjackgrove/gym-SolarPVDER-environment",
  year = 2019,
  note = "[Online; accessed 18-March-2019]"
}