First usage example: play the “grid2op game”

Once you started a grid2game server with a given grid2op environment, you can easily start operating the grid on a given scenario (a scenario is analogous to a “level” in a video game).

Choose a scenario

To choose a scenarion, you simply need to click on the dropdown list:

As some environments are stochastic, you might also want to chose a seed for this scenario ( if you use the same seed on the same scenario, all the “stochastic” components will be the same. For example, maintenance operations will happen exactly at the same time, concern the same powerline, the opponent will attack at the same time for the same duration etc.)

This can also be achieved in the GUI by entering a number in the appropriate field:

Once done, you will need to “reset” the environment by pressing the “reset” button:

And you can start operating the grid !

To make sure the scenario is properly configured, you can have a look at the information just above the grah of the powergrid:

(in this example, the scenario named “0002” has been selected and no seed has been set)

Play interactively

Once the scenario is loaded, you can interact with the grid just like any “agent”. Here are the main commands similar to “env.step(action)”

Note

We will explain in the next section (Choose an action) how to chose the action

All the interesting buttons are located there:

In this “quickstart” we only detail a few of them. More information can be found in the dedicated page: Step, back, +XXX and “end” .

Step

This button allows you to perform a single “step” on the environment. It is rather straightforward.

Simulate

Compatible with the grid2op framework, we added the possibility to use a model of the grid (a simulator) to assess if a given action is good or not. This is done with obs.simulate(action) (more info on the official grid2op documentation).

This “simulate” button allows you to do that. The resulting grid state (of the simulation) is displayed on the right.

+12

A scenario can last more than 1000 steps (usually even more than 8000). In this case it would be very impractical to clik on the “step” button more than 1000 times in order to get to the end of the scenario.

For this purpose, the “+12” button allows you to perform more than one step when you click on it.

Note

You can customize how many steps you want to perform per click by entering a digit in the “steps” field.

It has the following behaviour:

• when the assistant raises an alarm, the computation is stopped, and you see the grid state just after the assistant raised the alarm.

• when there is a game over, the computation is also stopped.

• the action performed is determined by the value of the dropdown “action” panel (see image bellow). By default the action is the one suggested by the “assistant” (eg the agent helping the human), but you can replace it with:

  • “do nothing”: the do nothing action will be performed

  • “assistant” (default): let the assistant (if any has been entered) chose. (if no assistant is present, it is replaced with “do nothing”)

  • “previous”: always do the same action.

Once you click on it, you the application will be in a “computing” state, you can see it with the appropriate flag:

When this is happening, we do not recommend you to interact with the application anymore: do not press any button.

Warning

TL;DR: Do not attempt to press any “grey” button.

End

It behaves the same way as “+12” (see section above). The only difference is that it does not compute a certain number of steps but rather tries to go until the “end”. In other words, it will stop the computation when:

• the assistant raises an alarm

• there is a game over (total blackout)

• the scenario is over (you successfully managed the entire scenario)

Choose an action

There are multiple ways to chose an action, that you can control from the “Action” part of the UI:

You can chose an action using different methods.

Automatic actions

When you select some action types in the approppriate dropdown:

• “do nothing”: in this case you ask the app to not take any action (you do nothing)

• “previous”: always do the same type of action, chose an action once, and it will be done again and again

• “assistant” (default): let the “assistant” (an AI) chose an action (by defaut not agent are given, in this case it will do nothing)

Manual action

The UI lets you define some manual actions that you want to perfom on the grid.

To select which action you want to do, you can click on the “real time” graph of the grid:

Most of the time these actions ar straightforward. If you want to modify the setpoint of a storage unit, click on it, and modify it with a slider. If you want to disconnect a powerline, click on it and select “disconnect” etc.

Note

We remind that more information for the action are availble on the dedicated page: Perform action on the powergrid

Line status

To do such action, you first need to clic on a powerline. For example, for the environment displayed, you can click in each on the arrow:

And then (in this example we clicked on the higher powerline) this appear:

And you can confirm which action you took:

Redispatching

These actions can be made by clicking on a dispatchable generator (for example of type “nuclear”, “hydro” or “thermal”).

They consist in increasing / decreasing the amount of power the produce.

In our example, there are given there:

And you can chose the amount of redispatching you perform by moving the slider. As for the other type of action, you have a text representation of this action.

Curtailment

These actions can be made by clicking on generator coming from renewable energy sources (for example “solar” and “wind”).

They consist in limiting the amount of power that can be produced by this generator.

In the example grid, you can click on these one:

You can chose the limit you want to impose by moving the slider (to the right if you want to “limit more” - eg when a problem arises* - or to the left if you want to “limit less” - eg when a problem is over and you want to allow as much renewable energy on the grid).

Topology

For making topological actions, you have to click on substations, represented by large disks on the interface:

In this graphical user interface, each object at the substation you clicked on are visible in a “zoomed in” plots, as shown here:

On this plot, each “radial line” represents an object connected to this substation as shown here:

On each substation, in grid2op, you have the choice to assign elements on one of the two “buses”. Each being represented by a circle inside the substation, as shown here:

In the GUI it can be achieved by clicking on the intersection of each elements (shown as straight lines) and the bus to which you want to connect it (represented by circles). And as always, the resulting action is shown (as text) on the right.

In this example, we clicked on the 3 circles showed bellow:

Storage units

These actions can be achieved by clicking on storage units, as shown here.

And as for many possible actions in the GUI, you can modify it by acting on the slider:

• to the left: you ask the storage unit to produce something. It will empty the storage unit and provide some power to the grid (behaving similarly as a generator). It is only possible if the storage unit is not empty.

• to the right: you ask the storage unit to absorb power. It will charge the storage unit and take some power from the grid (behaving like a load). It is only possible if the storage unit is not full.