`sustaingym.algorithms.evcharging.baselines`#

This module implements baseline algorithms for EVChargingEnv.

Module Contents#

`GreedyAlgorithm`	Per-time step greedy charging. Whether the action space is continuous or
`RandomAlgorithm`	Random action.
`MPC`	Model predictive control.
`OfflineOptimal`	Calculates best performance of a controller that knows the future.

`MAX_ACTION`
`D_MAX_ACTION`

class sustaingym.algorithms.evcharging.baselines.GreedyAlgorithm(env: sustaingym.envs.evcharging.EVChargingEnv)[source]#

Per-time step greedy charging. Whether the action space is continuous or discrete, GreedyAlgorithm outputs the maximum pilot signal allowed.

get_action(observation: dict[str, Any]) → numpy.ndarray[source]#

Returns greedy charging action.

class sustaingym.algorithms.evcharging.baselines.RandomAlgorithm(env: sustaingym.envs.evcharging.EVChargingEnv)[source]#

Random action.

get_action(observation: dict[str, Any]) → numpy.ndarray[source]#

Returns random charging action.

class sustaingym.algorithms.evcharging.baselines.MPC(env: sustaingym.envs.evcharging.EVChargingEnv, lookahead: int = 12)[source]#

Model predictive control.

See BaseAlgorithm for more attributes.

Parameters:

lookahead#: number of timesteps to forecast future trajectory. Note that MPC cannot see future car arrivals and does not take them into account.

get_action(observation: dict[str, Any]) → numpy.ndarray[source]#

Returns first action of the MPC trajectory.

class sustaingym.algorithms.evcharging.baselines.OfflineOptimal(env: sustaingym.envs.evcharging.EVChargingEnv)[source]#

Calculates best performance of a controller that knows the future.

Parameters:: env (sustaingym.envs.evcharging.EVChargingEnv) – EV charging environment

reset() → None[source]#

Reset timestep count.

get_action(observation: dict[str, Any]) → numpy.ndarray[source]#

On first call, get action for all timesteps (should be directly after environment reset).