Source code for smartgrid.rewards.numeric.differentiated.over_consumption

from smartgrid.agents.agent import Agent
from smartgrid.rewards.reward import Reward
from smartgrid.world import World




class OverConsumption(Reward):
    """
    Reward representing the *over-consumption* of an Agent.

    The *over-consumption* is the quantity of energy that was consumed by
    the society of agents, but which was not available in the grid.
    (We assume that the grid automatically buys from the national grid to
    compensate, which has a negative impact. Over-consumption should thus be
    avoided).

    We compare the quantity of energy *taken* (i.e., consumed + stored from
    the grid) by all agents to the quantity of energy over-consumed by all
    agents. This gives us a *global* component (the current environment).

    Then, we compare the quantity of energy over-consumed, minus the agent's
    taken energy, and we compare to the sum of energy taken by all agents.
    This gives us a *local* component (the hypothetical environment, had the
    agent not acted).

    The reward follows the Difference Reward principle, and thus is the global
    component minus the local component.
    """

    name: str



    def __init__(self):
        super().__init__()




    def calculate(self, world: World, agent: Agent):
        # The energy taken and given from/to the grid by each agent
        sum_taken = 0
        sum_given = 0
        for a in world.agents:
            sum_taken += a.enacted_action.grid_consumption
            sum_taken += a.enacted_action.store_energy
            sum_given += a.enacted_action.give_energy

        # Total quantity of energy over-consumed (if negative, clipped to 0).
        global_oc = max(0, sum_taken - sum_given - world.available_energy)

        # Global reward: proportion of over-consumption compared to sum of taken
        global_reward = 1.0 - global_oc / (sum_taken + 10E-300)

        # Local reward: proportion of over-consumption (minus what the agent
        # consumed) compared to sum of taken
        taken_by_agent = agent.enacted_action.grid_consumption + agent.enacted_action.store_energy
        local_reward = (global_oc - taken_by_agent) / (sum_taken + 10E-300)

        return global_reward - local_reward