Concept of hybrid modeled digital twins and its application for an energy management of manufacturing systems

Sorry you have no rights to view this entry!

3 Comments

  1. Dear Mr. Langlotz,

    My question for you is the following.
    In your presentation, the use of a reinforcement learning approach is mentioned for decision making on utilization of energy from either the battery or the public power grid.
    Could you give some clarification on the “rewarding” utilised in your reinforcement learning application?

    Thanks for your interesting presentation and best regards.

    Roberto Teti

    • Hello Mr. Teti,

      thank you for your question. The reward depends on several system factors, like the state of charge, the predicted energy price as well as the assumed energy consumption of the model scaled factory. For example, the algorithm receives a negative reward if the algorithm wants to charge the battery in the case the state of charge is about 100 % or if the algorithm wants to charge the battery in the case of high energy prices. Otherwise, a positive reward is gained if the energy price is high and the model scaled factory is powered by the battery. The result of the reward function can be positive or negative for a positive or negative reward.
      Best regards
      Pascal Langlotz

  2. Hello Mr. Teti,

    thank you for your question. The reward depends on several system factors, like the state of charge, the predicted energy price as well as the assumed energy consumption of the model scaled factory. For example, the algorithm receives a negative reward if the algorithm wants to charge the battery in the case the state of charge is about 100 % or if the algorithm wants to charge the battery in the case of high energy prices. Otherwise, a positive reward is gained if the energy price is high and the model scaled factory is powered by the battery. The result of the reward function can be positive or negative for a positive or negative reward.
    Best regards
    Pascal Langlotz

Leave a Reply

Your email address will not be published. Required fields are marked *