METHOD FOR TRAINING A CONTROL DEVICE OF A CONTROLLED SYSTEM

According to various embodiments, a method is described for training a control device of a controlled system, the control device having a regulating device and an actuator operating according to a control strategy. The method comprises generating control actions by means of the control device, each control action being generated by detecting a measurement variable which indicates a state of the controlled system, determining a correction term of the actuator for the detected measurement variable according to the control strategy, the detected measurement variable is adapted using a correction term for the detected measurement variable, and the control action is generated by feeding the adapted measurement variable as an actual value to the control device. The method also has training the control strategy by reinforcement learning to maximize the revenue achieved by the generated control action. 按照各种实施方式，描述了一种用于训练受控系统的控制装置的方法，其中该控制装置具有调节装置和按照控制策略来工作的执行器。该方法具有：通过该调节装置来产生控制行动，其中每个控制行动都通过如下方式来被产生：检测说明受控系统的状态的测量参量