INTEGRATED ACTIVE-REACTIVE POWER DISPATCHING MODEL BASED ON DEEP REINFORCEMENT LEARNING

The integrated active and reactive power dispatching model based on deep migration reinforcement learning is proposed. Reinforcement learning can guide the agent to acquire optimal scheduling knowledge in the source task, and the migration network can realize the application of the knowledge learned in the source task to new operating scenarios. Through the simulation of IEEE 14 buses, the design method of key elements of agent is given, including the construction of state, action and reward function. By calculating the similarity of the scene, the dispatching strategy under the new scene can be quickly obtained.