Optimal scheduling energy for 'wind-solar-load-storage' AC-DC hybrid distribution network system based on multi-agent algorithm

Aiming at the real-time optimisation problem of AC/DC hybrid distribution network with high proportion of new energy access, a 'wind-solar-load-storage' collaborative scheduling framework based on multi-agent reinforcement learning (MARL) is proposed. Firstly, the Markov game model is constructed, and wind power, photovoltaic (PV), energy storage and flexible load are modelled as heterogeneous agents, and a mixed action space integrating DQN (Deep Q-Network) and Actor-Critic is designed, and the federated-edge collaborative mechanism is introduced to realise the privacy protection training of 'data-fixed model moving'. The single step decision-making time is less than 70 ms, and the voltage fluctuation is strictly controlled within ±5%. It achieves the coordinated optimisation of economy, safety, and privacy, providing a new paradigm for real-time scheduling of high proportion new energy distribution networks.