This paper proposes a microgrid optimization strategy for new energy charging and swapping stations using adaptive multi-agent reinforcement learning, employing deep reinforcement learning methods to achieve coordinated control of new energy output and charging–swapping loads, effectively reducing the fluctuation of new energy grid power.
In, operates in a flywheel storage power plant with 200 flywheels of 25 kWh capacity and 100 kW of power. Ganged together this gives 5 MWh capacity and 20 MW of power. The units operate at a peak speed at 15,000 rpm. The rotor flywheel consists of wound fibers which are filled with resin. The installation is intended primarily for frequency control. This service is sold.
