Multi-armed Bandit Based Load Aggregation for Power System Frequency Regulation

The intermittence and stochasticity of distributed generators (DG) bring new challenges to the stability of power system. While thermostatically controlled loads can provide efficient frequency regulation for power system operation stability. This paper proposes a novel risk-averse multi-armed bandit theory based frequency regulation framework considering users’ random and uncertain behavior. The framework consists of online learning residents’ electricity consumption behavior and aggregation residential air conditioning system. Owing to these two control system, residential loads can be aggregated to achieve primary and secondary frequency regulation. The experimental simulation based on IEEE RTS 24 bus system verifies that the proposed framework has good frequency regulation performance with fewer users being selected and opting out of the control.