Learning-based flexible load aggregation for secondary frequency regulation in co-simulated transmission and distribution networks

Aggregated flexible loads offer a promising solution for secondary frequency regulation (SFR) in power systems with increasing intermittent renewable energy sources. However, uncertainties in users’ behaviors may create a mismatch between the aggregated power of flexible loads and the control target of SFR. Furthermore, as these loads are dispersed across distribution networks, distribution network’s topology and its interplay with the transmission network may affect the performance of aggregated flexible loads in SFR. Therefore, this paper proposes an adaptive combinatorial multi-armed bandit (CMAB) flexible load aggregation strategy to enhance SFR performance in co-simulated transmission and distribution (T&D) networks. First, a dynamic T&D co-simulation framework is proposed based on the HELICS platform. Then, the combinatorial upper confidence bound-average (CUCB-Avg)-based CMAB algorithm is employed to manage users’ uncertain responses. Case studies on the IEEE 14-bus system with five IEEE 8,500-node feeders demonstrate the effectiveness of the proposed framework and method. The SFR performance of the proposed strategy based on CUCB-Avg algorithm outperforms the average and CUCB strategies in terms of accuracy, rapidity, robustness, and the number of affected users.