Distributed resource low-delay interaction method for low-voltage substation based on communication quality-aware reinforcement learning

In the distribution substation, edge computing can optimize the allocation and utilization of overall distributed resources, which is essential for ensuring efficient energy utilization and stable operation of the power grid. However, because of electromagnetic interference during data transmission at the device, the traditional upper confidence bound (UCB) algorithm neglects the impact of communication quality when estimating confidence bounds, leading to a deviation in the value estimation of actions and an increase in data processing delay. To tackle this issue, a distributed resource low-delay optimization algorithm for substations based on communication quality-aware UCB is proposed. Firstly, we propose the low-delay interaction architecture for low-voltage substations with communication quality-aware distributed resources. Moreover, by optimizing the data migration strategy, the problem of minimizing edge-edge collaborative data processing delay is formulated. To address this issue, we convert it into a multi-armed bandit (MAB) problem and solve it by using the proposed algorithm. This algorithm incorporates communication quality weight into the calculation formula for edge server preference value, enabling real-time awareness of migration channel quality during the data migration process and improving the efficiency of distributed resource interaction in the distribution substation. Finally, the proposed algorithm’s effectiveness is confirmed through simulation.