Peer-to-Peer Market and DSO Coordination Simulation on GPU

The energy transition requires increasing renewable energy production and consumer flexibility. These changes will likely take place in the distribution grid. In addition, consumer flexibility should be managed in a decentralized way to respect people’s privacy. Solutions for a decentralized market that can handle the physical transport state already exist in the literature. Nevertheless, these implementations are not suited for the distribution grid as they do not scale up well with the problem size when simulated on a single machine. In this contribution, we propose an endogenous Peer-to-peer market on a graphics processing Unit (GPU) to allow the scaling up of these algorithms. The algorithm relies on a consensus between a peer-to-peer market and the distribution system operator that solves an optimal power flow. As the bottleneck of this computation is the optimal flow, we used the speed-up brought by the GPU parallelization to conduct a study on significant test cases to gain time by modifying the interactions. It has been shown that if the coordination is poorly designed, it can prevent the computation from converging. Nevertheless, the average computation time in the European TestFeeder can be halved with a different coordination. The open-source framework of this article allows anyone to conduct a parametric study on their case.