Asynchronous algorithm of an endogenous peer-to-peer electricity market

Abstract

chronous versions of an optimal power flow (OPF) were studied. In [5], the asynchronous, ADMM based, distributed OPF reduced significantly the convergence time while also increasing the gap of the objective function. In [6], an ADMM based distributed OPF was also studied, but this time on a lateral branching network. Different strategies were investigated to replace the loss of information due to the asynchrony, which lead to fluctuations in the solutions. A Lagrangian relaxation based decentralized OPF was studied in [7], in which individual markets communicate with each other asynchronously. The asynchronous updates lead to the optimal solution with up to 50% time reduction. The convergence of the asynchronous resolution of a multi-time step energy optimization problem in a microgrid, using ADMM is discussed in [8]. The results may deviate from the optimal solution if the communication delays are too high. In all of these decentralized schemes, there is a communication link between agents that are physically connected by a power line, as opposed to our algorithm in which the communication links are independent of the power network topology.After formulating the endogenous peer-to-peer market in II, we propose in III an asynchronous algorithm to avoid communication or computation latencies. The simulation platform presented in IV is based on a 31 agents testcase. The simulation results discussed in V confirm that the asynchronous algorithm converges towards the same solution as the synchronous one regarding the power dispatch. The impact of various delays and asynchronism parameters are further investigated. II. ENDOGENOUS PEER-TO-PEER MARKET A. AC regularized peer-to-peer electricity marketThe problem addressed in this paper is an optimal power flow based on a peer-to-peer market clearing algorithm described in [1], consisting of N agents part of a community Ω, performing multi-bilateral trades with each other. A system operator (SO) is added here to verify that the AC networks constraints are respected and to induce the market peers to change the solution otherwise. In order to account for the network losses, a new agent is added to the peer-to-peer market, called the loss provider. This agent is buying the losses to the producers of the market. We note the extended set of peer-to-peer market agents Ω * = Ω ∪ {Loss} with Abstract-As the number of actors in the electricity market increases, peer to peer decentralized markets gain interest despite facing two challenges. First, the amount of exchanged messages increase significantly, w hich m akes t he r esolution d ependent on communication delays, as well as computation delays. Second, the physical limitations of the power network are not taken into account. An asynchronous implementation of an endogenous peerto-peer market is here introduced to address these difficulties. The proposed algorithm is tested on a 31 agents testcase and a study of the influence o f t he v arious a lgorithm p arameters on the convergence time is performed.

GPU Optimisation of an Endogenous Peer-to-Peer Market with Product Differentiation

“…However, before a possible actual deployment, numerous simulations are required from the research community, network managers, and normative authorities. In a non-exhaustive way, it will be necessary to judiciously parameterize the distributed management mechanisms [8], to verify the behavior during extreme events [9], or to ensure the robustness of the proposed solutions in case of faults on the power or communication network [10]. All these simulations cannot yet rely on the computational power provided in a distributed manner and must therefore be performed centrally on a single machine or computing cluster.…”

“…In particular, because of the diversity of mechanisms currently proposed in the literature, any contribution to this domain bears the risk of being specific to a particular case. To overcome this point, we choose here the so-called endogenous market problem [10], [14], [15]. This solves a peer-to-peer market (P2P) constrained as an OPF by the electrical network's physical equations and limits.…”