Communication-efficient distributed strategy for reactive power optimisation considering the uncertainty of renewable generation

Due to the non-linear power flow equality constraints and uncertain injected power of renewable generation, the reactive power optimisation problem is stochastic and non-convex, which is hard to be solved efficiently. Therefore, the discrete probability model of renewable generation is utilised to build the multi-scenario deterministic formulation of the stochastic problem, which is further changed to a conic optimisation model via the second-order conic relaxation. By using the skill of splitting buses and the augmented Lagrangian decomposition, the problem is separated into several smaller manageable sub-problems. Moreover, on the basis of the accelerated local augmented Lagrangian, a fully distributed conic optimisation algorithm is developed and the global optimal solution can be obtained iteratively in a parallel fashion. During the iterative procedure of the algorithm, only partial local message, without central coordination, need to be exchanged between the neighbouring subsystems. Each subsystem is interested in knowing only some components rather than the entire information, which improves communication efficiency. Simulations verify the effectiveness of the proposed algorithm.