Decentralised optimisation for large offshore wind farms using a sparsified wake directed graph

Abstract

Considering the wake effects, developing an efficient wake-based graph clustering algorithm is essential for the decentralised coordinated online control of large offshore wind farms. In this study, the idea of graph sparse is introduced into the algorithm, in which the sparsified wake-directed graph is generated while preserving the original wake-directed graph critical wake coupling relationship between turbines. The key is to construct the original wake-directed graph by quantifying wake intensity and applying the graph sparseness constraints algorithm to achieve an ultra-sparse wake subgraph. Based on the sparsified wake-directed graph and wake intensity weighting matrix, the wind farm is split into almost uncoupled cluster subsets to establish decentralised sparse communication architectures. By doing so, a decentralised sequential quadratic programming optimisation strategy is proposed to solve a nonconvex optimisation problem, in which the thrust load problem is converted into a power controlling optimisation problem, and the thrust load distribution is balanced using the power margin. Simulation results reveal that the proposed scheme can maximise wind farm power production while minimising thrust loads in various turbulence intensities, making it functional for real-time operations on a large-scale wind farm.