Non-centralised coordinated optimisation for maximising offshore wind farm power via a sparse communication architecture

Abstract

The motivation for establishing sparse communication architecture is to reduce the complexity and size of the power optimisation problem being solved by centralised controllers of a large-scale offshore wind farm. In this regard, this article introduced the sub-graph extraction technique to construct sparse communication architectures while ignoring the weakly coupled wake effects between turbines to develop two non-centralised coordination optimisation strategies for maximising power output. The non-centralised optimisation framework mainly uses the idea of graph decomposition that creates an original wake digraph by quantifying wake intensity, defining the spectral norm of the wake weight matrix as a constraint to obtain pruned wake sub-digraph. Then, for determining communication neighbours, we used a depth-first tree search arithmetic to cluster subsets of turbines by which the impact of neighbouring turbines was considered. On this basis, designing a suitable optimisation-based distributed or decentralised strategy, defining shared variables and consensus stages among the local controllers, yaw angles from such controllers are optimally coordinated to maximise the total power of the whole wind farm. Finally, a case study of a wind farm with 36 turbine simulation results showed that the scale of the non-convex optimisation problem could be efficiently reduced without affecting the power output while reducing execution time under various turbulence intensities.