eduzhai > Applied Sciences > Engineering >

Deep Learning-Aided MPC of Wind Farms for AGC Considering the Dynamic Wake Effect

  • Save

... pages left unread,continue reading

Document pages: 8 pages

Abstract: To provide automatic generation control (AGC) service, wind farms (WFs) arerequired to control their operation dynamically to track time-varying powerreference. Wake effects impose significant aerodynamic interactions amongturbines, which may remarkably influence the WF dynamic operation. Thenonlinear and high-dimensional nature of dynamic wake flow, however, bringsextremely high computation complexity and obscure the design of optimal WFcontrol coping with dynamic wake effects. To solve this problem, this paperproposes a deep learning-aided model predictive control (MPC) method.Leveraging recent advances in computational fluid dynamics (CFD) to providehigh-fidelity data that simulates WF dynamic wake flows, two deep neuralnetwork (DNN) architectures are designed to learn a dynamic WF reduced-ordermodel (ROM) that captures the dominant flow dynamics from high-dimensional CFDdata. Then, an MPC framework is constructed that incorporates the learned WFROM to coordinate different turbines while considering dynamic wind flowinteractions. Case studies validate the accuracy of the DNN-based WF ROM andshow the effectiveness of the proposed strategy for WF AGC performance. Therange of traceable power reference is improved compared to the traditionalcontrol scheme. By reducing the number of model states by many orders ofmagnitude, the proposed method is computationally practical for controlapplications.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×