集电系统拓扑优化是大型海上风电场规划建设的核心问题,本质上是一个涉及多约束、多目标的复杂混合整数优化问题.针对该问题,提出了一种基于大语言模型(large language model,LLM)辅助的大型海上风电场集电系统拓扑优化方法.首先,基于大语言模型辅助对风电机组群进行聚类,通过链式提示法使LLM理解优化目标,并利用LLM将大型海上风电场分割为若干小型区域,以降低优化问题维度,提升求解速度和质量.然后,构建集电系统拓扑优化模型,基于混合整数线性规划求解器,获得海上风电场的最优集电系统拓扑设计方案.最后,利用1个含有75台风电机组的大型海上风电场系统进行方法性能验证,仿真结果表明:与传统优化技术相比,所提方法获得的聚类风机数量更加均衡,在考虑拓扑功率损耗的同时,生成的拓扑方案经济性最优.LLM在集电系统拓扑辅助优化中具有较高的有效性,为海上风电场集电系统拓扑设计优化提供了一种新思路.
Topology Optimization of Large-scale Offshore Wind Farm Collector Systems Based on Large Language Models
The optimization of the collector system topology is a core issue in the planning and construction of large-scale offshore wind farms(LSOWF),and it is inherently a complex mixed-integer optimization problem involving multiple constraints and objectives.To address this problem,this paper proposes a method for optimizing the collector system to-pology of LSOWF assisted by large language model(LLM).Firstly,groups of wind turbines(WT)with the assistance of LLM are clustered and the method of chain prompting is applied to help the LLM understand the optimization objectives.Then,the LLM is employed to divide LSOWFs into smaller areas.This reduces the dimensions of the optimization prob-lem and enhances both solving speed and quality.Next,the optimization model for the collection system topology is constructed and a mixed-integer linear programming solver is utilized to achieve the optimal design for the LSOWF's collector system topology.Finally,the method's performance is validated using a LSOWF with 75 WTs.The simulation results show that,compared to traditional optimization techniques,the proposed method can be adopted to achieve more balanced clustering of WTs and generate a topology design that is economically optimal while considering power loss.The LLM demonstrates high effectiveness in assisting with the collector system topology optimization,providing a new approach for LSOWF collector system design optimization.
large language modelcollector system topology optimizationoffshore wind farmmixed-integer linear programmingclustering algorithm
万方数据
维普
