数据驱动的大规模风电场智能聚合等值建模方法

A Data-driven Method for Intelligent Aggregated Equivalent Modeling of Large-scale Wind Farms

摘要:

大规模风电场接入新型电力系统时，仿真面临风机数量众多、模型复杂带来的精度与效率难以兼顾的问题。为此，提出了一种基于数据驱动的风电场智能聚合等值建模方法，在保证风电场动态特性不失真的前提下提高仿真效率。首先，基于库普曼理论提出一种可解释性的风机特征值提取方法，构建了动态特性视角下的多风机分群规则；其次，利用Autoformer网络建立了基于风电集群端口量测数据的集群等效模型；最后，基于CloudPSS平台，设计并搭建了等效模型的接入接口，实现了统一仿真任务。通过在含百台风机接入的典型风电场场景中进行验证，所提方法在等效精度与计算效率上均优于现有方法，能够满足工程应用需求。

Abstract:

The integration of large-scale wind farms into new power systems poses significant challenges for simulation, the vast number of wind turbines and model complexity make it difficult to achieve both precision and efficiency. In response, a data-driven intelligent aggregated equivalent modeling method is proposed for wind farms, which can improve simulation efficiency while ensuring that the dynamic characteristics of wind farm are not distorted. Firstly, an interpretable method for extracting turbine eigenvalues is introduced based on Koopman theory, enabling the formulation of a multi-turbine clustering strategy from the perspective of dynamic behavior. And then, an equivalent model for wind farm cluster based on the measurement data from cluster ports is constructed using the Autoformer network. Finally, an access interface for the equivalent model is developed and integrated into the CloudPSS platform, facilitating the unified simulation tasks. Validation with a typical wind farm comprising hundreds of wind turbines demonstrates that the proposed method surpasses the current approaches in both equivalent accuracy and computational efficiency, which can meet the requirements of engineering applications.