基于时延神经网络的风电场集总调压动态建模

Dynamic Modelling of Lumped Voltage Regulation Characteristics for Wind Farm Based on Time-delay Neural Network

风电场并网调压依赖于场内风电机组无功支撑能力,而机组分散运行及其工况差异使场级集总调压动态表征面临很大挑战,不利于场级调压特性监测及调压控制器的改进设计。针对这一问题,本文基于数据驱动时延神经网络算法提出一种风电场全工况下集总调压特性动态等值建模方法。首先,基于风电场集总调压原理及策略分析明确调压响应的输入和输出,根据多工况下调压实验数据,采用传递函数、子空间辨识的线性建模及间隙测度方法评估各工况下调压动态差异度。然后,根据场级调压动态超短时响应特点,提出基于时延神经网络的全工况集总动态等值建模方法。最后,根据东北某风电场调压运行数据进行仿真验证,仿真结果表明,所提方法可高精度逼近具有超短时采样周期的场级集总调压特性,有效提升风电场并网友好性。

The grid-connected voltage regulation of wind farms depends on the reactive power support capacity of wind turbines in the field,and the dispersed operation of units and the differences in working conditions make the dynamic characterization of field-level lumped voltage regulation face great challenges,which is not conducive to field-level voltage regulation characteristicsand the improved design of voltage regulation controllers.To solve this problem,a dynamic equivalent modelling method for the lumped voltage regulation characteristics of a wind farm under all working condi⁃tions is proposed in this paper,which is on the basis of a data-driven time-delay neural network algorithm.First,the in⁃put and output of voltage regulation response are clarified based on the principle and strategy analysis of the wind farm lumped voltage regulation.Then,based on the voltage regulation experiment data under multiple working conditions,the linear modelling and gap measurement methods of transfer function and subspace identification are adopted to evaluate the dynamic difference degree of voltage regulation under various working conditions.According to the characteristics of field-level voltage regulation dynamic ultra-short-time response,a lumped dynamic equivalent modelling method based on time-delay neural network under all working conditions is put forward.Finally,simulation verification is carried out based on the voltage regulation operation data of one wind farm in Northeast China,and results show that the proposed method can approximate the field-level lumped voltage regulation characteristics with an ultra-short sampling period at a high accuracy,thus effectively improving the grid-connection friendliness of the wind farm.