考虑注意力机制的CNN-LSTM高渗透风电并网暂态电压稳定性研究

Study on high penetration wind power integration and transient voltage stability in power grid using CNN-LSTM with attention mechanism

为了能够快速且精确地判断出风电场接入电网后系统暂态电压稳定性,文章基于注意力机制提出一种卷积-长短时记忆网络(Convolutional Neural Networks-Long Short-Term Memory,CNN-LSTM)暂态稳定评估指标。为了更好地捕捉输入数据中空间和时间的相关性,基于核主成分分析(Kernel Principal Component Analysis,KPCA)进行特征降维;针对高比例的新能源电网中,整个系统的短路容量下降、短路电流水平攀升问题,提出了安装超导故障限流器的主动支撑措施,限制故障过程短路电流水平,维持并网点电压稳定。最后,在PSD-BPA中搭建含风电的IEEE39节点系统进行仿真计算和数据采集。结果表明,KPCA方案能有效筛选电力系统暂态稳定评估中重要度高的特征,所提评估指标具有更高的辨识能力,所提改进措施对高比例风电并网系统暂态电压稳定具有积极作用。

In order to achieve rapid and accurate assessment of transient voltage stability in the power system following the integration of wind farms into the grid,a transient stability assessment metric is proposed based on Convolutional Neural Networks-Long Short-Term Memory(CNN-LSTM)and attention mechanisms.To better capture spatial and temporal correlations in the input data,feature dimensionality reduction is carried out using Kernel Principal Component Analysis(KPCA).Addressing challenges related to decreased short-circuit capacity and increased shortcircuit current levels in high-proportion renewable energy grids,an active support measure is introduced by installing superconducting fault current limiters to restrict short-circuit current levels during fault processes and maintain voltage stability at grid connection points.Finally,simulations and data collection are performed on an IEEE39-node system with wind power integration using PSD-BPA.The results indicate that the KPCA approach effectively screens features of significant importance in the transient stability assessment of power systems.The proposed evaluation metric demonstrates higher discriminative capability,and the suggested improvement measures are observed to play a positive role in enhancing transient voltage stability in high-proportion wind power integration systems.