基于多源数据驱动的电力系统暂态稳定性分析方法

Transient Stability Analysis Method of Power System Based on Multi-source Data Drive

【目的】当前,以深度学习为代表的数据驱动方法已广泛应用于电力暂态稳定性分析中。然而,现有研究数据驱动的暂态稳定模型在面对小样本、弱样本等实际场景时,存在泛化能力有限、模型精度不足等问题。为了提高模型的表达能力,提出一种基于运行数据和故障数据的精细化暂态稳定评估方法。【方法】首先,根据电力系统暂态稳定机理模型构建故障时间、故障位置、受扰线路和负荷水平4个故障信息特征。然后,提出并行融合和串行融合两种特征融合方式,实现运行特征和故障特征的统一表达,并对多源特征融合方式对暂态稳定分析模型的影响进行深入分析。【结果】新英格兰系统算例的实验结果表明,基于多源数据混合驱动的暂态稳定分析方法有利于提高暂态稳定评估模型的准确度,在面对小样本、弱样本等实际场景时仍具有较高的准确率。

【Purposes】At present,the data driven method represented by deep learning has been widely used in power transient stability analysis.However,the existing transient stability models for researching data driving have some problems,such as limited generalization ability and insufficient model accuracy,when facing small samples,weak samples,and other actual scenarios.In order to improve the expression ability of the model,a refined transient stability assessment method is proposed in this paper according to operation data and fault data.【Methods】First,four fault information characteristics,namely fault time,fault location,disturbed line,and load level,are constructed according to the transient stability mechanism model of power system.Then,two feature fusion methods,parallel fusion and serial fusion,are proposed to realize the unified expression of operation features and fault features.The influence of multi-source feature fusion on transient stability analysis model is analyzed in depth.【Findings】The experimental results of the New England system example show that the transient stability analysis method based on multi-source data hybrid drive is conducive to improving the accuracy of the transient stability assessment model,and still has a high accuracy in practical scenarios such as small sam-ples and weak samples.