轨迹分析方法与长短期记忆网络的电网暂态稳定裕度评估

Power frid transient stability margin assessment based on long-short term memory networks

为了能够快速、准确地判断电力系统发生暂态故障后系统稳定性,提出了基于轨迹分析方法和长短期记忆网络的暂态稳定评估方法。以故障前、故障中和故障清除后的测量数据,构建全阶段时间序列的输入特征模型,利用轨迹分析法构造暂态稳定指标,定量评估发电机的暂态稳定裕度,在给定预想故障集下,给出LSTM神经网络刻画电网特征量与发电机稳定指标之间的映射关系,采用准确率、漏判率和误判率评价神经网络的暂态稳定区分能力。结果表明,LSTM模型的评估准确率最高97.50%,误判率和漏判率均低于其他模型,验证了长短期记忆网络可以通过其学习机制深度挖掘时序信息间的关联关系。

This paper proposes a transient stability assessment method based on trajectory analysis method and long-short term memory network,method designed for a quick and accurate evaluation of the stability of the power system following the occurrence of the transient faults.The study is focused on the efforts to use the measurement data before,during and after the fault clearance to construct the input characteristics of full-stage time series;to use the trajectory analysis method to construct the transient stability index for quantitatively evaluating the transient stability margin of the generator;given the expected fault set,construct a LSTM neural network to describe the mapping relationship between power grid characteristics and generator stability index;and to use accuracy rate,missed judgment rate and misjudgment rate to evaluate neural network’s ability to distinguish transient stability.The results show that the LSTM model boasts the highest evaluation accuracy rate of 97.50%,and a lower false positive rate and the missed positive rate than other models,validating the ability of the long and short-term memory network to deeply mine the association relationship between time series information through its learning mechanism.