基于PMU的电力系统状态估计与卷积神经网络异常检测OA
PMU-based Power System State Estimation and Convolutional Neural Network-based Anomaly Detection
为进行电力系统监测,提出一种卷积神经网络(CNN)数据过滤器,运用Nesterov Adam梯度下降和分类交叉熵损失方法对相量测量单元(PMU)数据进行验证,以识别针对状态估计器的异常数据流.为评估过滤器性能,将其与深度学习算法以及传统分类器进行比较,在IEEE-30和IEEE-118总线系统上进行实验模拟.实验结果表明,CNN过滤器能有效识别旨在篡改状态估计的伪造数据流,作为额外安全层为决策支持和电网稳定运行提供保障,优于循环神经网络(RNN)、长短期记忆(LSTM)网络和其他传统分类器.
For power system monitoring,a convolutional neural network(CNN)data filter is proposed to validate phasor meas-urement unit(PMU)data by applying the Nesterov Adam gradient descent algorithm and the categorical cross-entropy loss function to identify anomaly data streams targeting the state estimator.To evaluate the performance of the proposed filter,it is compared with other deep learning algorithms as well as traditional classifiers,and experimental simulations are conducted on IEEE-30 and IEEE-118 bus systems.The results demonstrate that the CNN-based filter effectively detects forged data streams intended to tamper with state estimation,serving as an additional security layer for decision support and stable grid operation.Furthermore,it outperforms the recurrent neural network(RNN),the long short-term memory(LSTM)network,and other conventional classifiers.
孙萌;黄宇;王奇;肖耀辉;胡明辉
中国南方电网有限责任公司超高压输电公司电力科研院,广东,广州 510700中国南方电网有限责任公司超高压输电公司,广东,广州 510700中国南方电网有限责任公司超高压输电公司电力科研院,广东,广州 510700中国南方电网有限责任公司超高压输电公司电力科研院,广东,广州 510700中能国研(北京)电力科学研究院,北京 100055
信息技术与安全科学
卷积神经网络错误数据注入混合状态估计多变量时间序列
convolutional neural networkfalse data injectionhybrid state estimationmulti-variate time series
《微型电脑应用》 2026 (3)
1-4,4
国家自然科学基金项目(61602530)
评论