目的 电力系统面临着虚假数据注入攻击的威胁，而已有的虚假数据注入攻击检测方法存在特征学习能力不足和检测速度较慢的问题，以至于无法对虚假数据注入攻击进行快速精确定位，因此提出一种基于Levy麻雀优化深度极限学习机的电网虚假数据注入攻击定位检测方法。 方法 所提方法将深度极限学习机作为特征提取算法和基础分类器来实现对攻击的快速精确定位；同时采用具有强局部搜索能力、融入Levy飞行策略的麻雀搜索算法对其初始权重与偏置进行优化，以进一步提高方法的定位检测精度。 结果 在IEEE-14和IEEE-57节点系统进行了大量仿真分析，所提方法的检测准确率在94%以上。 结论 与其他检测方法对比，所提方法具有更优的检测精度，可以实现更为快速的虚假数据注入攻击定位检测。

Objectives Power systems are facing threats of false data injection attacks. Existing detection methods for false data injection attacks have the problems of insufficient feature learning ability and slow detection speed, making it difficult to locate false data injection attacks rapidly and accurately. Therefore, this study proposes a method for locating false data injection attacks in power grid based on deep extreme learning machine optimized by Levy flying sparrow search algorithm. Methods The proposed method uses a deep extreme learning machine as the feature extraction algorithm and basic classifier to achieve rapid and accurate attack location. At the same time, a Levy flying sparrow search algorithm with strong local search ability is employed to optimize the initial weight and bias to further improve the location detection accuracy of the method. Results Extensive simulation analyses are conducted on IEEE-14 and IEEE-57 bus power systems. The proposed method achieves a detection accuracy rate of over 94%. Conclusions Compared with other detection methods, the proposed method demonstrates better detection accuracy and enables faster location detection of false data injection attacks.

在可再生能源高渗透率的背景下，电力系统的负荷频率控制（load frequency control，LFC）面临虚假数据注入攻击（false data injection attack，FDIA）的安全威胁。现有检测方法难以有效区分控制输入攻击和测量数据攻击，影响系统的稳定性和安全性。为此建立了包含可再生能源及储能系统的LFC状态空间模型，并分析了FDIA对系统动态特性的影响。通过状态空间分解方法将攻击信号解耦为控制输入攻击和测量攻击，提高检测精度。基于滑模观测器设计攻击估计方法，实现对攻击信号的实时检测。进一步结合 H ∞控制理论，提出了抗攻击控制（attack⁃resilient control，ARC）策略，以增强系统在攻击环境下的鲁棒性。仿真算例表明：与传统方法相比攻击估计均方误差降低约30 %，系统频率响应稳定性显著提升。结果表明，该方法能够有效检测FDIA并提高电力系统的安全性和抗干扰能力。

With the growing integration of renewable energy, load frequency control （LFC） in power systems faces security risks from false data injection attack （FDIA）. Existing detection methods struggle to differentiate control input attacks from measurement attacks, compromising system stability and security. This paper develops a state-space model for LFC incorporating renewable energy and energy storage systems and analyzes the impact of FDIA on system dynamics. A state-space decomposition method is employed to decouple attack signals into control input and measurement attacks, improving detection accuracy. A sliding mode observer-based attack estimation method is proposed for real-time detection. Additionally, an attack-resilient control （ARC） strategy is designed using $H_{\mathrm{\infty }}$ control theory to enhance system robustness. Simulations show that the proposed method reduces the attack estimation mean squared error by nearly 30% and significantly improves frequency response stability compared to traditional methods. These results demonstrate the method′s effectiveness in detecting FDIA and enhancing power system security.

虚假数据注入攻击严重威胁了电网安全稳定运行。由于电力量测数据维度高、特征复杂，传统攻击定位检测方法存在定位精度不足的问题。为此，提出一种基于相关特征-多标签级联提升森林的电网虚假数据注入攻击定位检测方法来精确定位电网受攻击的位置。所提方法通过融入极端梯度提升算法来增强多标签级联森林对复杂电力量测数据的拟合能力，进而识别系统各节点状态量的异常；引入“相关特征”算法来对原始电力量测数据中的高信息性特征进行提取，提升多标签级联森林的泛化能力，以获得更精确的定位检测。在IEEE-14和IEEE-57节点系统中进行仿真测试，验证了所提方法的有效性，且与其他方法相比，所提方法具有更优的准确率、查准率、灵敏度和F1分数。

False data injection attack seriously endanger the safety and stability of the power grid operations. Due to the high dimension and complex characteristics of the electricity measurement data, the attack locational detection accuracies of the existing methods are insufficient. For this reason, a false data injection attack locational detection method based on relevant features multi-label cascade boosting forest is proposed to locate the attacked position of the power grid. The proposed method enhances the fitting ability of the multi-label cascade forest processing the complex electricity measurement data by incorporating the extreme gradient boosting algorithm, so as to identify the abnormal state variables of each bus. Furthermore, the "relevant features" algorithm is integrated to extract the highly informative features from the original electricity measurement data to improve the generalization ability of the multi-label cascade forest, so as to obtain more accurate location detection. The simulation results on IEEE 14-bus and IEEE 57-bus test systems verify the effectiveness of the proposed method, and compared with many other methods, the proposed method has better accuracy, precision, sensitivity and F1-score.

虚假数据注入攻击对电力信息物理系统造成严重安全威胁。由于受到攻击样本与正常样本之间存在类别不平衡特性，导致机器学习检测方法偏向于多数类的预测，影响其对攻击的检测精度。为此，提出了基于Focal Loss^IM-Transformer的虚假数据注入攻击检测。Transformer利用其自注意力机制能够捕捉数据中的长期依赖性，进而识别不平衡的虚假数据注入攻击数据。Focal Loss^IM通过引入调制因子来更好地匹配虚假数据注入攻击样本的分布和特性，来增强检测方法对不平衡数据的识别能力，以提高检测方法对攻击的检测精度。通过在IEEE 14节点系统、IEEE 30节点系统和IEEE 57节点系统进行仿真，验证了所提方法的有效性。相较于传统损失函数和其他检测方法，所提方法显示出更好的泛化能力和对少数类的识别能力，且辨识精度高、误报率低。

False data injection attacks pose serious security threats to cyber-physical power system. Due to the class imbalance property between the attacked samples and the normal samples, machine learning detection methods tend to predict the majority of classes, which affects their detection accuracy of the attacks. Therefore, a false data injection attack detection based on Focal Loss^IM Transformer is proposed. Transformer utilizes itself attention mechanism to capture long-term dependencies in data, thereby identifies imbalanced false data injection attack data. Focal Loss^IM enhances the detection method's ability to identify imbalanced data by introducing modulation factors to better match the distribution and characteristics of false data injection attack samples, thereby improving the detection accuracy of the detection method for attacks.The effectiveness of the proposed method is verified through simulations on IEEE 14⁃node system, IEEE 30⁃node system, and IEEE 57⁃node system. Compared with traditional loss functions and other detection methods, the proposed method exhibits better generalization ability and recognition ability for minority classes, with high recognition accuracy and low false alarm rate.

In this paper, from the perspectives of defenders, we consider the detection problems of false data-injection attacks in cyber-physical systems (CPSs) with white noise. The false data-injection attacks usually modify the sensor data to make CPSs unstable and keep stealth for the χ detector. To guarantee system security, a novel detector, that is, the summation (SUM) detector, is proposed to detect the false data-injection attacks. Different from the χ detector, the SUM detector not only utilizes the current compromise information but also collects all historical information to reveal the threat. Its evaluation value also satisfies χ distribution when no attacks compromise the systems, and the false alarm rate can be restricted to less than any given value by choosing the proper threshold value. Furthermore, an improved false data-injection attack with a time-variable increment coefficient is introduced based on the existing approaches. The effects of the SUM detector are also verified for the traditional and the improved false data-injection attacks, respectively. Finally, some simulation results are given to demonstrate the effectiveness and superiority of the SUM detector.

随着新型能源互联网的发展,大规模的传感量测系统为基于数据驱动的虚假数据注入攻击检测方法提供了数据支持,然而攻击样本数据不平衡问题会影响此类方法的性能。提出了基于改进生成对抗网络(generative adversarial network, GAN)和极端随机树的数据重平衡攻击检测模型。首先,为了生成高质量数据,设计GAN的结构使其训练稳定;其次,使用Copula函数构建电力系统状态量之间的空间关联性以适应分布式能源的接入;然后,对改进的GAN进行对抗训练得到重平衡的数据集,采用极端随机树分类器实现攻击检测。此外,设计基于多种分类器的数据有效性指标评估生成数据的质量。通过对比实验对所提方法进行验证,结果表明该方法能生成高质量的量测数据,可以有效解决数据不平衡问题,攻击检测率达98.95%。

With the development of new-type energy internet, large-scale sensing measurement systems provide data support for data-driven detection of false data injection attack. However, the problem of unbalanced attack data will affect the performance of such methods. Therefore, a data rebalance attack detection model based on improved generative adversarial network (GAN) and extremely randomized tree is proposed. Firstly, the GAN structure is designed to make the training procedure stable enough to generate high-quality data. Secondly, the Copula function is used to construct the spatial correlation between the power system states to adapt to the integration of the distributed energy resources. Then, a rebalanced dataset is obtained through the adversarial training of the improved GAN, and the extremely randomized tree classifier is used to detect the attack. In addition, the data validity index based on multiple classifiers is designed to evaluate the quality of the generated data. The effect of the proposed method is verified by comparative experiments. Results show that the method can generate high-quality measurement data, solve the problem of data imbalance, and the attack detection rate is 98.95%.

虚假数据注入攻击通过篡改数据采集与监视控制系统采集的数据，进而破坏电力系统的稳定运行。传统虚假数据注入攻击检测方法无法对受攻击位置进行定位，亦或定位精度低。首先提出一种改进海鸥优化卷积神经网络的虚假数据注入攻击检测方法，所提方法利用具有共享权值和局部连接特性的卷积神经网络来对高维历史量测数据进行高效的特征提取及分类。然后引入具备平衡全局搜索和局部搜索能力的改进海鸥优化算法进行超参数寻优，以获得虚假数据检测的高度匹配网络结构，进而对不良数据进行检测和定位。最后通过对IEEE-14和IEEE-57节点系统进行大量攻击检测实验，验证了所提方法的有效性，并与其他多种检测方法对比，验证了所提方法的具有更优的分类性能、更高的准确率、精度、召回率和F1值。

False data injection attacks disrupt the stability of power systems by tampering with the data collected by data acquisition and monitoring control systems. Traditional methods for detecting false data injection attacks are unable to locate the attacked location or have low accuracy. Firstly, an improved method for detecting false data injection attacks using seagull optimized convolutional neural networks is proposed. The proposed method uses a convolutional neural network with shared weights and local connectivity to efficiently extract and classify features from high-dimensional historical measurement data. Secondly, an improved seagull optimization algorithm with balanced global and local search capabilities is introduced to perform hyperparametric optimization to obtain a highly matched network structure for false data detection. The network structure is then used to detect and locate bad data. Finally, the effectiveness of the proposed method is verified through extensive attack detection experiments on IEEE-14 and IEEE-57 node systems, and compared with various other detection methods to verify that the proposed method has better classification performance, higher accuracy, precision, recall, and F1 value.

目前，随着人工智能研究的进步，人工智能被大规模采用，数据监管等领域的需求也促使人们对隐私保护的认识和关注越来越多，这促进了联邦学习（federated learning，FL）框架的流行。但现有的FL难以应对异构问题以及用户的个性化需求。为了应对上述问题，研究了个性化联邦学习（personalized federated learning，PFL）的相关方法并提出了展望。列举了FL的框架并指出了FL的不足，在FL场景的基础上，引出PFL的研究动机对PFL中的统计异构、模型异构、通信异构、设备异构进行分析并提出可行性方案；将PFL中的客户端选择、知识蒸馏等个性化算法分类并分析各自的创新与不足。最后，对PFL的未来研究方向进行了展望。

Currently, with the advancement of artificial intelligence research, artificial intelligence is being widely adopted, and the increasing demand in areas such as data governance has led to growing awareness and concern for privacy protection, this has promoted the popularity of the federated learning (FL) framework. However, existing FL frameworks struggle to address heterogeneous issues and personalized user needs. In response to these challenges, methods of personalized federated learning (PFL) are studied and prospects are proposed. Firstly, the FL framework is outlined and its limitations are identified, leading to the research motivation for PFL based on FL scenarios. Subsequently, the analysis of statistical heterogeneity, model heterogeneity, communication heterogeneity, and device heterogeneity in PFL is conducted, and feasible solutions are proposed. Then, personalized algorithms in PFL such as client selection and knowledge distillation are categorized, and their innovations and shortcomings are analyzed. Finally, future research directions for PFL are discussed.

Identifying interesting relationships between pairs of variables in large data sets is increasingly important. Here, we present a measure of dependence for two-variable relationships: the maximal information coefficient (MIC). MIC captures a wide range of associations both functional and not, and for functional relationships provides a score that roughly equals the coefficient of determination (R(2)) of the data relative to the regression function. MIC belongs to a larger class of maximal information-based nonparametric exploration (MINE) statistics for identifying and classifying relationships. We apply MIC and MINE to data sets in global health, gene expression, major-league baseball, and the human gut microbiota and identify known and novel relationships.