基于嵌入分层决策树的电网完整性保护

为了提升系统抗攻击能力以及泛化能力,提出了一种基于支持向量机嵌入分层决策树的多智能体电网抗攻击系统完整性保护方案。首先利用多智能体来增强系统完整性保护的网络弹性,重点是系统的态势感知和自适应能力。进一步将分布式系统完整性保护设置中的数据驱动异常检测问题转化为多分类问题。然后提出了一种支持向量机嵌入分层决策树的有监督学习算法和一种自适应的负荷卸载策略,以提升检测能力,减少时间成本。实验结果表明,所提出的系统完整性保护方案能够检测出电网的异常运行状态,并能够调整其补救措施,以适应各种网络攻击。

Integrity Protection of Multiple Agent Power Grid Anti Attack System Based on SVMLDT

In order to improve the anti attack ability and generalization ability of the system, an integrity protection scheme of multiple agent power grid anti attack system based on support vector machine and hierarchical decision tree was proposed. Firstly, multiple agent was used to enhance the network flexibility of system integrity protection, focusing on the situation awareness and adaptive ability of the system. Furthermore, the problem of data-driven anomaly detection in distributed system integrity protection settings was transformed into a multiple classification problem. Then, a supervised learning algorithm based on SVM embedded in hierarchical decision tree and an adaptive load shedding strategy were proposed to improve the detection ability and reduce the time cost. The experimental results show that the proposed system integrity protection scheme can detect the abnormal operation state of power grid, and adjust its remedial measures to adapt to various network attacks.