多跳D2D组网下基于数据驱动的配电网在线异常检测

随着配电网数据信息的急剧增长，为了保证配电网供电可靠性，在配电网和基站间建立多跳D2D网络进行数据传输，提出多跳D2D组网下基于数据驱动的配电网在线异常检测方法。因配电网中每个时刻都会产生新的测量数据，提出一种基于一类支持向量机的配电网运行状态在线检测算法，该算法可根据每个时间周期智能电表上报的用电数据更新模型参数，实时推测配电网当前的运行状态。为了保证用电数据的正常传输，提出基于双边主成分分析的在线流量监测方法监督多跳D2D组网的流量状态。通过仿真实例验证，证明了提出的基于数据驱动的配电网在线异常检测算法可在提高检测速率和精确度的同时节约大量的计算时间和存储空间。

Data-driven online anomaly detection for power distribution networks under multi-hop D2D networking

With the rapid increase of data information generated in the power distribution network, we propose a data-driven online anomaly detection method for the power distribution network under multi-hop D2D networking to ensure the reliability of the power supply. Firstly, a multi-hop D2D network for data transmission is established between the power distribution networks and the base stations. Then, because new measurements will be generated in the power distribution networks in each period, an online detection algorithm for the power distribution network is designed based on one-class support vector machine. The algorithm can update model parameters based on the power data reported by smart meters in each time period and infer current working states of the power distribution network in real time. To ensure the normal transmission of power data, we propose an online traffic monitoring method based on the bilateral principal component analysis to monitor the traffic volume of the multi-hop D2D networking. Finally, the simulation results prove that the proposed data-driven online anomaly detection algorithm for power distribution network can improve the detection rate and accuracy while saving a lot of computing time and storage space.