基于Autoformer AI架构的SD-OTN专线流量预测模型  

SD-OTN Private Circuit Traffic Prediction Model Based on Autoformer AI Architecture

作　　者：骆益民[1] Luo Yimin(China Unicom Guangdong Branch,Guangzhou 510627,China)

机构地区：[1]中国联通广东分公司,广东广州510627

出　　处：《邮电设计技术》2023年第10期79-83,共5页Designing Techniques of Posts and Telecommunications

摘　　要：根据政企SD-OTN专线流量时间周期分布相似的特点,在业内首次采用流量序列时频域特征结合训练的Autoformer AI架构对流量序列进行预测,以获得更高预测准确性,并首次应用于某省联通的生产实践中,通过对流量趋势的精准预测,提前做好流量越限和丢包预警,有效降低SD-OTN的投诉量,可为SD-OTN客户提供付费的临时调速和周期性调速方案等增值业务功能。Based on the characteristics of similar time cycle distribution of traffic on government enterprise SD-OTN dedicated lines,it is the first time in the industry to use the time-frequency domain characteristics of traffic sequences combined with the trained Autoformer AI architecture to predict traffic sequences,in order to achieve higher prediction accuracy.It is also the first time in the production practice of China Unicom in a certain province to accurately predict traffic trends,provide early warning of traffic exceeding limits and packet loss,and effectively reduce the number of SD-OTN complaints,Which can provide value-added business functions such as paid temporary speed regulation and periodic speed regulation solutions for SD-OTN customers.

关 键 词：SD-OTN专线 流量预测 时频域特征训练 Autoformer AI算法 

分 类 号：TN913[电子电信—通信与信息系统]