联合模糊聚类和多样本SSO优化SVR的配电通信网络流量预测  

Prediction Method Based on FCM and Multi-Sample Social Spider Optimization SVR for Distribution Communication Network Traffic

作　　者：袁华 陈景召[2] YUAN Hua;CHEN Jingzhao(Jidong Oilfield Power Supply Company, Tangshan 063299, Hebei, China;Sias International College, Zhengzhou University, Zhengzhou 450000, Henan, China)

机构地区：[1]冀东油田供电公司,河北唐山063299 [2]郑州大学西亚斯国际学院,河南郑州450000

出　　处：《电网与清洁能源》2018年第5期36-43,共8页Power System and Clean Energy

基　　金：河南省科技厅科技攻关项目(172102210506);河南省科技厅科技攻关项目(162102210316);河南省科技厅基础与前沿技术研究计划项目(162300410269)~~

摘　　要：针对电力通信网络流量预测问题,提出了一种联合模糊聚类和多样本群居蜘蛛优化(social spider optimization,SSO)SVR的配电网络流量预测方案。针对配电网络流量时序非线性、周期性特点,首先采用自适应模糊聚类方法(adaptive fuzzy C-means clustering)对流量样本数据进行预处理,实现了样本数据聚类自动划分,有效降低了流量数据短相关性对预测精度的影响;然后利用多样本SSO优化算法(multi-sample social spider optimization algorithm)对SVR预测模型参数进行优化,通过引入多样本、网格迭代进化策略,从而得到不同流量数据聚类对应的最佳SVR参数组合;最后,运用多样本SSO优化SVR模型对预测数据进行预测分析。仿真结果表明,同ARIMA、神经网络等配电网路流量预测模型相比,提出的预测方案预测精度提高了18.8%～34.1%。Aiming at the prediction of distribution communication network traffic, a network traffic prediction algorithm based on fuzzy C-means clustering and multi-sample SSO optimization SVR is proposed. First, according to the nonlinear and periodic characteristics of network traffic time series, the adaptive fuzzy C-means clustering method is used to process the training traffic sample set, which can automatically divide the traffic sample data into different clusters and effectively reduce the influence of sample data difference for prediction performance. Second, the multi-sample social spider optimization algorithm is used to optimize the SVR prediction model parameters,through the introduction of multiple samples and mesh iterative evolution strategy, the best SVR parameters combination is obtained for different traffic data clusters. Finally, the improved SVR model is used to predict the test data. The simulation results show that, compared with ARIMA and the neural network prediction model, the prediction accuracy of the proposed method is improved by 18.8% ~ 34.1%.

关 键 词：配电通信网络 时间序列 模糊C均值聚类 群居蜘蛛优化 支持向量回归(SVR) 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]