基于风电场SCADA系统的云平台设计  

Design of cloud platform based on wind farm SCADA system

作　　者：梁涛 李燕超 许琰 杨改文 LIANG Tao;LI Yanchao;XU Yan;YANG Gaiwen(School of Artificial Intelligence,Hebei University of Technology,Tianjin 300130,China)

机构地区：[1]河北工业大学人工智能与数据科学学院

出　　处：《中国测试》2019年第10期114-119,共6页China Measurement & Test

基　　金：河北省科技计划项目(16214510D)

摘　　要：为充分利用风电场集控中心采集爆炸式增长的海量数据,解决风电场SCADA系统数据存储和处理能力不足的问题,设计基于SCADA系统的云计算平台,实现数据的高效分布式存储以及处理。传统LS-SVM适用于小样本数据,对风电场海量数据的处理比较乏力,针对此问题,采用MapReduce云计算平台实现算法并行化,并用遗传算法对其进行参数寻优。最后利用某实际风电场SCADA系统的历史数据,实现短期风功率预测功能。在数据不断增加的情况下,采用MapReduce云计算平台实现LS-SVM算法并行化训练比单机LS-SVM训练缩短的时间十分显著,可增强SCADA系统的实时性;当数据增加到1.2 GB时,时间缩短一半,而准确度相差0.1%。In order to make full use of the wind farm centralized control center to collect exploding massive data,solve the problem of insufficient data storage and processing capacity of wind farm SCADA system,a cloud computing platform based on SCADA system is designed to realize efficient distributed storage and processing of data.The traditional LS-SVM is suitable for small sample data,and the processing of massive data of wind farm is relatively weak.To solve this problem,the algorithm is parallelized by MapReduce cloud computing platform,and the parameters are optimized by genetic algorithm.Finally,using the historical data of a real wind farm SCADA system,the short-term wind power prediction function is realized.In the case of increasing data,The LS-SVM algorithm parallelization training using the MapReduce cloud computing platform is significantly shorter than the stand-alone LS-SVM training.the real-time performance of the SCADA system is enhanced.When the data is increased to 1.2 GB,the time is reduced by half,and correct rate differs by 0.1%.

关 键 词：SCADA 云计算 LS-SVM 并行 风功率预测 遗传算法 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]