摘要
在大数据环境下,现代企业生产数据的挖掘与利用对提升企业经济效益与提质增效尤为重要。目前企业工艺数据的利用与维护,多依赖专家经验与传统数据分析方法,针对这些企业数据处理与分析存在的局限性和共性问题,以大数据环境下火电厂锅炉系统运行状态预测为例,提出了一种基于Hadoop框架的卷积神经网络(Convolutional Neural Networks,CNN)-人工蜂群算法(Artificial Bee Colony,ABC)-双向门控循环网络(Bidirectional Gate Recurrent Unit,BiGRU)模型的主蒸汽流量预测方法。基于分布式预测模型对辽宁某电厂监测数据进行分析,结果表明,在预测波动较大的主蒸汽流量时,该方法在提升速度的同时,相较于传统BP、LSTM、GRU、BiGRU、CNN-BiGRU等模型,MAE值分别降低了61.188%、51.348%、46.342%、38.005%和20.560%,预测精度有所提高。
In the big data environment,the mining and utilization of modern enterprise production data is especially important to improve the economic efficiency and increase the quality and efficiency of enterprises.At present,the utilization and maintenance of process data mostly rely on expert experience and traditional data analysis methods.In view of the limitations and common problems of data processing and analysis in these enterprises,a prediction method based on Convolutional Neural Networks(CNN)-Artificial bee colony(ABC)-bidirectional gate recurrent unit(BiGRU)model with Hadoop framework is proposed for boiler system operation status prediction in thermal power plants under the big data environment as an example.Based on the distributed prediction model to analyze the monitoring data of a power plant in Liaoning,the results show that in predicting the fluctuating main steam flow,the method improves the speed while reducing the MAE value by 61.188%、51.348%、46.342%、38.005%and 20.560%,compared with the traditional BP,LSTM,BiGRU,CNN-BiGRU models,respectively.and the prediction accuracy has been improved.
作者
李萌
宗学军
连莲
何戡
杨忠君
LI Meng;ZONG Xuejun;LIAN Lian;HE Kan;YANG Zhongjun(College of Information Engineering,Shenyang University of Chemical Technology,Shenyang 110142,China)
出处
《电子设计工程》
2023年第22期11-15,20,共6页
Electronic Design Engineering
基金
辽宁省“兴辽英才计划”项目资助(XLYC2002085)。
