摘要
为进一步分析设备能耗异常问题,以及面向海量的能耗数据,在提取能耗特征数据基础上,提出改进注意力机制结合Bi-LSTM的能耗异常分类模型,然后将分类模型部署到Spark并行框架中,以提高其海量数据的能力。结果表明,在引入能耗特征的分类模型上,其准确率为95.11%,高于只以原始数据作为数据的分类模型;引入注意力机制的Bi-LSTM对能耗的分类准确率明显高于Bi-LSTM,准确率达97.76%。同时通过Spark并行框架运行,可实时监测能耗异常问题。由此通过以上构建,得出本研究构建的分析模型及平台可行,可在企业设备能耗监测中应用。
In order to further analyze the problem of equipment energy consumption anomalies and face massive energy consump-tion data,based on the extraction of energy consumption characteristic data,the improved attention mechanism combined with Bi-LSTM energy consumption anomaly classification model is proposed,and then the classification model is deployed in the Spark paral-lel framework to improve its capacity of massive data.The results show that the accuracy of the classification model with energy con-sumption characteristics is 95.11%,which is higher than that of the classification model with only original data as data;The classifi-cation accuracy of energy consumption of Bi-LSTM with attention mechanism is significantly higher than that of Bi-LSTM,with an ac-curacy of 97.76%.At the same time,the Spark parallel framework can be used to monitor abnormal energy consumption in real time.Based on the above construction,it is concluded that the analysis model and platform built in this study are feasible and can be applied in the monitoring of enterprise equipment energy consumption.
作者
张俊丽
ZHANG Juni(Xi’an Eurasia University,Xi’an 710065,China)
出处
《自动化与仪器仪表》
2023年第6期31-34,39,共5页
Automation & Instrumentation
基金
陕西省教育科学“十三五”规划项目《基于OBE理念的数据科学与大数据人才培养创新实践研究》(SGH20Y1480)
西安欧亚学院教育教学改革研究项目《基于产教融合的数据科学与大数据技术专业人才培养研究与实践》(2021ZD002)。
