APP下载

基于 GSA-AGRU 的挤压机能耗预测

2021-08-20陈铭俊印四华

机电工程技术 2021年11期

陈铭俊 印四华

摘要:在铝型材的生产过程中,挤压机是核心的生产机器,其能耗占铝型材生产能耗的60%以上。针对当前挤压机能耗预测精度低和预测速度慢的问题,提出基于引力搜索优化的注意力机制门控循环单位网络模型(GSA-AGRU)用于预测挤压机的能耗,首先构建注意力机制的门控循环单位网络模型(AGRU),然后加入引力搜索算法(GSA)优化该网络的权重,最后得到最优的 GSA-AGRU 预测模型。利用某铝型材企业的挤压机生产能耗数据进行实验,结果表明 GSA-AGRU 模型相比于传统的 GRU、 LSTM、BP 和 AGRU模型具有更高的预测精度和更快的预测速度。

关键词:门控循环单位;能耗预测;挤压机;注意力机制;引力搜索算法

中图分类号:TG375文献标志码:A

文章编号:1009-9492(2021)11-0021-05

开放科学(资源服务)标识码(OSID):

Energy Consumption Prediction of Extruder Based on GSA-AGRU Chen Mingjun1,Yin Sihua2

(1. School of Computers, Guangdong University of Technology, Guangzhou 510006, China;

2. School of Electromechanical Engineering, Guangdong University of Technology, Guangzhou 510006, China)

Abstract: In the process of aluminum profile production, the extruder is the core production machine, its energy consumption accounts for more than 60% of the aluminum profile production energy consumption. In view of the current low extrusion machine energy consumption prediction precision and slow speed of prediction problem. The concentration mechanism gating cycle unit network model (GSA-AGRU) based on gravity search optimization was proposed to predict the energy consumption of extruder. Firstly, the attention mechanism gated cycle unit network model (AGRU) was constructed, and then the gravity search algorithm (GSA) was added to optimize the weight of the network. Finally, the optimal gsa-agru prediction model was obtained. The experiment was carried out by using the production energy consumption data of an aluminum profile enterprise, the results show that the GSA-AGRU model has higher prediction accuracy and faster prediction speed than the traditional GRU, LSTM, BP and AGRU models.

Key words: GRU; energy consumption; extruding machine; attention mechanism; GSA

0 引言

我國是铝型材生产、出口和消费大国。2017年中国挤压铝材产量攀升,达到了19500 kt/a ,占全球总产量的55%,拥有各种挤压力的现代化油压机约1850台,约占全球总台数的70%[1]。铝材生产与消费规模在不断扩大,对铝型材生产过程的进一步分析,已经成为促进铝材生产进一步发展的迫切需求。挤压机的能耗一直是铝型材生产企业高度关注的问题,而传统的物理能耗模型和仿真分析是常用的方法。张聪聪[2]通过挤压机生产理论计算得到连续挤压工艺参数与挤压模具结构参数的联系,设计出合理的模具以减少生产能耗成本。蒋攀[3]对挤压机的泵控液压系统进行改进与原系统相比能耗有所降低。

随着深度学习的发展,各种网络模型用于工业能耗预测中,Zhou B[4]使用长期短期记忆(LSTM)网络构建核心预测模型,LSTM区别于传统神经网络使用3个门和1个“记忆细胞”实现长距离信息传递。……

登录APP查看全文