基于DAK-FNN的出水氨氮软测量建模方法  

Soft measurement based on DAK-FNN for effluent ammonia nitrogen

在线阅读下载全文

作  者:张伟[1] 张春辉 ZHANG Wei;ZHANG Chunhui(School of Electrical Engineering and Automation,Henan Polytechnic University,Jiaozuo 454000,Henan,China)

机构地区:[1]河南理工大学电气工程与自动化学院,河南焦作454000

出  处:《河南理工大学学报(自然科学版)》2023年第2期134-143,159,共11页Journal of Henan Polytechnic University(Natural Science)

基  金:国家自然科学基金资助项目(61703145);河南省科技攻关项目(222102210213);河南省高校科技创新团队项目(20IRTSTHN019)。

摘  要:针对因出水氨氮实时数据不完备建立的模型性能较差问题,在迁移学习框架下,提出一种基于数据与知识驱动的模糊神经网络(data-knowledge-driven fuzzy neural network,DAK-FNN)出水氨氮软测量建模方法。首先,为有效结合源场景已有知识和当前场景数据,提出一种基于迁移学习思想的知识传递方法,将从大量历史数据获取的参考模型蕴含的知识传递到目标模型,并利用在线学习的方式微调目标模型参数,提高模型精度。此外,提出一种基于长短时记忆机制的结构调整方法,将目标模型神经元分为核心神经元与非核心神经元,通过设定不同的神经元增删阈值调整结构,提高模型泛化性能。出水氨氮预测实验结果表明,与其他方法相比,所提方法具有更高的在线预测精度和更好的实时性。The performance of the established model with incomplete real-time data is poor.To solve this problem,a soft measurement method based on data-knowledge-driven fuzzy neural network(DAK-FNN)was proposed for effluent ammonia nitrogen under the framework of transfer learning,which not only could use data of the current scene,but could make full use of knowledge of the source scene.First,in order to ef⁃fectively combine knowledge of the source scene with data of the current scene,a knowledge transfer method was proposed based on the idea of transfer learning.It could obtain knowledge contained from a large amount of historical data in the reference model,and could transfer knowledge to the target model.The parameters of the target model were fine-tuned through online learning method to improve the model accu⁃racy.Second,in order to improve the generalization performance of the model,a structure adjustment method was proposed based on long and short-term memory mechanism,which could divide the neurons of target model into core neurons and non-core neurons.The structure could be adjusted by setting different neuron addition and deletion thresholds.Effluent ammonia nitrogen prediction experiment were carried out,the re⁃sults showed that the proposed method had higher online prediction accuracy and better real-time perfor⁃mance compared with other methods.

关 键 词:模糊神经网络 数据驱动 知识驱动 微调 出水氨氮 软测量 

分 类 号:TP301[自动化与计算机技术—计算机系统结构]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象