学独您w. xliii ull.con
第35卷第10期
仪器仪表学报
Vol 35 No 10
2014年10月
Chinese Journal of Scientific Instrument
Oct,2014
基于双能
γ射线的煤质灰分软测量技术研究
程栋,温和,滕召胜,黎福海
(湖南大学电气与信息工程学院长沙410082)
摘要:为提高煤质灰分测量精度,提出了基于双能γ射线的煤质灰分智能软测量方法,该方法以Cs和Am作为中能和低
能的γ射线源,并以探测器检测到的Y计数作为辅助变量,利用混沌算法优化的函数链神经网络实现灰分软测量辨识建模,最
后对煤质灰分进行软测量预测和验证。研究结果表明:混沌算法优化的函数链神经网络预测方法的预测精度高,具有较强的泛
化能力;基于混沌算法优化函数链神经网络的灰分智能软测量值与实测值的平均误差为0.7%,最大误差为0.9%,煤质灰分测
量准确度高。
关键词:煤灰分;软测量;函数链神经网络;混沌优化;双能量γ射线
中图分类号:TH83文献标识码:A国家标准学科分类代码:480.30
Study on soft-sensing of coal ash content based on dual-energy y-ray
Cheng Dong, Wen He, Teng Zhaosheng, Ii Fr
ual
College of Electrical and Information Engineering, Hunan University, Changsha 410082, China)
Abstract In order to enhance the measurement accuracy of coal ash content, An intelligent soft-sensing method with dual-energy y-ray
based on functional link neural network is proposed. The proposed method takes Cs and Am as the sources of medium and low ener
gy Y-ra s, and the results of the y countering as auxiliary varables. The coal ash content is measured and verified after accomplishing
the modeling of soft-sensing applying the functional link neural network optimized with Chaos optimization algorithm. The study results
show that the functional link neural network forecasting method based on Chaos optimization algorithm has higher accuracy and stronger
generalization capability than other forecasting methods. The intelligent soft-sensing based on optimized functional link neural network
with the Chaos algorithm has good measurement accuracy, and the maximum error and average error between the soft-sensing value and
real value are 0. 9% and 0. 7%, respectively
Keywords coal ash content; soft-sensing; functional link neural network; chaos optimization; dual-energy Y-ray
速中子活化分析技术、中子感生瞬发射线技术等6。
上述
核技术法均根据所探测的γ射线强度与灰分之
间的1次线性经验方程,再通过参数标定过程实现煤
质灰分测定。此外,在煤质灰分检测方面已经有关于
灰分是煤炭成分、质量分析的主要技术指标之软测量技术的报道910,主要以电站锅炉系统的运行
常用煤炭充分燃烧后的剩余物(即灰分)来近似参数作为软测量的辅助变量,而利用DEγ进行智能
代表其中矿物质的含量?。测量煤质灰分的方法主软测量的相关研究文献甚少。
要包括燃烧法和核技术法等3。核技术法又可以分
国内外在软测量建模方面开展了较为深入的研究
为反散射与穿透方式,如低能γ射线反散射和穿透如回归分析、状态估计、模式识别、神经网络建模等。
法、双能Y射线(dual- energy y-rays,DEY)穿透法、快其中,神经网络由于其非线性逼近性能良好的特点,在解
收稿日期:201402 Received Date:2014-02
*基金项目:国家白然科学基金(61370014,5137049)资助项目
免费区 