摘要
在输送压力可达3.7 MPa,固气比可达660 kg.m-3的气力输送实验台上进行系统的研究,考察输送压力、输送差压、流化风量、充压风量、补充风量、煤粉含水率等条件对固气比的影响.采用改进学习算法的BP神经网络,对固气比进行有效预测.结果表明,固气比随着输送差压的增大而增大;随着流化风量的增大先增大后减小;注入风量一定时,充压风量增大固气比先减小后增大;固气比随着补充风量的增大而显著减小;随着煤粉含水率的增大而减小;随着表观气速的增大而减小.获得了系统最佳的操作条件.建立的BP网络,最大训练误差为2.7%,最大预测误差为5.8%,具有很好训练结果和预测能力.
Experiments were carried out on an experimental setup at the pressure up to 3.7 MPa and the solid-gas ratio up to 660 kg · m^-3. The influences of such factors as conveying pressure, conveying differential pressure, flow rates of fluidizing gas, pressurized gas and supplying gas, and moisture in coal on the solid/gas ratio were investigated. BP neural network was established, which successfully predicted the solid/gas ratio. The results indicated that solid-gas ratio increased with the increase in differential pressure; first increased and then fell down with the increase in fluidizing gas flow rate; first decreased and then grew with the increase in pressurizing gas flow rate when injecting gas flow rate kept instant; decreased dramatically with the increase in supplying gas flow rate; decreased with the increase in the moisture content in coal; decreased with the increase in superficial gas velocity. The optimum operating condition of the system is obtained. BP network constructed showed good training results and predicting ability with the maximum training error of 2.7% and the maximum predicting error of 5.8%.
出处
《燃烧科学与技术》
EI
CAS
CSCD
北大核心
2008年第5期423-428,共6页
Journal of Combustion Science and Technology
基金
国家重点基础研究发展计划(973)资助项目(2004CB217702)
关键词
气力输送
高压
超浓相
固气比
神经网络
pneumatic conveying
high-pressure
dense-phase
solid/gas ratio
neural network
