Coal blending has now attracted much attention in coal industry of China, and has been investigated extensively to meet the often conflicting goals of environmental requirements and reliable and efficient boiler operation in power plants. However, most of the existing blending projects are guided by experience, or linear-programming (LP), whose main assumption is that all the quality parameters of a blend can be approximated as the weighted average of the corresponding indexes of its component coals at any condition. This has been proved incorrect for some blend properties. Now, more and more evidence indicates that a strong non-linearity exists between some quality parameters of a coal blend and those of its component coals. Thus the unreliable assumption impairs the resulting coal-blending scheme. To remedy this situation, a novel coal blending technology for power plants, i.e. using nonlinear programming (NLP) based on neural network models, was proposed, and has now been successfully applied at the Hangzhou Coal Blending Center. The application attests that this new technology is much better than the existing linear-programming coal-blending method.
[1]
R. Sakurovs,et al.
The prediction of the fusibility of coal blends
,
1994
.
[2]
Nick J. Miles,et al.
Determination of the handleability of a coal blend using an extrusion trough
,
1997
.
[3]
Kefa Cen,et al.
Predicting coal ash fusion temperature with a back-propagation neural network model
,
1998
.
[4]
Yao Qiang,et al.
Nonlinear-programming mathematical modeling of coal blending for power plant
,
1997
.
[5]
Klaus R. G. Hein,et al.
Effect of coal blending and particle size on NOx emission and burnout
,
1994
.
[6]
Dilip Sarkar,et al.
Methods to speed up error back-propagation learning algorithm
,
1995,
CSUR.
[7]
Alan W. Scaroni,et al.
T.g.a. and drop-tube reactor studies of the combustion of coal blends
,
1993
.