﻿ 基于GA-BP的煤质发热量在线监测 On-Line Monitoring of Coal Calorific ValueBased on GA-BP

Advances in Energy and Power Engineering
Vol.04 No.04(2016), Article ID:18305,8 pages
10.12677/AEPE.2016.44014

On-Line Monitoring of Coal Calorific Value Based on GA-BP

Yifan Zhao, Zhongguang Fu

Key Laboratory of Condition Monitoring and Control for Power Plant Equipment, North China Electric Power University, Beijing

Received: Jul. 22nd, 2016; accepted: Aug. 14th, 2016; published: Aug. 17th, 2016

ABSTRACT

Due to a variety of factors, such as the origin, transport and price, the types of steam coal for power changed in a large range. Thus coal calorific value varied widely. It has a great impact on the safe operation of the boiler. In order to keep abreast of the unit’s operating conditions, this thesis applied reverse modeling ideas, combined with Genetic Algorithm optimizing BP neural network to simulate the relationship between the parameter values easily measured in the process of boiler operation and coal heat, and constructed an online coal calorific value monitoring model. The result of model validation with plant operation data showed that the accuracy met the engineering requirements, and the model was practical.

Keywords:Coal Calorific Value, Reverse Modeling, Neural Networks, The Mean Impact Value Method,

Online Monitoring

1. 引言

2. 反向建模

3. BP神经网络

BP神经网络作为一种机器学习方法，内嵌了很多数学思想和学习准则，基本思想是将输入数据经过多次迭代运算得到输出值。BP神经网络由输入层、隐含层和输出层三层组成。如果把输入值作为自变量，输出值作为因变量，那么我们可以把BP神经网络当成一个非线性函数 [8] 。

BP神经网络算法流程有三步：BP神经网络构建、BP神经网络训练和BP神经网络预测，如图1所示。

Figure 1. The algorithm flow

BP神经网络训练是用输入输出数据训练神经网络，通过训练使网络具有联想记忆和预测能力，从而训练后的网络能够预测模型输出。BP神经网络的训练过程包括以下几个步骤。

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

4. 遗传算法

5. 输入变量分析

(1)

(2)

(3)

Q1——锅炉的有效利用热；

Q2——排烟热损失；

Q3——可燃气体未完全燃烧热损失；

Q4——固体未完全燃烧热损失；

Q5——散热损失；

Q6——灰渣物理热损失。

(4)

——燃料物理显热；

——外来热源加热空气时所带入的热量；

——雾化燃油所用蒸汽带入的热量。

6. 煤质发热量在线监测模型建立

6.1. GA-BP神经网络构建及训练

6.2. MIV法筛选变量

6.3. GA-BP神经网络预测

Figure 2. The effects of input variables on calorific value

Figure 3. Forecast output of coal calorific value

7. 结论

Figure 4. The prediction error of BP network

On-Line Monitoring of Coal Calorific ValueBased on GA-BP[J]. 电力与能源进展, 2016, 04(04): 103-110. http://dx.doi.org/10.12677/AEPE.2016.44014

1. 1. 王舰. 煤质(种)特性对电站锅炉影响的试验研究[D]: [硕士学位论文].杭州: 浙江大学, 2002.

2. 2. 杨婷婷, 曾德良, 王玮, 等. 大型燃煤发电机组煤质实时校正方法[J]. 电站系统工程, 2010, 26(5): 7-10.

3. 3. 姚伟, 韩立芳, 薛宁, 等. 煤质变化对电站锅炉运行的影响[J]. 热力发电, 2005, 34(7):22-27.

4. 4. 胡宏伟. 煤种(质)变化对电站锅炉运行经济性影响的实验研究[D]: [硕士学位论文]. 杭州: 浙江大学, 2004.

5. 5. 景源. 电站锅炉入炉煤质发热量的反向建模研究[D]: [硕士学位论文]. 北京: 华北电力大学(北京), 2013.

6. 6. 靳涛. 火电机组反向建模方法的研究[D]: [博士学位论文]. 北京: 华北电力大学(北京), 2011.

7. 7. 张越. 600 MW超临界机组能耗混合建模及能耗敏感度研究[D]: [硕士学位论文]. 北京: 华北电力大学(北京), 2014.

8. 8. Matlab中文论坛. Matlab神经网络30个案例分析[M]. 北京: 北京航空航天大学出版社, 2010.

9. 9. Sisworahardjo, N.S., EI-Shark, M.Y. and Alam, M.S. (2008) Neural Network Controller for Micro Turbine Power Plants. Electric Power Systems Research, 78, 1378-1384. http://dx.doi.org/10.1016/j.epsr.2007.12.004

10. 10. Zhang, C.-F., Zhao, N., Li, L.-P., Wang, H.J. and Gao, A.-G. (2007) The Variable Load Operation Characteristic and Energy-Saving Optimizing Control System of Power Plant. 2007 International Conference on Machine Learning and Cybernetics, 1, 443-448. http://dx.doi.org/10.1109/ICMLC.2007.4370185

11. 11. Energy Information Administration (IEA) (2011) China Energy Data, Statistics and Analysis-Oil, Gas, Electricity. Washington DC.

12. 12. 樊泉桂, 阎维平, 闰顺林, 等. 锅炉原理[M]. 北京: 中国电力出版社, 2004.