基于智能算法的燃煤电站锅炉NO_x排放模型及优化研究

发布时间：2018-04-02 02:12

本文选题：燃煤锅炉　切入点：神经网络　出处：《南昌大学》2015年硕士论文

【摘要】：随着经济的发展,环境污染愈加严重,目前燃煤发电机组作为我国的主要发电设备的状况不会短期内改变,而2011年版《火电厂大气污染排放标准》的出台,对燃煤电站的高效低污染运行提出了更加严苛的要求。近年来我国燃煤机组大都加装的脱硝设备,使NOx排放值低于国家标准,这样必然影响燃煤电站的经济性,因此锅炉的高效低污染燃烧优化运行的研究具有重要意义。对某660MW燃煤电站锅炉进行混煤掺烧热态试验,选取该电厂经常运行的三个工况,按照实验室得出的配煤方案进行效率及污染排放物试验。试验结果表明,在660MW负荷下,采用方案二的上煤方式,锅炉在变氧量测试中氧量为2.5%时锅炉的热效率最高,为93.57%;600MW负荷下,采用方案四的上煤方式,锅炉在变氧量测试中氧量为2.0%时锅炉的热效率最高,为93.68%;550MW负荷下采用方案六的上煤方式,锅炉在变氧量测试中氧量为2.5%时锅炉的热效率最高,为93.60%。锅炉NOx的排放浓度随着氧量的增加而升高,因此在运行过程中应保证氧量稳定在最佳氧量值,使锅炉高效低污染运行。在热态试验的基础上应用BP神经网络建立锅炉排放特性模型,得到了较好的结果。网络能够很好得映射输入与输出之间的关系,NOx排放模型的平均相对误差为0.73%,其中最大相对误差出现在样本9处,最大相对误差为4.6%。三个测试样本的相对误差分别为0.46%,0.59%和2.34%,平均相对误差为1.13%。锅炉效率的网络的平均误差为0.13%,平均相对误差为0.4%。结合遗传算法对所建立的网络模型进行优化,优化后的模型精确度和泛化能力有所提高。优化后的网络平均误差为0.18%,较优化前的0.73%大大降低,校验样本的相对误差分别为0.39%、0.51%、0.8%,平均误差为0.57%。优化结果表明,遗传算法对BP网络训练的初始权值的优化是有效的,可以提高网络的精确度和泛化能力。在网络建立的基础上,对锅炉的NOx排放进行优化,优化前的习惯工况下NOx浓度为458.4mg/m3,采用优化后的运行方式NOx浓度降低为329.7mg/m3,降低了28%,效果明显。优化后的操作运行方式能够体现锅炉燃烧的燃料分级与配风分级,从而抑制NOx的生成。
[Abstract]:With the development of economy, environmental pollution is becoming more and more serious. The status of coal-fired generating units as the main power generation equipment in China will not change in the short term. However, the 2011 edition of the Standard of Atmospheric pollution emissions from Thermal Power plants has been issued. More stringent requirements have been put forward for the operation of coal-fired power stations with high efficiency and low pollution. In recent years, most coal-fired units in China have installed denitrification equipment, which makes the NOx emission value lower than the national standard, which will inevitably affect the economy of coal-fired power stations. Therefore, it is of great significance to study the optimal operation of boiler combustion with high efficiency and low pollution. For a 660MW coal-fired power plant boiler, the heat state of mixed coal combustion is tested, and the three operating conditions of the power plant are selected. According to the coal blending scheme obtained in the laboratory, the efficiency and pollution emissions are tested. The results show that under the 660MW load, the boiler has the highest thermal efficiency when the oxygen quantity of the boiler is 2.5 in the test of the variable oxygen quantity, and the coal feeding mode of the second scheme is adopted. Under the load of 93.57MW and 600MW, the boiler has the highest thermal efficiency when the oxygen quantity of the boiler is 2.0 in the test of the variable oxygen quantity, and the coal feeding mode of scheme six is adopted under the load of 93.68MW / 550MW. The boiler has the highest thermal efficiency of 93.60 when the oxygen quantity is 2.5 in the variable oxygen quantity test. The NOx emission concentration of the boiler increases with the increase of oxygen quantity, so the oxygen quantity should be stabilized at the optimum oxygen value during the operation. The boiler is operated efficiently and low pollution. On the basis of thermal test, BP neural network is used to establish the boiler emission characteristic model. The network can well map the relation between input and output. The average relative error of NOx emission model is 0.73, and the maximum relative error appears in 9 samples. The maximum relative error is 4.6. The relative error of the three test samples is 0.46% and 2.34%, the average relative error is 1.130.The average error of boiler efficiency network is 0.13 and the average relative error is 0.4. The accuracy and generalization ability of the optimized model are improved. The average error of the optimized network is 0.18, which is much lower than the 0.73% before the optimization. The relative error of the calibration sample is 0.390.51 and the average error is 0.57. Genetic algorithm is effective to optimize the initial weights of BP network training, which can improve the accuracy and generalization ability of the network. Based on the establishment of the network, the NOx emission of boiler is optimized. The concentration of NOx is 458.4 mg / m ~ (3) under normal working conditions before optimization, and the concentration of NOx is reduced to 329.7 mg / m ~ (3) by using the optimized operation mode, which reduces 28mg / m ~ (3), and the effect is obvious. The optimized operation mode can reflect the fuel classification and air distribution classification of boiler combustion. Thus, the formation of NOx was inhibited.
【学位授予单位】：南昌大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：X773;TP18

【参考文献】