燃烧优化论文：于神经网络和遗传算法的锅炉燃烧优化.docVIP

燃烧优化论文：基于神经网络和遗传算法的锅炉燃烧优化 【中文摘要】提高锅炉运行效率,降低烟气NO_X的排放是电站锅炉燃烧优化的主要目标,而燃烧特性模型是燃烧优化的核心。通过对BP神经网络的发展、特点、结构,以及神经网络模型的原理,结构和学习规则的分析,人工神经网络模型能够拟合任意非线性函数并具有良好的泛化能力,对复杂问题具有自适应和自学习能力,在非线性系统辨识方面得到了广泛地研究和应用。基于某电站锅炉燃烧系统的稳态试验数据,通过对锅炉系统模型结构的分析,应用人工神经网络建立了NO_x排放量和锅炉效率的预测模型,实现了其飞灰含碳量、排烟温度、炉膛温度、NO_x排放量等参数的软测量和锅炉效率的预测,为锅炉燃烧优化奠定了基础。遗传算法具有隐含并行性、全局解空间搜索和不受问题形式约束等特点,本文介绍了遗传算法的基本原理和操作,比较了二进制编码和实数编码的区别,对遗传算法的交叉、变异操作和初始种群的选取作了改进,以所建立的锅炉燃烧特性模型为基础,采用遗传算法,对锅炉燃烧运行工况进行优化。根据优化目标的不同,对锅炉稳态燃烧优化进行了效率约束下NO_x低排放的优化和NO_x排放约束下锅炉热效率的优化,为电站DCS基础控制层提供锅炉各操作参数的最佳设定值。算例表明,该优化方法可以实现锅炉高效、低污染排放运行。 【英文摘要】To improve the boiler efficiency and reduce NO_X emissions is the main goals of the power plant flue gas boiler combustion optimization. Combustion characteristics model is the core of combustion optimization. Through the analysis of the development, characteristics, structure of BP neural network and the the theory of the neural network model ,structure and the rules of learning, artificial neural network model can fit any nonlinear function and has good generalization ability and has the self-learning of the complex issues.It has been widely studied and applied in nonlinear system identification. Based on the steady-state test data of a power plant boiler combustion system and through the analysis of boiler system model structure,the thesis applys the artificial neural network method to build the prediction model of NO_x emissions and boiler efficiency.The model achieves the soft sensor of its carbon content in fly ash, flue gas temperature, furnace temperature, NO_x emissions and the prediction and boiler efficiency, and then lays the foundation for the boiler combustion optimization. Genetic algorithms has the characteristics of the implicit parallelism global search of solution space,and it don抰 has constraints of problems form. This article describes the basic principles and operations of genetic algorithms and compares the difference between the binary coding and real coding.It also improves the crossover