新型神经网络算法研究及其在锅炉燃烧优化中的应用

As a large consumption of the coal resource, power plants are faced with multiple demands of generating additional profit and reducing the pollutant emissions. So it is an urgent problem to be solved for the combustion of the high efficiency and low pollution of coal-fired boilers. However, due to the extremely complex physical and chemical changes in boilers’ combustion process, it is very difficult to set up models of combustion process and to achieve the combustion optimization objective. Therefore, in this project, a great deal of research will be done on the learning algorithm of neural networks and how to apply it to make a boiler combust at high efficiency and low pollution. This project will propose a novel incremental learning algorithm of neural networks; set up a dynamic prediction model of a boiler combustion system and design a state estimation method; present an operation strategy for achieving combustion optimization; design and develop new boiler combustion optimization system software. This project not only has theoretical significance, but also has great significance to the clean utilization of coal in the development of science and technology, and simultaneously, it has wide application prospect and great social benefits.

火电厂是煤炭消耗的大户，其面临着降低运营成本和减少污染物排放量等多重标准，故电站锅炉的高效低污染燃烧是亟待解决的问题。然而由于电站锅炉燃烧过程是一个极其复杂的物理化学变化过程，传统方法很难描述出燃烧特性模型，不易实现锅炉燃烧优化。因此，本项目首先对人工智能领域内的神经网络算法进行深入的研究，然后针对其如何应用于锅炉燃烧过程实现锅炉的高效低污染燃烧进行深入的研究。本项目预计提出一种新型增量式神经网络学习算法、建立锅炉燃烧系统的动态预测模型并设计状态估计方法、给出锅炉实现燃烧优化的运行策略、设计开发一款新型智能锅炉燃烧优化系统软件。该项目的研究不仅具有理论意义，而且对煤炭洁净利用科技的发展具有重大的意义，同时拥有广泛的应用前景与巨大的社会效益。

该项目为了提高锅炉热效率降低污染物排放量，提出了一种基于人工神经网络和启发式优化算法的锅炉燃烧优化策略。该项目实现的主要研究目标：1)提出了两种新型离线神经网络，分别命名为双线性快速学习网和深度双向学习机，并将其应用于锅炉热效率和NOx排放量建模上；2) 提出了一种增量式在线最小二乘快速学习网，并在其基础上建立了锅炉燃烧系统的动态特性模型；3) 在上述模型的基础上，给出了一种基于人工蜂群优化算法的优化策略；4) 实验表明该项目能很好的达到锅炉燃烧优化的目的。在本项目资助下，项目组在国内外重要学术刊物、会议上发表论文27篇论文（含录用），其中13篇SCI期刊论文，9篇国内核心期刊，3篇国际会议论文；并申请发明专利1项，获河北自然科学技术三等奖1项、香江学者奖1项。该项目完成并远超预期目标。