Boiler combustion optimization is an important means to improve boiler efficiency and reduce NOx emission and to achieve energy conservation and emission reduction in power plant. Nowadays,most combustion optimization methods are based on steady- state model of the boiler combustion system; it is difficult to achieve dynamic optimization under variable load conditions. To solve the problem,an improved online adaptive least squares support vector machine dynamic modeling algorithm is proposed. Firstly,off line support vector screening is conducted to reduce the sample number and ensure the sparseness of support vector. Then,three of the online update strategies of support vector,i. e. ,replacement,addition and delete are used to make the algorithm to well adapt the variation of characteristics of the object. With the algorithm above,the dynamic model of the boiler combustion system is established for certain 600 MW power unit. The results of simulation show that the proposed model can accurately reflect the dynamic characteristics of the boiler efficiency and NOx emission with the load change. Compared with the steady state model established by traditional online adaptive least squares support vector machine algorithm,higher accuracy and predictive capability are obtained. In addition,the structure of model is simplified,and less online calculation is needed,which is the basis for further research on the dynamic optimization control strategy of boiler combustion.%通过燃烧优化提高电站锅炉效率并降低NOx排放,是实现电厂节能减排的重要手段.目前大多数的燃烧优化方法都是基于锅炉燃烧系统的稳态模型,因而难以实现动态变负荷情况下的燃烧优化.针对该问题,提出了一种改进的在线自适应最小二乘支持向量机动态建模算法.该算法首先进行离线的支持向量筛选,不仅减少了建模所需样本数,也确保了支持向量的稀疏性;然后,采用替换、新增、删除3种支持向量的在线更新策略,使算法能够更好地适应对象特性的变化.将上述算法应用于建立某600 MW机组锅炉燃烧系统的动态模型.仿真结果表明,所建模型能够准确反映锅炉效率和NOx排放随负荷变化的动态特性.相比原有基于在线最小二乘支持向量机建立的稳态模型,其具有更高的精度和预测能力.同时,该模型结构简单、在线计算量小,为进一步研究锅炉燃烧动态优化控制策略奠定了基础.
展开▼
