【摘要】 燃煤电站脱硝系统在变负荷工况下具有非线性、大滞后的特性,传统的控制方式很难保证喷氨量的精确控制。随着燃煤发电厂超低排放标准的实施,有必要对脱硝系统进行运行优化。通过挖掘海量脱硝系统的历史运行数据,提出一种基于混合群智能算法优化的核极限学习机NO_X排放动态预测模型。首先,对选择性催化还原(SCR)脱硝反应系统进行理论分析和实际运行研究,研究了采用核函数代替极限学习机中隐含层节点的显式映射的方法,从而无需事先给定隐含层节点数。然后,采用混合蚁群和粒子群优化的混合智能算法,对核极限学习机的学习参数进行优化。最后,以某电站锅炉脱硝系统为例,利用提出的方法进行验证,得到较高的建模精度。该研究为下一步脱硝系统控制优化打下良好基础。更多还原

【Abstract】 The denitrification system of coal-fired power station has nonlinear and large hysteresis characteristics under variable load conditions; traditional control methods are difficult to guarantee precise control of ammonia injection.With the implementation of ultra-low emission standards for coal-fired power plants,it is necessary to optimize the operation of denitrification system.By mining the historical operation data of mass denitrification system,a dynamic prediction kernel limit learning machine(KELM) NO_X emission model based on hybrid swarm intelligence algorithm optimization is proposed.Firstly,the theoretical analysis and practical operation of SCR denitrification reaction system are studied.The explicit mapping method of using kernel function instead of hidden layer nodes in limit learning machine is studied,so that the number of hidden layer nodes need not be given beforehand.Then,the parameters of kernel limit learning machine are optimized by hybrid ant colony optimization and particle swarm optimization.Finally,the parameters of kernel limit learning machine are optimized by using hybrid intelligent algorithm of ant colony optimization and particle swarm optimization.With a certain power plant boiler denitrification system as example,using the proposed method to verify,the higher accuracy of modelling is obtained,which lays a good foundation for optimization of denitrification system control in future.更多还原