基于智能算法的分解炉掺烧污泥脱硝经济分析

Economics analysis of denitrification with co-combustion of coal and sludge in precalciner based on intelligent algorithm

为研究水泥分解炉在掺烧污泥和选择性非催化还原(SNCR)脱硝共同作用下,以脱硝系统运行成本最小为目标,对相关主要参数进行优化控制.以广东某6000 t/d水泥分解炉为研究对象,以污泥掺烧量、喷氨量、烟风温度、压差等运行参数为输入变量,采用遗传算法优化的BP神经网络(GA-BPNN)建立出口烟气NOx浓度的预测模型.再采用多种群遗传算法(MPGA),在满足NOx排放浓度达标的前提下,以脱硝系统成本最小为优化目标,得到经济性最佳的水泥分解炉运行参数.结果表明:所建立的烟气NOx浓度预测模型相对误差仅为2.63%,相关系数高达0.976;MPGA较GA和PSO具有更强的全局搜索能力和进化能力,更适合运用在高维度非线性解空间中.所选对象增加污泥掺烧量至18 t/h,同时将喷氨量减少至500 L/h,可获得最低的脱硝运行成本.

The optimal control of the main parameters was studied to minimize the operation cost of denitrification system,when the co-combustion of sludge and coal and selective non-catalytic reduction(SNCR)denitrification worked together in the cement precalciner.A 6000 t/d cement calciner with combustion chamber in Guangdong Province was taken as the research object.The BP neural network with genetic algorithm optimization(GA-BPNN)was used to establish the prediction model of NOx concentration in outlet flue gas,with input variables of sludge burning amount,ammonia injection amount,furnace flue gas temperature,pressure difference and other operating parameters.Under the premise of meeting the NOxemission concentration standard,the minimum cost of denitration system was set as the optimization objective,and multigroup parallel genetic algorithm(MPGA)was used to obtain the optimal operating parameters of cement precalciner.Results show that the relative error of the established NOx concentration prediction model is only 2.63%,and the correlation coefficient is 0.976.Compared with GA and PSO,MPGA has stronger global search ability and evolutionary ability,and is more suitable for high dimensional nonlinear solution space.The cement plant can increase the sludge burning amount to 18 t/h and reduce the ammonia injection amount to 500 L/h to obtain the lowest denitration operation cost.