Research On Modeling And Optimizational Operation Of Combustion System For Pulverized-coal Boiler Based On Artificial Intelligence Technology

The fuel cost takes great proportion in the cost of thermal powergeneration. In domestic situation, because of the problem of manage, coalquality and so forth, the coal consumption of boilers is high, resulting inthe increased productive cost. And, it is important means to improve thecompetitive ability of power plant by enhancing the running level ofboilers, saving energy and reducing consumption. This paper aims at theproblem that the timely adjustment basis is lacked for operators of boilerduring combustion adjustment. The main works are as follows:(1)Based on experimental data, the artificial neural networkstechnique was used to correlate many operation parameters(boiler load,air distribution mode, coal characteristics and so forth) with temperatureof flue gas, carbon content in fly ash and clinker, and the hybrid model topredict the combustion efficiency of boiler has been established utilizingcounter-balance calculating method of boiler efficiency.(2)In the modeling, BP neural network has been adopted to comparethe following two methods for advancing the generalization performanceof neural network: normalization mean-square error function and Baggingneural network ensemble. The result of experiment reveals that themaximum relative errors of neural network ensemble constituted by 7individual BP neural network are individually 3.92%, 4.8% and 4.93% foraspects of temperature of flue gas, unburned coal in fly ash and carboncontent of the clinker in the validation collection. And it has bettergeneralization performance than the neural network using normalizationmean-square error function.(3) On the foundation of predictive model of boiler efficiency,adaptive real-coded genetic algorithm was adopted to optimizeconsidering the disadvantage of simple genetic algorithm. Theoptimization result indicated that the average efficiency enhanced by 1%approximately on work conditions. Moreover, each optimized value ofoperation parameters value tally well with the experiment data analysis.The optimizational operation of boiler combustion system was studied by artificial intelligence technology. Generalization performanceof modeling was emphasesed, neural network ensemble technology wasutilized, achieving excellent effect.