基于Elman神经网络的烟煤与生物质混烧灰熔点预测

Forecasting on ash fusion temperatures of bituminous coal and biomass co-firing based on Elman neural network

基于实验法测定烟煤与不同生物质混烧灰熔点时存在工作量大、时效性差的问题,通过分析烟煤分别与玉米秸秆和木屑混烧灰的成分,并以灰成分为输入量,建立基于Elman神经网络的灰熔点预测模型。采用Levenberg-Marquardt反向传播算法训练模型,利用残差检验与后验差检验法检验模型预测性能。研究结果表明:玉米秸秆灰、木屑灰分别含有较高的K_2O和CaO,与烟煤灰相比,2种生物质灰的碱性氧化物(Na_2O,K_2O,CaO,MgO)质量分数较高,酸性氧化物(Al_2O_3和SiO_2)质量分数较低;检验结果验证了该模型具有较高的预测精度和较强的泛化能力。

Considering that experimental determination of ash fusion temperatures of bituminous coal and different blended biomass is often onerous and time-consuming, an Elman neural network-based model to forecast ash fusion temperatures was presented by analyzing the ash compositions of bituminous coal blended corn straw or sawdust with ash ingredients as inputs. The model was trained by adopting LM back-propagation algorithm and its forecasting performance was assessed by employing the method of residual and posterior error test. The results show that corn stalk ash and sawdust ash are rich in K2O and CaO, respectively. The two biomass ashes contain more alkaline oxides（Na2O,K2O,CaO,MgO）, but less acidic oxides（Al2O3, SiO2） compared to bituminous coal ash. The test results validate that the model has high accuracy and strong generalization ability.