ANN模型对自热式固定床气化的动态过程预测

Prediction of the dynamic process of an autothermal fixed bed gasifier based on ANN model

在考虑气化炉顶部温度、中部温度、底部温度、空气流速、蒸气流速、反应时间以及生物炭添加量等多参数的情况下,建立自热式固定床气化过程合成气动态神经网络模型,以实现对气化过程中的气体浓度(H2,CH4,CO,CO2)动态过程预测.采用不同策略实现对网络参数选择与模型结构优化,同时通过嵌入适当数量的随机噪声来避免模型陷入局部最优.模型对气化过程中H2浓度变化曲线的预测决定系数(R^2)达到0.8646.

A dynamic neural network model was established based on the gasification process of an autothermal fixed bed gasifier to predict the real-time concentration of the syngas (H2,CH4,CO,CO2).Seven parameters were considered,including the temperature at the different height of the fixed bed gasifier (top,middle,and bottom),air flow rate,steam flow rate,reaction time and biochar addition. To avoid local optimum points when training by embedding random noise of appropriate size,various strategies were applied to optimize the network parameters and model structure.The prediction coefficient (R^2) of the hydrogen concentration curve in the gasification process was 0.8646.