Research On Pyrolysis Characteristics And Prediction Method Of Litsea Cubeba Kernel Meal Based On Neural Network

In this paper,the pyrolysis characteristics of Litsea cubeba kernel meal and the prediction model of the calorific value of the target product bio oil were studied with the aim of obtaining high calorific value bio oil,combined with the artificial neural network method,so as to provide a basis and a new method for the process design of Litsea cubeba kernel meal pyrolysis to produce bio oil.In order to master the weight loss characteristics of Litsea cubeba kernel meal,the thermogravimetric experiments of Litsea cubeba kernel meal in N2 atmosphere were carried out,the effects of heating rate and catalyst addition ratio on the weight loss process were investigated,and the apparent reaction activation energy of the weight loss process was calculated by equal conversion method(fwo and KAS).The results showed that: 1)at different heating rates,the weight loss law of Litsea cubeba kernel meal was similar,which was divided into water analysis stage(50 ℃ ～160 ℃),volatile analysis stage(160 ℃ ～570 ℃)and slow carbonization stage(570 ℃ ～800 ℃).2)With the increase of heating rate,TG and DTG curves shifted to the right,and the maximum weight loss rate was 72.5%.3)The addition of HZSM-5(20%-50%)reduces the initial pyrolysis temperature by about 13 ℃,which indicates that HZSM-5 can effectively reduce the activation energy required for the reaction.4)The apparent reaction activation energies obtained by fwo method and KAS method are 229.47 kj/mol and 226.08 kj/mol respectively,and the correlation coefficient is more than 0.98,which indicates that the assumption of equal conversion is reasonable.This part of the conclusion provides a reference for the subsequent rapid pyrolysis of Litsea cubeba kernel meal in terms of pyrolysis temperature range and catalyst addition ratio.In order to explore the pyrolysis conditions required for the preparation of high calorific value bio oil and provide samples for the artificial neural network in Chapter 4,this paper then carried out a rapid pyrolysis experiment of Litsea cubeba kernel meal,studied the effect of pyrolysis temperature(400 ℃ ～700 ℃)on the yield of gas,liquid and solid products,and investigated the effect of zeolite based catalyst(HZSM-5)on the reduction of oxygenates in bio oil,The optimal pyrolysis conditions were also investigated by response surface methodology.The results showed that: 1)the highest yield of bio oil was56.4% at 500 ℃.The yield of biochar decreased gradually with the increase of temperature,reaching a minimum of 27.8% at 600 ℃,and the volume proportion of pyrolysis gas increased gradually with the increase of temperature,reaching a maximum of 22.3%.This indicates that high temperature will promote the precipitation of volatile matter and conduct secondary pyrolysis.2)The proportion of oxygenates in pyrolysis oil is high.At 500 ℃,there are more long-chain alkanes in bio oil,among which cetane,octadecane and eicosane account for 1.74%,1.78% and1.39% respectively.CO2 in the pyrolysis gas decreases with the increase of temperature,and the content of CO increases with the increase of temperature.High temperature is conducive to the endothermic reaction,which promotes the reaction between CO2 and C to produce more Co.HZSM-5 increased the aromatic hydrocarbons in bio oil during thermal cracking from 1.91% to 16.76%,and reduced the oxide content from93.25% to 53.69%,indicating that HZSM-5 can promote the conversion of oxygenates to aromatic hydrocarbons,and then gradually increase the calorific value from 22.31 mj/kg to 29.56 mj/kg.3)The response surface methodology was used to optimize the pyrolysis conditions of Litsea cubeba kernel meal.The optimum conditions were N2 182.76ml/min,HZSM-5 36.21% and temperature 563.83 ℃.The predicted calorific value of bio oil under these conditions was 30.16mj/kg.Under the optimal process,the numerical value is verified,which shows that the predicted value of response surface method is close to the actual value,indicating that the model has good application value.In order to develop a faster and more accurate prediction method for the process design of Litsea cubeba kernel meal pyrolysis to produce bio oil,based on the above research,the prediction model is studied by using BP neural network in MATLAB.Three algorithms(Levenberg Marquardt,Bayesian regularization and scaled converge gradient)are used to explore their prediction accuracy.The results show that: 64 groups of bio oil calorific value data are predicted,and 70% of them are selected as the training set,15% as the verification set,and 15% as the test set.The highest correlation coefficients of the Bayesian regularization algorithm training set and the prediction model are 0.999 and 0.994 respectively,and the lowest MSE is only 0.012.This shows that the algorithm can accurately predict the calorific value of bio oil under different N2 purging rates,HZSM-5 addition ratio and pyrolysis temperature,and provide guidance for the subsequent pyrolysis process of Litsea cubeba kernel meal.There are 16 figures,16 tables and 90 references.