This modelling approach allows for the formation of intermediate components and their subsequent conversion to final products, as shown in equation M.5.

biomass —> product component j M.5

where the rate of formation of component j with a yield Vj at a given time t is given by equation M.6:

koj exp (- Ep (Vj* — V) M.6

eft RT

where:

Vj* ultimate attainable yield of product j, i. e, the yield at high temperature and long residence times (77).

The constants k0j, Ej and Vj* cannot be predicted beforehand and must be

estimated from experimental data, a problem that increases as the number of reactions postulated increases. The model provides a simple scheme that can be used to predict product yields, This method has been used by Krieger et al. (77).

Some models have taken into account the competitive nature of some of the pyrolysis reactions which have been postulated to account for the variations in product yield. This is done by Bradbury et al. (44) for cellulose as shown in Figure 2.7.

Figure 2.7 Proposed Pyrolysis Model for Pure Cellulose (44)

Bradbury et al. (45) used this approach for their kinetic model as shown above. This model was based on the pyrolysis of pure cellulose. Theoretical and experimental results for weight loss agreed to within ± 5%. As a consequence this mode) has been used to account for char yields in the models of large particle pyrolysis (77, 78, 79, 80).

Koufopanos et al. (41,42) and Nunn et al. (76) proposed that the biomass pyrolysis rate could be related to the individual pyrolysis rate of the biomass components i. e.

biomass = a (cellulose) + b {hemicellulose) + c (lignin)

where:

• bracketed terms () represent the fractions of the biomass components not transformed into gases or volatiles

• a, b, c are the weight fractions of the corresponding biomass components in the virgin biomass.

The reaction scheme of the individual components then followed a similar reaction scheme as shown in Figure 8. In both cases, theoretical and experimental results for the weight loss agreed to within ± 10%. However, for the model of Koufopanos et al. (41,42) there was no indication that it could be used to predict product yields. Nunn et a!. (76) found that, in general, the calculated values fitted laboratory data within ± 7% for temperatures up to about 950-1000°C. Similar approaches have been adopted by Simmons (65), Salazar (68), Samolada (69) and Varhegyi (55).