Characteristics Of Slurrvability Of Carbon-fuel And Solid Concentration Prediction OfBlending-CWS By Neural Networks

As a new type of gasification material and clean fuel which can substitue petroleum and, Coal water slurry (CWS) has been widely applied in our country, which has obtained significant social and economic benefits. Along with development of economic and improvement of technology research and development, kinds of coal for CWS are needed to broaden, in order to achieve diversification of fuel and raw materials. On the other hand, slurrying of blending coals is required to meet national situation of varying coals quality. Therefore, preparing high concentration slurry by using various carbon based materials have become the development trend of slurry fuel. In this thesis, slurrying characteristics of single fuel and blending fuel were rearched for petroluem coke, oil sand and coals with different metamorphism. Artificial neural networks were constructed for predicting slurry concentration of blending coals. Fractal theory was used to deeply analyze microscopic pore structure characteristic of single coal and blending coals, and its relationship with slurryability of these coals.Physical and chemical properties of 20 coals with different metamorphism and slurryability, rheological behavior, thixofropy and stability of CWS were studied. It was concluded that inner water, Oxygen content, coal rank (O/C、O/H), Oxygen-containing fuctional groups were main factors influencing slurryability of coal. Coal with high content of inner water, Oxygen, Oxygen-containing fuctional groups and low coal rank had poor slurryability, and the slurry concentration was low. Additives alse had important effect on the characteristics of CWS. During the process of slurrying carbon-fuel, additives improved its slurryability by improving wettability of particle’s surface, reducing solid-liquid interfacial tension and increasing Zeta potential.Blending the coal with nice slurryability into coal with poor slurryability could efficiently improve slurryability of blending coals, and vice versa. Rational blending schemes even made the slurry concentration of blending coals higher than each single coal used for blending. During the blending process, increasing the content of coal with strong hydrophilic made the slurryability of blending coals poorer, and increasing the content of coal with strong hydrophobic made the slurryability of blending, coals better. Particle size distribution of coals used for blending also influenced the slurryability of blending coals. Different particle size distribution of coals used for blending caused large error between the actual slurry concentration and value of linear weighed sum, which meaned that, different particle size distribution caused the concentration of blending coals appearing non-linear properties, and two coals with great differece of particle size made the blending coals’slurryability higher. Single coal and blended coals with grading sizes both increased concentration by 1-2%.Fractal theory was used for researching the microscopic pore structure characteristic of single coal and blending coals, and its relationship with slurryability. Results showed that, fractal dimension D1 had no significant linear relativity with coals’pore stucture and slurryability, and the fractal dimension D2 had nice linear significant relativity with blending coals’pore structure and slurryabilty, with correlation coefficient R2 between 0.8562 and 0.9434. As increasment of blending coals’D2, the BET specific surface area and total pore volume monotonically increased, and average pore diameter decreased monotonically. Single coal’s fractal dimension D2 had nice linear relativity with its BET specific surface area and average pore diameter, and the correlation coefficient R2 was 0.8977 and 0.8625, respectively. To not only single coal but also blending coals, D2 had high linear relativity with pore structure, and D2 could be used for describing spatial roughness coals’pore structure. Therefore, fractal dimension D2 was fitter for evaluate coal’s slurryability.Characteristics of slurryability of blending carbon-fuel and coal were studied. Results showed that, blending these carbon-fuel, coal, oil sand and petroluem coke, could take advantages from each other, and get better slurrying effect. Blending coal into oil sand decreased slurry’s viscosity and increase fixed viscosity concentration significantly. As increasment of petroleum coke in the mixture of petroleum and brown coal, its slurry concentration increased markedly, and the slurry’s pseudoplastic behavior weakened. As the content of brown coal increasing, the slurry’s water separating proportion decreased significantly, which meaned the slurry’s stability improved.With average particle diameter (D), content of inner water (Mad), ash (Aad), volatiles (Vad), Oxygen (Oad) and coals’blending ratio as input factors, various BP neural networks were constructed to predict blending coals’ slurry concentration. Results showed that, the average absolute error of prediction by the best BP neural network was 0.47%, and the average absolute error of linear fitting was 0.90%. Taking advantage of PSO algorithm’s feature of global optimization, PSO-BP neural network was constructed. The average absolute error of prediction by PSO-BP neural network was 0.32%, which was better than BP neural network.Researh in thesis is useful for broadening slurry materials and improving accuracy of blending coals’ slurry concentration prediction. It also offer technical support arid theoretical reason for preparing slurry with high solid concentration of various carbon-fuel.