Study On Flow Law Of Particles With Large Density Difference In Dual Circulating Fluidized Bed Based On Pressure Signal

At present,the gasification technology carried out in fluidized bed equipments is an important way to utilize biomass energy.In these fluidized bed devices,the dual circulating fluidized bed system uses a combination of a bubbling fluidized bed and a fast fluidized bed to sub-regionally enhancing the gasification and combustion processes in the biomass gasification reactions,which can effectively improve the quality and yield of gas productions.In this system,the differences in the flow states of the two beds results in a complex particles flow pattern,and the physical differences in the mixed particles of the biomass-inert fluidized medium increases the complexity.For this reason,this paper used particles with large density difference composing of quartz sand and rice husks as the experimental bed material,and carried out pressure signal analysis on bubbling bed and dual circulating fluidized bed cold state experimental device to study the relationships between wave characteristics and particle flow law.Then a data-driven model and hydrodynamic model were established to achieve particles circulation flow rate prediction and fault diagnosis,which provided a theoretical basis for the operation and design of the dual circulating fluidized bed gasification units.(1)The initial fluidization characteristics of the rice husk-quartz sand particles with large density difference were experimentally studied in the bubbling bed cold experimental device and the regression fitting of the initial fluidization velocity empirical formula was applied.It was found that the increase of rice husk mass fraction and particle size of the quartz sand resulted in an increase in the initial fluidization velocity of the particles with large density difference.The feature extraction method for bed pressure signals under different superficial gas velocity,bed material mass,average quartz particle size and rice husk mass fraction indicated that the motion of particles in the bubbling bed is largely affected by the influence of the bubbles phase,so the main frequency of the pressure signal was distributed at about 5 Hz,and the mid-frequency band after HHT transformation,the 3 scale(6.25?12.5 Hz) and 4 scale(3.125?6.25 Hz) after wavelet multi-resolution analysis accounted for a large amount of energy,and the study of recurrence plot and recurrence parameters(Laminar flow rate)of pressure signals showed showed that the particle motion in the bed is obviously intermittent due to the influence of the bubble phase by the nonlinear analysis method.(2)In the experimental system of dual circulating fluidized bed cold state,the study on the flow law of particles with large density difference betwwen two bed was carried out.It was found that the changes of circulation flow rate and the circulating material composition which represent the particle motion law were related to the gas velocity of the gasificer,the gas velocity of the riser,the mass inventory of bed material,particle size of quartz sand and initial rice husk mass fraction.The mass inventory of bed material had a great influence on the particle motion law,and in actual production,the monitoring of this parameter should be strengthen controled.Based on the experimental results,BP neural network,genetic algorithm optimized BP neural network,support vector machine,least squares support vector machine,kernel limit learning machine and nuclear limit learning machine model were established for particles circulation flow rate and circulating material composition change prediction.The average absolute percentage error of the kernel extreme learning machine model for the prediction of parameters in the above two parameters were 2.35% and 1.48% respectively,with high generalization ability,high prediction accuracy and short prediction time.It could be used as a better model to monitor and warn the state parameters during operation.(3)Through the pressure signal analysis,the particle flow law of the dual circulating fluidized bed system riser(fast bed) under different control parameters were studied.It was found that the average frequency distribution of the pressure signal fluctuation is around 25 Hz,and the high frequency part after HHT transformation and the 1 scale(25-50 Hz) and 2 scale(12.5?25 Hz) energy after wavelet multiresolution decomposition are large,indicating there is a strong particle collision and friction during the movement of the particles in the riser.In addition,it was also found that the control of the gas velocity of the riser affects the particle motion law in the gasifier by controlling the particles circulation flow rate between the two beds,making the main frequency of the pressure signal is distributed at about 10 Hz,the energy ratios of the 2 scale(12.5?25 Hz) and 3 scale(6.25?2.5 Hz) after wavelet analysis were relatively larger,and the corresponding recurrence parameter LAM also shows a certain change law.(4)The agglomeration and blockage faults were simulated,by adding the fluidized bed biomass agglomerates to the dual circulating fluidized bed and blocking different areas of the gasification chamber air distribution devices.The relationship between the characteristic parameters of pressure signals and the degree of agglomeration and clogging position under various fault conditions was studied.It was found that the agglomeration and clogging faults caused the particle flow law to deteriorate,and the corresponding pressure signal fluctuation characteristics presented different laws of change with different faults.On this basis,the wavelet decomposition(variational mode decomposition) and sample entropy(feature energy)were combined to extract the feature of the pressure signal,and the kernel limit learning machine model was established to realize the fault classification diagnosis.learning machine model based on wavelet decomposition-feature energy extraction had the training and test accuracy of up to 100% and 82.50%,respectively,and the application of pressure signal in diagnosising of dual circulating fluidized bed system could be realized.(5)Starting from the spatial pressure distribution of the dual circulating fluidized bed,the riser pressure drop model was established according to the particles concentration distribution(dense phase-dilute phase partition) and particles velocity change(acceleration-sufficient development zone).Compared and selected the optimal riser pressure drop mode,the hydrodynamic model was established based on the two-bed pressure balance and bed mass conservation,and realize the prediction of circulating flow rate of particles with difference.During the process of model establishment,the particles with large density difference were replaced by homogeneous particle or phase-separated particle model according to the characteristics of the partition.The prediction results show that although there is a certain prediction error(the maximum prediction error was-22.18%),the hydrodynamic model has the same trend for prediction with the experimental measurement,so it has a higher applicability in the prediction of circulating flow rate in a dual circulating fluidized bed system,especially for particles with large density difference.