EFFICIENT RDFS COMBUSTION IN LAB-SCALE FLUIDIZED BED: PROCESS MODELLING AND EXPERIMENTAL VALIDATION

Abstract

The recent initiatives on the reduction and total abstention from the further investment in fossil fuels mining/extraction and utilization enforced the pace of the energy shift. For the Republic of Kazakhstan, where the share of coal in the energy generation is around 68.9% and renewables generate only 2.2%, as stated in the National Energy Report (KazEnergy, 2019), the energy transition has a core impact on the decarbonization initiatives declared in Paris Agreement. Hence, an investigation of the available wastes applicability for energy recovery is a noteworthy direction to achieve the goal for GHG emissions reduction. Utilization of the wastes is a subject of research relevant for both industry and academia. Two types of waste-derived fuels are commonly considered, 1) refuse-derived fuel (RDF) and 2) solid-recovered fuel (SRF). RDF is a fuel derived out of combustible components of Municipal solid waste (MSW) such as textile, wood, leather, plastics, food wastes, and some inert materials. In this study a new model has been established in Aspen Plus® for co-firing of coal and RDF in a fluidized bed reactor using the reactor’s hydrodynamics and detailed reaction kinetics. Experiments were conducted using laboratory scale bubbling fluidized bed reactor for the co-combustion of the RDF and locally available Ekibastuz coal. For the process development, two-stage validation was performed using data available in the literature and our previous experimental results from the semi-industrial fluidized bed reactor. The mean-absolute deviation between the model predictions and literature data were less than 2% for the emissions of CO2, CO, NO2 and SO2. Based on the developed model, optimum temperature and RDF/coal ratio were defined using the series of sensitivity analyses. The results demonstrated good agreement between literature and predicted values for flue gas composition. The proposed model is further being developed in order to provide real-time optimization of fuel mixture, air flow rate, and the flue-gas emissions. Machine learning algorithms were used to develop the artificial neural network to predict the thermal behavior of RDF with a root mean square error between the experimental and predicted value equal to 0.0027%.