Modeling of biomass fast pyrolysis products by artificial neural networks

The harvest wastes of tomato, pepper and eggplant plants that are mostly produced in greenhouses have an increasingly high impact in a negative way on environment both in the world and in our country Investigations of the disposal of vwastes with traditional methods are carried out within the framework of solid waste management in our country. The efforts to develop technologies that are supposed to reduce the negatif impact of waste biomass on the environment and transform it into value-added outputs are encouraged by global policies and action plans. Funding support provided to academic and sectoral research on obtaining energy from biomass within the framework of climate, environment and bioeconomy in the national and international arena maintains its popularity. In this study, the disposal of the harvest wastes of tomato, pepper and eggplant plants, which are classified as biomass originating from vegetable / agricultural waste within the definition of solid waste, by pyrolysis method and their potential in terms of energy have been evaluated. The thesis work consists of three main parts. In the first part, preliminary, component, elemental, FT-IR and ICP-MS characterizations of the harvest wastes of tomato, pepper and eggplant plants, which are defined as greenhouse wastes, were carried out. In the literature, it has been determined that it has similar properties with many biomasses that can be used as a source in the axis of energy and transformation into different products by pyrolysis method. In the second part, thermogravimetric analysis (TGA) of the harvest wastes of tomato, pepper and eggplant plants one by one, of equal mass double and triple mixtures in pyrolysis medium were carried out. Some important information on the thermal behavior of biomass with TG and DTG has been obtained from the information achieved within the framework of TGA. FWO, KAS and Starink kinetic models were created with TGA data of biomass. Activation energies (Ea) 69.41-104.44 kjmol-1 of single, double and triple mixtures of biomass were determined using kinetic models. When greenhouse wastes are evaluated in terms of FWO, KAS and Starink Kinetic model compatibility, activation energies and thermodynamic parameters, it has been evaluated that single, double and triple mixtures can be an important source for biomass energy production. The activation energy of the mixtures formed by tomato in single use of biomass and tomato in binary mixtures with pepper and eggplant provides important information in terms of mixture synergy. The diversity of biomass resources is an important factor in the number and diversity of the research on the need for technology infrastructure that will be used in obtaining energy from biomass. Especially in the research carried out to obtain energy from potential new biomass sources, the large number of parameters that will affect the product has created the need to develop new approaches in resource use such as time and cost. In this context, the literature include studies where different mathematical approaches are used in experimental designs The studies carried out by using artificial neural networks, which is one of the mathematical approaches and the working principle of the human neural network structure, in the solution of different research problems, and the pyrolysis method in obtaining energy from biomass are limited in the literature. The studies indicate that, the data sets examined by the researchers are usually provided from the literature. In the third part of the thesis, two different application data sets were created from the data obtained from the fast pyrolysis experiments carried out in the pyrolysis device in the Department of Environmental Engineering in Akdeniz University by using the artificial neural networks (ANN) method. The research on ANN architectures were carried out with two application datasets. Preliminary, component and elemental analysis results of biomass used in rapid pyrolysis experiments were clustered together with process parameters, and ANN architectures were examined on two application dataset 14 models. In terms of 4-21 input parameters and fast pyrolysis products, ANN models in 4 different output frameworks were examined in terms of the number of hidden layer neurons, the number of hidden layers and the percentage of training/testing/validation of the data. As a result of the studies, it was determined that the model performances for process parameters and elemental and component analysis clusters were more successful than other models. Theoretical results for the experiments that were not realized within the framework of single, double and triple mixtures of tomato, pepper and eggplant wastes used in this study, which were found to be successful in terms of performance, were produced for the liquid product. According to the results, it was determined that two of the models obtained from the first application data were more successful than the other models. It can be said that the model obtained as a result of the study has the potential to be used as a pre-application tool in terms of determining the liquid product before new fast pyrolysis experiments. It is envisaged that the pre-application tool can be developed with the new rapid pyrolysis test data to be performed in the pyrolysis device in the Department of Environmental Engineering of Akdeniz University and can be used as a pre-application tool for other products other than liquid products