Online-Klassifizierung unterschiedlicher Brennstoffe und Brennstoffcharakteristika als Basis für eine Flexibilisierung der Feststoffverbrennung

The heterogeneity of biomass characteristics is a major challenge for all thermochemical conversion processes. This work describes the development from the idea to a working prototype for a system with which online fuel characteristics can be identified. The idea as well as the scientific work are based on the following hypothesis: If warm air flows through a solid bed of biogenic fuels, it interacts with the solid bed during the flow. The physical processes occurring in this process measurably change the properties of the air. Based on these changes, it is possible to draw conclusions about relevant fuel prop erties (water content, bulk density) and the type of fuel.
In a first step, a test rig was developed with which defined fuel samples can be dried and the changes in temperature, water content and pressure drop can be analyzed. First, a screening test series with a wide range of feedstocks (wood chips, forest residues, wood pellets, saw dust, corn cobs, biochar from digestate and aluminum oxide spheres) was conducted to create an initial data basis. Through this series of tests, the general feasi bility of the idea could be demonstrated. The data from the test series were used to develop an evaluation logic by combining data-based and analytical approaches. Based on this preliminary work, a prototype was developed and connected directly to an auto matically fed grate furnace in the Fraunhofer UMSICHT technical center. After commis sioning of the prototype, test series were carried out with wood chips, forest residues, saw dust and wood pellets. These tests created the data basis for training and evaluating a machine learning algorithm. For evaluation the furnace was alternately fed with wood chips, forest residues and wood pellets. Regarding the accuracy, it could be shown that the identification of the feedstock succeeds with a probability of about 90% for wood chips and forest residues and 100% for pellets. The determination of the water content and the bulk density showed that the orders of magnitude were correctly determined. The deviation in the water content was on average between 1.7% and 4.1% in absolute terms. The precision in the determination of the bulk density lies with a relative error of 10.42% for wood chips, 15.41% for forest residues and 5.98% for wood pellets. In the end, the hypothesis was confirmed, and the basic idea was turned into a working proto type.