Machine learning approach applied for interpretation of data from multisensor technique used in combustion
Typ
Examensarbete för masterexamen
Program
Innovative and sustainable chemical engineering (MPISC), MSc
Publicerad
2021
Författare
Stenbäck Storm, Benjamin
Hernández Leal, Maria Alejandra
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
As the need for more sustainable energy production has increased, the thermochemical conversion
of biomass as an alternative to combustion of fossil fuels has gained more attention. One of the
most e cient technologies used for combustion of biomass is
uidized bed combustors, since the
bed provides a good solid-gas mixing and heat transfer. One of the major problems, related to the
technology, is bed material agglomeration; a clustering of bed particles caused by the formation of
low-melting compounds such as K- and Na-silicates. The agglomeration causes disturbance in the
uidization of the bed and as a consequence the operation has to be stopped and the bed has to
be exchanged, thereby increasing operational costs. In this context, a method used to follow the
quality change of the bed could be used to prevent premature stops. This project investigates the
feasibility of the use of an electronic tongue (ET); a sensor consisting of electrodes with overlapping
selectivities that can be used to analyze quality parameters, as a tool to monitor the bed quality. For
this purpose bed samples as well as data from a six months period from the bubbling
uidized bed
boiler combusting wood at Idb acksverket were analyzed. The received data from the plant included
fuel composition, operational data and response signals obtained from an ET. The data from the
ET, using three working electrodes of Pt, Au and Rh, was gathered by using pulse voltammetry
on a leachate solution from used bed material. Chemometrics was used to analyze the response
signals from the ET and the used bed material was analyzed using SEM-EDS, AAS and XRF. The
analysis of the bed material showed that the ash-forming matter from the fuel is responsible for
layer build-up around the bed particles. The layer consists of a Ca-rich outer layer and a Na-rich
inner layer. Furthermore, the analysis showed that an increase in the Ca/Si ratio in the fuel resulted
in an increase of the Ca/Si ratio in the outer layer which in turn results in, not only Ca-silicates
but also other, more leachable, compounds. The results of the analysis, together with the fuel data
were used to understand the connection between bed material quality and the response from the
ET. From the analysis of the used bed material further experiments were conducted with the same
ET on solutions with known concentrations of Ca, K and Na, to determine their in
uence on the
ET signal. The ndings in this study indicate that the ET can be used to follow the change of
the amount of leachable ions from the bed material. From the studied elements, changes in Ca
can be detected by the ET. The results also show that the ET can be used to distinguish between
mixtures of several compounds. The conclusion of this study is that the ET is a promising tool
for monitoring the quality change of a
uidized bed during combustion. Further investigation of
electrolyte, pulse-train, electrodes and leachable compounds as well as experiments on bed material
of di erent quality can add to the optimization of this tool. Nevertheless, the results suggest that
the ET could be used to detect changes in the composition of the bed material as well as to predict
when the bed has to be replaced in order to avoid agglomeration.
Beskrivning
Ämne/nyckelord
agglomeration , ash-forming matter , ash-layers , bed material , biomass , calcium , combustion , electronic tongue