Machine learning-based fuel characterization and fuel property effects on bubbling fluidized bed boiler operation
Moisio, Kalle (2022)
Moisio, Kalle
2022
Ympäristö- ja energiatekniikan DI-ohjelma - Programme in Environmental and Energy Engineering
Tekniikan ja luonnontieteiden tiedekunta - Faculty of Engineering and Natural Sciences
This publication is copyrighted. Only for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2022-05-10
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202204263680
https://urn.fi/URN:NBN:fi:tuni-202204263680
Tiivistelmä
Tightening environmental obligations and climate change cause pressure to use more renewable energy sources in the energy industry. Renewable energy sources are commonly more challenging to utilize than fossil fuels. Bubbling fluidized bed (BFB) boiler technology allows using a wide range of fuels efficiently for power and heat generation. Fuel properties vary between different fuel types and affect boiler operation. These kinds of fuel properties are, for example, elementary composition, moisture content, the proportion of volatiles, and ash composition.
Boiler operators use an automation system to operate the boiler. Nowadays, advanced automation systems can control the boiler mainly independently. However, automation systems do not necessarily have information about currently used fuel and cannot automatically optimize boiler operation based on used fuel.
In this thesis, a literature survey was done to investigate how different fuels affect BFB boilers’ operation and how the boiler automation system works. It was also studied if used fuel could be characterized by utilizing existing measurements and how could fuel characterization results be communicated into the boiler automation system.
Data used in the study was from a BFB boiler which is designed to use biomass and a mixture of peat and biomass. Information on used fuel was available from guarantee tests and later executed tuning tests. Machine learning-based methods were used for fuel characterization and finding suitable measurements. As a result, a fuel characterization model was created. The input features for the model were mainly from the fuel feeding system. The model was able to characterize fuel correctly with good accuracy.
The developed fuel characterization model was used to investigate the effects of different fuel types on the boiler’s measurements. Effects were found from expected measurements, such as NOx and SO2 emissions, but the fuel type seemed to have also a significant impact on bed temperature and the boiler’s stability.
The fuel characterization results could be communicated into the boiler automation system by utilizing possibilities of Open Platform Communication Unified Architecture (OPC UA) specification. It is used widely for data transferring in automation, and it provides a secure and reliable way to transfer data.
As a follow-up to this work, it is proposed that similar fuel characterization methods would be studied on different boilers and with different fuel types. It might be possible to characterize fuel more specifically with a greater data set. In addition, utilizing fuel characterization results in the boiler automation could be studied with a separate test campaign.
Boiler operators use an automation system to operate the boiler. Nowadays, advanced automation systems can control the boiler mainly independently. However, automation systems do not necessarily have information about currently used fuel and cannot automatically optimize boiler operation based on used fuel.
In this thesis, a literature survey was done to investigate how different fuels affect BFB boilers’ operation and how the boiler automation system works. It was also studied if used fuel could be characterized by utilizing existing measurements and how could fuel characterization results be communicated into the boiler automation system.
Data used in the study was from a BFB boiler which is designed to use biomass and a mixture of peat and biomass. Information on used fuel was available from guarantee tests and later executed tuning tests. Machine learning-based methods were used for fuel characterization and finding suitable measurements. As a result, a fuel characterization model was created. The input features for the model were mainly from the fuel feeding system. The model was able to characterize fuel correctly with good accuracy.
The developed fuel characterization model was used to investigate the effects of different fuel types on the boiler’s measurements. Effects were found from expected measurements, such as NOx and SO2 emissions, but the fuel type seemed to have also a significant impact on bed temperature and the boiler’s stability.
The fuel characterization results could be communicated into the boiler automation system by utilizing possibilities of Open Platform Communication Unified Architecture (OPC UA) specification. It is used widely for data transferring in automation, and it provides a secure and reliable way to transfer data.
As a follow-up to this work, it is proposed that similar fuel characterization methods would be studied on different boilers and with different fuel types. It might be possible to characterize fuel more specifically with a greater data set. In addition, utilizing fuel characterization results in the boiler automation could be studied with a separate test campaign.