Investigation of the chemical exergy of torrefied biomass from raw biomass by means of machine learning

Kartal F., ÖZVEREN U.

Biomass and Bioenergy, cilt.159, 2022 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 159
  • Basım Tarihi: 2022
  • Doi Numarası: 10.1016/j.biombioe.2022.106383
  • Dergi Adı: Biomass and Bioenergy
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, PASCAL, BIOSIS, Biotechnology Research Abstracts, CAB Abstracts, Compendex, Environment Index, Geobase, INSPEC, Pollution Abstracts, Veterinary Science Database, DIALNET, Civil Engineering Abstracts
  • Anahtar Kelimeler: Torrefaction, Biomass, Artificial neural network, Chemical exergy prediction, Machine learning, PROXIMATE ANALYSIS, WOODY BIOMASS, HEATING VALUE, TORREFACTION, WASTE, GASIFICATION, PREDICTION, CARBON, MODEL, TRANSFORMATION
  • Marmara Üniversitesi Adresli: Evet

Özet

© 2022 Elsevier LtdTorrefaction is one of the most important pretreatment processes to improve the quality of biomass as fuel and overcome its disadvantages. Predicting the chemical exergy of torrefied biomass is essential for evaluating and optimizing the performance of biocoal-based power plants. Therefore, the authors report on a wide range of research that has been conducted to accurately measure the chemical exergy of solid fuels. Nowadays, many researchers are working on computational methods to reduce the number of actions in experimental research. However, until now, researchers have not presented a model that predicts the chemical exergy of torrefied biomass considering the experimental conditions. This study is novel in two ways: first, the exergy of torrefied material was calculated using parameters of torrefaction conditions prior to the torrefaction process. Second, the developed model ANN predicts the chemical exergy of torrefied material directly from the results of proximate analysis of raw biomass samples. Statistical performance indicators show that the predictive capacity of the ANN model is satisfactory. The R2 value was greater than 0.92 for training and 0.79 for testing, while the MAPE value was less than 4% for both training and testing.