Szczegóły publikacji
Opis bibliograficzny
Comparative analysis of predictive models for boiler types: impacts of fuel, technology, and operational parameters on flue gas emissions / Katarzyna SZRAMOWIAT-SALA, Kamil Krpec, Roch Penkala, Jiří Ryšavý // W: ECOS 2025 [Dokument elektroniczny] : 38th international conference on Efficiency, Cost, Optimization, Simulation and Environmental impact of energy systems : Paris, France, 29 June - 4 July 2025. — Wersja do Windows. — Dane tekstowe. — [France : ECOS], [2025]. — S. 1–13. — Wymagania systemowe: Adobe Reader. — Tryb dostępu: https://v4.event-vert.org/uploads/AbstractsBundle/ECOS2025/fu... [2025-07-21]. — Bibliogr. s. 12, Abstr.
Autorzy (4)
- AGHSzramowiat-Sala Katarzyna
- Krpec Kamil
- AGHPenkala Roch
- Ryšavý Jiří
Dane bibliometryczne
| ID BaDAP | 161331 |
|---|---|
| Data dodania do BaDAP | 2025-07-30 |
| Rok publikacji | 2025 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
Abstract
This study explores the differences in predictive models tailored to various boiler types, focusing on the effects of solid fuel type, combustion technology, and heat energy output on flue gas composition. Residential combustion appliances play a pivotal role in providing heating solutions worldwide, yet their environmental impact varies significantly. By analysing real-world data from multiple boiler systems, including an automatic boiler fuelled by wood pellets, a down draught boiler fuelled by lignite and a gasification boiler fuelled by hard coal, the research aims to refine pollutant emission predictions for carbon dioxide, carbon monoxide, nitrogen oxides, organic gaseous compounds and sulphur oxides. The study employs a comparative approach to model development, integrating empirical data collection with advanced computational techniques. Using neural networks (NN), long short-term memory networks (NN-LSTM), and convolutional neural networks (CNN), predictive models were developed and benchmarked for their accuracy and adaptability to different boiler types. These models, underpinned by rigorous statistical analysis and machine learning methods, uncover nuanced relationships between operational parameters and emission profiles. Key findings demonstrate the model-specific strengths and limitations for various boiler technologies. Neural networks provided robust performance in general applications, with NN-LSTM excelling in capturing time-dependent patterns and CNNs proving effective in handling spatial and categorical data. The study also examined specific boiler types, highlighting differences in emission profiles and combustion characteristics. For instance, the gasification boiler exhibited more stable emission patterns, resulting in higher predictive accuracy, while the down draught boiler fuelled by lignite presented challenges due to variability in combustion dynamics.
