การทำนายความสามารถในการละลายของแก๊สไฮโดรเจนซัลไฟด์ ในการทำบริสุทธิ์แก๊สธรรมชาติ โดยโครงข่ายประสาทเทียม
Hydrogen sulfide is found in petroleum and natural gas products. Natural gas containing more than 4 ppm is defined as a sour gas. Hydrogen sulfide pollutes an environment at the refinery and the gas separation area. It also causes corrosion in process equipment and pipelines. Solubility of hydrogen...
محفوظ في:
| المؤلفون الرئيسيون: | , |
|---|---|
| مؤلفون آخرون: | |
| التنسيق: | Senior Project |
| اللغة: | Thai |
| منشور في: |
จุฬาลงกรณ์มหาวิทยาลัย
2020
|
| الموضوعات: | |
| الوصول للمادة أونلاين: | https://digiverse.chula.ac.th/Info/item/dc:10940 |
| الوسوم: |
إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
|
| المؤسسة: | Chulalongkorn University |
| اللغة: | Thai |
| الملخص: | Hydrogen sulfide is found in petroleum and natural gas products. Natural gas containing more than 4 ppm is defined as a sour gas. Hydrogen sulfide pollutes an environment at the refinery and the gas separation area. It also causes corrosion in process equipment and pipelines. Solubility of hydrogen sulfide is a main parameter to design, develop, and optimize the natural gas purification process. Moreover, it represents the thermodynamic properties of hydrogen sulfide. Amine based solvents (such as MDEA, EDA, and DIPA) and Ionic liquids (such as [C₂MIM][eFAP], [BMIM][Tf₂N], [DMEAH][For], and [C₈MIM][PF₆]) have high potential to be used as hydrogen sulfide absorbent. However, an experimental study for the solubility of hydrogen sulfide is very costly. It is because of the high pressure experimental operating condition. Additionally, an expert in the natural gas purification process is essential for selecting of appropriate absorbents. In this research project, artificial neural networks (ANN) and data science were applied for predicting the solubility of hydrogen sulfide. The solubility data of hydrogen sulfide in 62 different absorbents over ranges of temperature and pressure were collected from the literatures. The collected data were then analyzed by ANN function in MATLAB program. In the ANN model, temperature, pressure, absorbent, and its concentration were set as inputs, while the solubility of hydrogen sulfide (mol H₂S/mol [subscript [absorbent]]) was set as an output. The model was optimized by Levenberge-Marquardt (LM) algorithm with data splitting (70% for training, 15% for validating, and 15% for testing) and 17 hidden neurons. It was found that very satisfactory predicted results were obtained with a coefficient of determination (R²) of 0.9817 and a mean squared error (MSE) of 0.0138 |
|---|