Review on the challenges and strategies in oil and gas industry's transition towards carbon neutrality in China

MDs-NP: a property prediction model construction procedure for naphtha based on molecular dynamics simulation

“…In model construction, models based on ANN achieved relatively poor performance, which was probably as a result of the following: (1) the quantity of data used for training in this work was too small to effectively determine the optimal parameters in ANN. (2) The distribution of the sample composition in this work was too narrow which resulted in insufficient data diversity for ANN. (3) ANN requires finer input data cleaning and data mining, whilst some molecular descriptors have unclear meanings and inter-correlation, which is not conducive Considering the number of descriptors in each group (RDKit, Mordred, and RDKFp + NK) is quite large compared with the number of data used to train the machine learning model, we chose ten of the descriptors that were most relevant to viscosity and had a correlation below 0.7 with each other from each group in order to improve the performance of the viscosity model.…”

“…Viscosity had a larger MAPE compared to density might be attributed to the following: (1) the simulation of viscosity requires a longer simulation process and can lead to larger errors. (2) Considering the number of representative molecules, we did not optimize the force field parameters to improve the viscosity accuracy of each substance individually, but directly used the original OPLS-AA force field parameters with the highest versatility. (3) The viscosity of representative molecules distributed over a wide range, which varied with carbon number and structure.…”

“…In this work, due to the inability to directly calculate the molecular descriptors of mixtures, the calculation of molecular descriptors used to characterize the structure of mixtures can be divided into the following three steps: (1) yield the simplified molecular input line entry specification (SMILES) expression of all the 50 representative molecules based on the PubChem database, which clearly describes the molecular structure with an ASCII string [42]. (2) Calculate molecular descriptors by inputting molecular SMILES into open-source toolkits such as RDKit and Mordred [41]. RDKit is a widely used open-source toolkit for chemoinformatics with functions include molecular descriptor generation and molecular operations.…”

“…new challenges [1][2][3]. Molecular-level understanding and interpretations of the complex refining process have received increasing attention, which render molecular management a focus of petrochemical industry development [4].…”

From the Lab to the Field: Organic Materials for Industrial Applications and Environmental Remediation

“…Particularly, development of non-conventional reservoirs requires technological innovation of new inputs. [30] In non-conventional reservoirs, one of the most important supplies for hydraulic fractures are proppants, used to keep the fracture open. Although there are various types of proppants, such as silica, ceramic and resincoated sand; a great challenge is the design of ultra-light weight proppants (ULWP).…”