Prediction of the Electrical Strength and Boiling Temperature of the Substitutes for Greenhouse Gas SF₆ Using Neural Network and Random Forest

Abstract

Finding substitutes for sulfur hexafluoride (SF6), a gas with extremely high global warming potential, has been a persistent effort for years in the field of high voltage power equipment, which focuses on the evaluation of the electrical strength and boiling temperature for the practical purpose. Following up the previous proposed linear regression models, this work introduces machine learning algorithms including artificial neural network (ANN) and random forest (RF) as the potential approaches to predict the electrical strength and boiling temperature. Based on a series of descriptors derived from the molecular structure of 74 molecules, the performance of three different methods: multiple linear regression, artificial neural network and random forest are compared and assessed in terms of the sensitivity to the sample size, prediction accuracy and stability, and the interpretability of predictors. Considering the available data are limited, random forest shows superior performance with higher robustness and efficiency. The same approaches were applied to the boiling temperature and random forest produced better results as well. Besides, the variable importance ranked by RF improves understanding of the correlation between the molecular properties and electrical strength. It provides important insights to analyze the properties of the SF6 substitutes during the design and synthesis of the new eco-friendly gases in power equipment.

Research and Analysis of Insulating Gas in Unified Test Conditions

“…However, the prediction error of insulation strength of polar molecules was large. Rong et al 12 used neural networks and random forests to predict the electrical strength and boiling temperature of the substitutes for greenhouse gases. However, they were unable to describe the breakdown voltage of environmentally friendly gas with the change of gas pressure and spacing.…”

Research on a prediction model for gas insulation performance based on Pareto optimisation

“…In this manuscript, data for important molecular structure parameters [9, 11–13] related to the insulation process are used; these parameters include the molecular surface electrostatic potential P A , molecular volume $${V}_{\mathrm{m}}$$, highest occupied orbital energy E HOMO , lowest non‐occupied orbital energy E LUMO , polarisability $$\alpha $$, dipole moment $$\mu $$, molecular surface area A s , molecular surface average statistical deviation $${s}_{\mathrm{t}\mathrm{o}\mathrm{t}}^{2}$$, average deviation $$\pi $$ of the electrostatic potential on the surface of the molecule, and balance v of the positive and negative electrostatic potentials.…”

“…The comparison between Formula () and Ref. [12] is shown in Figure 4. The above formula shows that A s and $${P}_{\mathrm{A}}$$ are positively correlated with the gas insulation strength and that $${V}_{\mathrm{m}}$$ and $$\alpha $$ are negatively correlated with the gas insulation strength.…”

“…Linear regression was used to screen the parameters with the highest correlations with the insulation strength of the gas, and the insulation strength and boiling point of the insulation gas were predicted with the random forest method. However, the random forest method is based on a black box model with poor interpretability, random generation results, and poor reproducibility [12]. Shuai Yang et al from Hubei University of Technology established a prediction model for the insulation strength of polar and non-polar molecules by calculating isosurface electrostatic potential parameters, molecular polarisability and dipole moments.…”

Study on Relative Electrical Strength of SF6 Substitute Gas Based on Density Functional Theory

“…Based on ionization energy, electronegativity and molecular diameter, 137 alternative gases were screened, among which 5 gas molecules were considered as promising alternative SF6 gas molecules in [16]. Through neural networks and random forests, the electrical strength and boiling point of SF6substituting gases were predicted in [17]. The arc extinguishing performance and electric strength of SF6 substitute gas C5F10O were theoretically studied in [18].…”