Prediction of Vacuum Glass Insulation Performance Based on Optimization of Support Vector Regression Model

In the intelligent detection process of vacuum glass insulation performance, the heat transfer coefficient is a key parameter for measuring the ability of vacuum glass to hinder heat transfer. The traditional steady state method for detecting the heat transfer coefficient requires stringent experimental conditions and is time-consuming. In order to improve the efficiency and accuracy of detection, this study constructs a multi-input single-output support vector regression model, introducing three different parameter optimization algorithms to enhance model parameters and improve prediction accuracy. The temperature change variables at the fifth minute of the unsteady state process of the vacuum glass and the heat transfer coefficients detected by the steady state method are used as the input and output variables for training and predicting the support vector machine regression model, respectively. Different optimization algorithms can search the solution space to find suitable parameters for the support vector regression model. Ultimately, the optimized support vector regression model predicts the heat transfer coefficient of the vacuum glass. Experimental results proves that the support vector regression model optimized using sparrow search algorithm performs better. Its MSE, MAE and R² are 0.007729, 0.072165 and 0.974050, respectively.