Optimization algorithms for modeling conversion and naphtha yield in the catalytic co-cracking of plastic in HVGO

Abstract

The catalytic co-cracking of 2.5–10 wt% low-density polyethylene (LDPE) in heavy vacuum gas oil (HVGO) was carried out in a fixed-bed microactivity test (MAT) unit and studies on the effect of LDPE loading, cracking temperature and nature of zeolite catalyst generated reasonable data that were used to model HVGO/LDPE conversion and naphtha yield using four metaheuristic-based nature inspired optimization algorithms. Using 13 different input parameters into the data mining, less dominant variables with < 0.49 correlation co-efficient to the targets were filtered out. The effect of deterministic based feature on variable filtering prior to modeling stage was conducted for both HVGO/LDPE conversion and naphtha yield. The prediction performance of the developed models in both training and testing was evaluated using the mean square error (MSE), root mean square error (RMSE), coefficients of correlation (R) and determination (R²). Among the four algorithms, GPR-BO showed highest conversion prediction performance for both data training and testing, with MSE = 1.14 × 10⁻⁹ and 2.02 × 10⁻⁹, RMSE = 3.37 × 10⁻⁵ and 4.62 × 10⁻⁵, R² and R = 1.00 respectively. For naphtha yield prediction, the ANN-PSO showed highest performance for both data training and testing, with MSE = 0.020 and 0.0663, RMSE = 0.143 and 0.189, R² and R = 1.00 respectively.