A novel strategy for efficient biodiesel production: Optimization, prediction, and mechanism

Abstract

Resistance of biodiesel industrial production came from high energy consumption and feedstock costs. To solve it, Na2CO3@BFD catalyst was prepared from blast furnace dust and used to catalyze biodiesel production at low temperature. Biodiesel yield of 99.04 wt% was obtained under conditions optimized by response surface methodology of methanol/oil molar ratio 13.72/1, catalyst dosage 9.77 wt % and 74.86 °C for 1.62 h. The order of influence of the four factors was temperature (245.9) > time (109.8) > methanol/oil molar ratio (23.83) > catalyst dosage (1.19). Back propagation neural network model (BPNN) was optimized using genetic algorithm (GA) and sparrow search algorithm (SSA) to predict biodiesel yield. The evaluation indexes of mean absolute error, mean square error, root mean square error and mean absolute percentage error of SSA-BPNN were 0.9236, 2.0184, 1.4207 and 1.0247 (vs. 2.4329, 9.1037, 3.0172 and 3.5000 for GA-BPNN and 4.3291, 43.4693, 6.5931 and 6.9227 for BPNN), indicating that SSA-BPNN model had excellent prediction ability to effectively reduce experimental costs and resource consumption. The reaction kinetics of Na2CO3@BFD for transesterification process showed that its activation energy was 65.73 kJ/mol, lower than that of reported solid base catalyst, indicating that it had significant potential in biomass conversion.