Biodiesel-diesel blend optimized via leave-one cross-validation based on kinematic viscosity, calorific value, and flash point

Abstract

This study investigated the kinematic viscosity, calorific value, and flash point of three biodiesel–diesel blends (Luoping rape oil biodiesel, corn oil biodiesel, and catering waste oil biodiesel). Novel prediction equations for the calorific value were established based on the kinematic viscosity and flash point, and an optimal mixing ratio of biodiesel–diesel blends was obtained. Lederer’s equation, theoretical calorific values calculated using mass fractions, and the Hu-Burns equation, which are based on mixing ratio, are effective for predicting the kinematic viscosity, calorific value, and flash point of the three biodiesel–diesel blends. The mean absolute percentage error (MAPE) values of Lederer’s equation were 0.44%, 0.33%, and 0.68%, which were considerably higher than the results predicted using Krisnangkura equation. The MAPE values of the theoretical calorific values of mass fraction were 0.09%, 0.06%, and 0.11%. The MAPE values of Hu-Burns equation were 1.10%, 1.50%, and 1.00%, and the error was also smaller than that of Wickey-Chittenden equation and Chevro equation, which are consistent with the experimental results. A prediction model of kinematic viscosity, calorific value, and flash point values was established using Lederer’s equation, mass fraction theory calorific value, and Hu-Burns equation through the leave-one-out cross-validation method. The power exponent model was the most suitable for simulating the relationship between kinematic viscosity and calorific value, and the maximum errors of the three blends were 0.55%, 0.81%, and 0.90%, and the MAPE values were 0.16%, 0.39%, and 0.38%. The exponential decay model was the most suitable for simulating the relationship between the flash point and calorific value; the maximum errors of the three biodiesel blends were 0.98%, 0.68% and 0.58%; and the MAPE values were 0.51%, 0.25%, and 0.30%. Therefore, the calorific values obtained using these two models were highly accurate. Considering kinematic viscosity, flash point, and calorific value as the optimization indexes, the optimization interval was determined, and the optimal biodiesel mass fraction mixing ratios were 24.4%, 25.1%, and 25.2%. This study has important practical implications for the efficient utilization of biodiesel–diesel blended fuel. • Kinematic viscosity, calorific value and flash point are predicted by mixing ratio. • The calorific value is predicted by kinematic viscosity and flash point respectively. • A model for predicting the calorific value of biodiesel blended fuel is proposed. • The predicted calorific value error of biodiesel blending fuel is less than 1%. • The biodiesel blending proportion of the three blended fuels is about 25%.