Optimization of Reducing Acid of High-acid Feedstock of Biodiesel Based on Artificial Neural Networks

In order to get the optimal conditions of reaction, based on single factor experiment, oleic acid as the high-acid feedstock of biodiesel, orthogonal experiment and artificial neural networks were applied to optimize the synthetic conditions for reducing acid of high-acid biodiesel feedstock catalyzed by SO42/ZrO2 solid super acid. Based on orthogonal experiment, the three layers error back-propagation network (BP network) model was trained to reflect correlation of experimental data. And the optimal conditions were obtained from this network model as follows: SO42-/ZrO solid super acid was 4%(W/W) as catalyst, reaction temperature was 97.5 degrees C, reaction time was 180mins, and flow rate of gaseous methanol was 1.65 L.min(-1). Under the optimal conditions, validated experiment showed that the conversion rate of oleic acid was 96.89%, and the relative error was 0.03% compared with the predicted value. Therefore, the optimal conditions obtained based on BP artificial neural network are reliable and have better practical value.