Combined optimization of petroleum hydrocarbon degradation rate prediction model using response surface methodology and artificial neural networks

This study aims to optimize the conditions for microbial degradation of petroleum hydrocarbons. The laboratory-cultivated petroleum hydrocarbon-degrading strain BM-W10 was used as the research object. Through 16S rRNA gene sequencing and phylogenetic tree construction, it was identified as Bacillus mycoides. To investigate the effects of temperature, pH, inoculation amount, and crude oil concentration on the degradation efficiency of the BM-W10 strain, a single-factor experimental design was employed. Furthermore, this study combined Response Surface Methodology (RSM) and Artificial Neural Networks (ANN) to establish a prediction model for petroleum hydrocarbon degradation rate based on a Box-Behnken experimental design. Simultaneously using Genetic Algorithm (GA) to optimize ANN. The results showed that the prediction results of the ANN model were closer to the actual values, with indicators such as R², average absolute deviation (AAD), and root mean square error (RMSE) being superior to those of the RSM model, indicating that ANN exhibited stronger prediction and fitting capabilities in nonlinear regression analysis. The optimal degradation conditions predicted by the ANN model were: temperature 33.89 °C, pH 6.99, inoculation amount 5.14 %, and crude oil concentration 0.98 %, under which the degradation rate of petroleum hydrocarbons by the BM-W10 strain could reach 62.58 %. The research findings will contribute to the promotion of biotechnology for solving environmental pollution.