A novel approach for improving carbon fixation of Chlorella sp. by elements in converter steel slag using machine learning

Abstract

This study investigates the biomass and carbon fixation rate of Chlorella sp. by the main valuable elements in converter steel slag in F/2 seawater medium. The results indicate that Ca, Mg, P, Si, Fe, and Mn can increase the biomass and carbon fixation rate of Chlorella sp., while Cu, Zn, Cr, and Al can decrease the biomass and carbon fixation rate of Chlorella sp.. Three machine learning methods known as Back Propagation neural network (BPNN), decision tree (DT), and random forest (RF) were applied to construct the prediction model for the carbon fixation rate of Chlorella sp. based on real-life experimental data obtained from single factor experiments. The overall results exhibited that the BPNN model is better than the DT model and RF model to predict the carbon fixation rate of Chlorella sp.. Finally, the maximum carbon fixation rate for Chlorella sp. predicted by BPNN model is 50.86 mg/(L·d), which was 2.46 times that of the control group, under the optimum conditions of Ca 5.77 g/L, Mg 4.74 g/L, P 1.27 g/L, Si 6.31×10−1 g/L, Fe 6.50×10−4 g/L, Mn 5.00×10−5 g/L, Cu 2.51×10−6 g/L, Zn 4.98×10−6 g/L, Cr 0 g/L, and Al 0 g/L in F/2 seawater medium.

Keywords

• Machine learning
• Converter steel slag
• Valuable elements
• Chlorella sp.
• Biological carbon fixation
• Prediction model