Experimental And Modeling Study On The Reaction Rate Constants For Photocatalytic Degradation Of Organic Pollutants Based On Random Forest And Bayesian Optimization

The residual organic pollution in the ecosystem causes water pollution,threatens human health,and increases the risk of human infection with viruses.Therefore,efficient and thorough treatment of organic pollutant wastewater is an urgent problem that needs to be solved.Photocatalytic treatment of organic pollutant wastewater has attracted much attention because of its advantages of thoroughness,efficiency,and environmental friendliness.The current research on photocatalytic degradation of organic pollutants is mainly focused on the stage of experimental measurement.Optimization of experimental parameters is labor-intensive,and there is a lack of relevant theoretical models for photocatalytic treatment of organic pollutant wastewater.Therefore,we use a combination of machine learning and optimization algorithms based on a large amount of experimental data to establish a theoretical model applicable to the photocatalytic treatment of organic pollutant wastewater systems for predicting the reaction rate constants of photocatalytic degradation of organic pollutants and providing a theoretical basis for the smooth implementation of photocatalytic treatment of organic pollutant wastewater technologies.Therefore,we developed a photocatalytic treatment of organic pollutant wastewater theoretical model by combining machine learning and optimization algorithms and trained the model with a large amount of experimental data.The model can be used to predict the reaction rate constants of photocatalytic degradation of organic pollutants and provide a theoretical basis for the smooth implementation of photocatalytic treatment of organic pollutant wastewater technology.In this study,we combined machine learning with photocatalytic technology to investigate the experimental and theoretical modeling of photocatalytic degradation of organic pollutant systems for the problem of the lack of theoretical models related to the photocatalytic treatment of organic pollutant wastewater systems.The main research contents are as follows:（1）Photocatalytic degradation experimentWe selected fluoroquinolones（FQs）,tetracyclines（TCs）,and sulfonamides（SAs）as experimental objects and used Ti O₂（Degussa P25）as a photocatalyst to conduct adsorption and hotocatalytic degradation experiments.The Photocatalytic degradation rate constants（k）were measured with different experimental parameters（including the initial concentration of antibiotics,the catalyst dosage and the initial p H of the solution）.The effects of different experimental parameters on the photocatalytic degradation of antibiotics were systematically investigated and analyzed,and the patterns of adsorption rate,degradation rate,and photocatalytic degradation rate constants with experimental parameters were obtained.（2）Photocatalytic degradation of organic pollutants theoretical modelBased on the literature and the experimental data of this study,a Random Forest-Bayesian Optimization model is established using a computational method combining Random Forest and Bayesian Optimization algorithm.The predicted values of the model are consistent with the experimental values,with a coefficient of determination（R²）of 0.9701,a mean square error（MSE）of 0.0094,an average absolute relative error（AARD%）of 4.32%,and a maximum absolute relative error（MARD%）of 79.67%.To further test the predictive ability of the model,the calculated results of the present model（RF-BO）were compared with the models reported in the literature.The R²（0.9701）of the RF-BO model was higher than the theoretical models reported in the literature for photodegradation of multiple organic pollutants,and the AARD%（4.32%）was lower than the models reported in the literature.A feature importance analysis was conducted on the model using the SHAP（SHapley Additive ex Planation）algorithm.The input variables（organic pollutant（P）,ultraviolet light intensity（I）,catalyst dosage（D）,initial pollutant concentration（C₀）,and solution p H（p H））had an impact on the model output results as follows:P>C₀>D>I>p H,and the impact of the input variables on the model output results was consistent with the results of photocatalytic degradation experimentsThe best experimental parameters（the organic pollutant was Gatifloxacin,the UV intensity at 50.83 m W/cm²,the catalyst dosage at 1.62 g/L,the initial concentration of the pollutant at 3.10 mg/L,and the initial p H of the pollutant in solution at 5.20）were obtained by the Bayesian optimization algorithm,and the corresponding validation experiments were carried out.The results showed that the experimental value（0.4079）was in good agreement（ARD%=1.81%）with the predicted value（0.4005）under the best combination of experimental parameters,and the Random Forest-Bayesian Optimization（RF-BO）model had good prediction ability.