含有难降解有机污染物的废水严重威胁生态环境健康。光催化技术因其能利用太阳能，具有相对成本低、无二次污染、占地面积小、效能高等特点，被认为是具有巨大发展潜力的难降解有机污染物深度处理技术。然而，目前常见光催化技术存在与入射光接触面较小、需要额外能耗进行搅拌、材料回收困难等问题，限制了实际应用。将纳米光催化剂负载到硅铝酸盐载体上制备漂浮型光催化剂不仅成本低、环境友好、无需搅拌、能耗低、易回收，而且有助于光催化剂与入射光和氧气的充分接触。铋系纳米光催化由于具有特殊层状结构和较好的可见光响应能力，成为目前最有应用前景的纳米半导体材料之一。因此，本文以地质聚合物和膨胀珍珠岩作为硅铝酸盐基光催化剂载体负载铋系光催化剂制备辅助/自漂浮型光催化剂，以常见有机染料亚甲基蓝（MB）和地表水中检出频率较高的抗生素环丙沙星（CIP）作为目标污染物，探究其光催化效能与反应机理。主要研究结果如下：

 （1） 利用偏高岭土和常见工业固体废弃物粉煤灰为原料制备地质聚合物 （MFGP） ，负载BiOBr和导电材料乙炔黑 （AB），制备出整体块状复合光催化剂BiOBr/AB/MFGP。利用SEM、XRD、XPS和UV-Vis对BiOBr/AB/MFGP的形貌组成、光电性能等进行了表征，发现BiOBr和AB均匀固定在MFGP中，且不会影响地质聚合物原有的结构。AB中的导电碳网和MFGP中的硅铝酸盐网格可为光生电子提供转移途径，降低载流子复合率。此外，AB诱导的光热效应可促进光催化剂微界面处的光催化反应。BiOBr/AB/MFGP被自制铜网悬置在水/空气界面，在模拟太阳光照射30 min后，对20 mg/L MB的去除率达到96%，40 min内总有机碳含量（TOC）降低了近45%。

 （2） 通过溶剂热法制备Bi改性P25光催化剂 （Bi@P25） ，利用原位合成法将其负载于膨胀珍珠岩（EP）基轻质载体表面，制得自漂浮型光催化剂 （Bi@P25/EP） 。采用DFT计算分析了Bi@P25的结构和性质，通过TEM、XRD、XPS和UV-Vis等表征了Bi@P25/EP的物化性质。EP作为载体降低了Bi@P25的团聚，暴露出更多活性位点，生成了促进电荷转移的Bi-O-Si电子转移通道。当CIP初始浓度为10 mg/L时，Bi@P25/EP在模拟太阳光照120 min后，CIP去除率达98%，h⁺ 和 • O₂^-是最主要的活性物质。在模拟太阳光强度为1 kW/m²时，利用Bi@P25/EP作为光催化剂催化降解湿地水、湖水和污水厂出水中的低浓度CIP （5 mg/L），光催化降解效率分别为67.1%、75.2%和52.9%。

 （3） 采用常见树回归模型（包括决策树模型、装袋法模型、随机森林模型以及提升法模型）对Bi@P25/EP光催化降解CIP的效率进行了预测，并以均方根误差 （RMSE） 与拟合优度 （R²） 为评价指标筛选最优模型。结果显示随机森林模型的预测效果最好，泛化能力最强，对验证集数据的拟合优度为0.9045，可以较为精确的反映CIP降解过程中各影响因素与降解效率之间的关系，并对其进行了可视化，结果显示预测值与实验结果一致，说明该机器学习模型具有良好的应用能力。

The wastewater containing refractory organic pollutants poses a severe threat to the ecological environment and human health. Photocatalysis, characterized by its utilization of solar energy and features such as relatively low cost, no secondary pollution, small footprint, and high efficiency, is considered a promising technology for the deep treatment of refractory organic pollutants. However, common photocatalytic technologies currently face challenges in terms of limited contact area with incident light, additional energy consumption for stirring, and difficulties in material recovery, which hinder their practical application. By loading nano photocatalysts onto silico-aluminate support to prepare floating photocatalyst has emerged as a viable solution. This approach offers advantages such as low cost, environmental friendliness, no need for stirring, low energy consumption, easy recovery, and enhanced contact between the photocatalyst, incident light, and oxygen. Among various nano-photocatalysts, Bismuth-based nano-photocatalysts have shown great potential due to their unique layered structure and excellent visible light response capability.

Therefore, in this paper, geopolymer and expanded perlite were chosen as silico-aluminate-based photocatalyst carriers, with Bi-based nano-photocatalyst loaded.  The photocatalytic efficiency and reaction mechanism were investigated using methylene blue (MB), a common organic dye, and ciprofloxacin (CIP), an antibiotic frequently detected in surface water, as target pollutants. The main research findings are as follows:

(1) Geopolymer (MFGP) was prepared using metakaolin and common industrial solid waste fly ash as raw materials. BiOBr and conductive material acetylene black (AB) were loaded onto MFGP to prepare a bulk composite photocatalyst (BiOBr/AB/MFGP). The morphology, composition, and photoelectric properties of BiOBr/AB/MFGP were characterized using SEM, XRD, XPS, and UV-Vis. It was found that BiOBr and AB were uniformly fixed in MFGP without affecting its original structure. The 3D network of silico-aluminate structures in MFGP and the conductive network fabricated by AB provided pathways for photogenerated electrons transfer, thus inhibiting the photogenerated carriers’ recombination. In addition, the photo-thermal effect induced by AB promoted photocatalytic reactions at the micro-interface of the photocatalyst. When suspended at the water/air interface using a self-made copper mesh, BiOBr/AB/MFGP achieved a removal efficiency of 96% for 20 mg/L of MB after 30 minutes of simulated solar light irradiation, and a nearly 45% reduction in total organic carbon (TOC) within 40 minutes.

(2) Bi-modified P25 (Bi@P25) was prepared by solvothermal method, and it was then loaded onto the surface of expanded perlite (EP) to prepare the self-floating photocatalyst (Bi@P25/EP) through an in-situ synthesis approach. The structure and properties of Bi@P25 were analyzed by DFT, and the physical and chemical properties of Bi@P25/EP were characterized by TEM, XRD, XPS and UV-Vis. EP as the carrier reduced the aggregation of Bi@P25, exposed more active sites, and generated Bi-O-Si electron transfer channels that promote charge transfer. Bi@P25/EP achieved a removal rate of 98% for CIP at an initial concentration of 10 mg/L after 120 minutes of simulated solar light irradiation, and h⁺ and •O₂^- are the dominant active species. under a simulated solar light intensity of 1 kW/m², the photocatalytic degradation efficiency using Bi@P25/EP as a photocatalyst for the degradation of low concentration CIP (5 mg/L) in wetland water, lake water, and wastewater treatment plant effluent was found to be 67.1%, 75.2%, and 52.9% respectively.

(3) Common tree regression models, including decision tree model, Bagging model, random forest model and Boosting model, were employed to predict the degradation efficiency of CIP photocatalytic removal by Bi@P25/EP. The evaluation criteria for selecting the optimal model were root mean square error (RMSE) and goodness of fit (R²). The results showed that the random forest model had the best prediction effect and the strongest generalization ability. It achieved a good fit with the validation dataset, with an R² value of 0.9045. This model accurately reflected the relationship between various influencing factors and degradation efficiency in CIP photodegradation process, and it was visualized accordingly. The results showed consistency between the predicted values and experimental results, demonstrating its strong applicability.