Among various solar cell architectures, dye-sensitized solar cells (DSSCs) and perovskite solar cells have demonstrated the capability of being bifacial as both can be fabricated on conducting glass electrodes. In both cells, TiO2 plays a key role in the optoelectronic properties of the cell. Various studies have reported a range of recipes and deposition techniques for TiO2 thin films. Such variety introduces some uncertainties into the optical properties of the prepared films as well as in the process repeatability. Here, we utilized machine learning methods to correlate the film porosity to the film refractive index, making it capable of studying the impact of varying the fabrication and deposition techniques. Image postprocessing for scanning electron microscope measurements was utilized to estimate the film porosity, and the refractive index was calculated from the T–λ spectra. Four sets of samples with complete bifacial DSSCs were fabricated and characterized. They recorded a maximum current of 23.42 mA. They were fabricated using carboxymethyl cellulose-based suspension and deposited via the spin-coating sol-gel method. The fabricated cells showed an overall conversion efficiency of 7.9% under optical injection of the AM1.5G spectrum from the front side and LED indoor lighting from the counter electrode.

中文翻译：

---

使用机器学习算法研究用于双面太阳能电池的 TiO2 薄膜光学特性的变化

在各种太阳能电池结构中，染料敏化太阳能电池（DSSCs）和钙钛矿太阳能电池已经证明了双面的能力，因为两者都可以在导电玻璃电极上制造。在这两种电池中，二氧化钛在电池的光电特性中起关键作用。各种研究报告了一系列用于 TiO2 薄膜的配方和沉积技术。这种变化给制备的薄膜的光学特性以及工艺可重复性带来了一些不确定性。在这里，我们利用机器学习方法将薄膜孔隙率与薄膜折射率相关联，使其能够研究改变制造和沉积技术的影响。用于扫描电子显微镜测量的图像后处理用于估计薄膜孔隙率，并且折射率由 T-λ 光谱计算。制造并表征了四组具有完整双面 DSSC 的样品。他们记录的最大电流为 23.42 mA。它们是使用基于羧甲基纤维素的悬浮液制造的，并通过旋涂溶胶-凝胶法沉积。在从正面光学注入 AM1.5G 光谱和从对电极进行 LED 室内照明的情况下，制造的电池的整体转换效率为 7.9%。