The invention of novel light-harvesting materials is one of the primary reasons behind the acceleration of current scientific advancement and technological innovation in the solar sector. Organometal halide perovskite (OHP) has recently attracted a great deal of interest because of the high-energy conversion efficiency that has reached within a few years of its discovery and development. Modern machine learning (ML) technology is quickly advancing in a variety of fields, providing blueprints for the discovery and rational design of new and improved material properties. In this paper, we apply ML to optimize the material composition of OHPs, propose design methods and forecast their performance. Our ML model is built using 285 datasets that were taken from about 700 experimental articles. We have developed two different ML models to predict the bandgap and performance parameters of solar cell. In the first model, we employed three ML algorithms to investigate the relationship between bandgap and perovskite material composition. We estimated the performance characteristics using projected and actual bandgap. Second, ML models are used to predict the performance parameters employing the bandgap of perovskite and energy difference between electron transport layer (ETL) and hole transport layer (HTL) with perovskite as an input parameter. Simulation results suggest that the artificial neural network (ANN) technique, which predicts the bandgap by taking into consideration how cations and halide ions interact with one another, demonstrates a better degree of accuracy (with a Pearson coefficient of 0.91 and root mean square error of 0.059). The constructed ML model closely fits the theoretical prediction made by Shockley and Queisser, and that is almost hard for a person to discover from an aggregation of datasets.

新型光收集材料的发明是当前太阳能领域科学进步和技术创新加速的主要原因之一。有机金属卤化物钙钛矿（OHP）最近引起了人们的极大兴趣，因为它在发现和开发的几年内就达到了高能量转换效率。现代机器学习 (ML) 技术正在各个领域快速发展，为发现和合理设计新的和改进的材料特性提供了蓝图。在本文中，我们应用 ML 来优化 OHP 的材料成分，提出设计方法并预测其性能。我们的 ML 模型是使用从大约 700 篇实验文章中提取的 285 个数据集构建的。我们开发了两种不同的 ML 模型来预测太阳能电池的带隙和性能参数。在第一个模型中，我们采用了三种 ML 算法来研究带隙和钙钛矿材料成分之间的关​​系。我们使用预计和实际带隙来估计性能特征。其次，ML 模型用于预测性能参数，采用钙钛矿的带隙和电子传输层 (ETL) 与空穴传输层 (HTL) 之间的能量差，钙钛矿作为输入参数。模拟结果表明，通过考虑阳离子和卤化物离子如何相互作用来预测带隙的人工神经网络 (ANN) 技术表现出更高的准确度（Pearson 系数为 0.91，均方根误差为0。059). 构建的 ML 模型与 Shockley 和 Queisser 所做的理论预测非常吻合，这对于一个人来说几乎很难从数据集的聚合中发现。