Organic solar cells are a new type of photovoltaic material with many advantages, such as low cost, environmental friendliness, large-scale production, and flexibility in device fabrication. In order to improve the efficiency and stability of organic solar cells, scientists have developed many different types of organic semiconductor materials. However, finding high-performance organic semiconductor materials remains a key challenge in organic solar cell research. With the development of big data and artificial intelligence technology, data-driven materials research has become a new trend in materials science. Combining large amounts of data generated by high-throughput screening with artificial intelligence technology for materials performance prediction and optimization helps accelerate the research process of organic solar cell materials. This study proposes a method for high-throughput screening of organic solar cells using artificial intelligence, combining deep learning and ensemble learning. Starting from molecular structure, the molecular physicochemical properties predicted by graph neural networks are used as inputs for the ensemble learning model, achieving accurate and rapid prediction of device properties. The team of Professor Li Youyong from the Functional Nanomaterials and Soft Matter Research Institute of Suzhou University, in collaboration with Professor Yuan Jianyu's team, has achieved high-throughput screening of organic solar cells through machine learning. This study uses a small dataset containing high-quality experimental data to train an ensemble learning model that predicts the photoelectric conversion efficiency (PCE) based on molecular physicochemical properties. Then, they train a Self-Learning Input Graph Neural Network (SLI-GNN) capable of accurately predicting molecular properties using a dataset containing a large number of molecular structures and properties. By combining these two models, a framework that can directly predict PCE based on molecular structure is designed. This framework demonstrates excellent performance and versatility in high-throughput screening and has been validated by experimental results. By integrating deep learning and ensemble learning, direct, fast, and accurate prediction of PCE based on molecular structure is achieved. This method not only improves the efficiency of organic optoelectronic material screening but also reduces computational costs, providing strong support for the design and optimization of optoelectronic devices.