Fault diagnosis in power grids is essential for ensuring uninterrupted and reliable electricity supply. Traditional approaches rely on expert systems, fuzzy logic, and machine learning, but they often underutilise the quantitative aspects of alarm information generated by monitoring systems. This paper proposes a novel fault diagnosis framework that integrates quantitative alarm data with machine learning techniques to enhance the accuracy and efficiency of fault identification. The proposed approach demonstrates improved diagnostic performance, offering greater reliability for modern power systems.