The invention discloses an operation and maintenance detection cost prediction method for physical assets of a power grid based on big data, and the method comprises a multisource data collection platform, a data warehouse, a MapReduce model, and a data visualization display platform. An operation and maintenance detection cost prediction system for physical assets of the power grid employs the big data analysis technology, the ETL technology, the data warehouse technology and the data visualization technology, takes the whole-life-cycle of assets as the guiding thought, takes the asset unit information data in a PMS system and an ERP system and the operation and maintenance cost as the objects, builds a statistical relation of mass unstructured data through the analysis methods of clustering analysis, classification analysis and correlation analysis, and achieves the middle-term and long-term analysis prediction of the operation and maintenance cost scale and development trend of the physical assets of the power grid. The method achieves the high-efficiency integration of data of various types of business systems in the operation and maintenance management of the physical assets of the power grid, and improves the prediction accuracy of the operation and maintenance cost.