The massive amount and multi-sourced, multi-structured data in the upstream petroleum industry impose great challenge on data integration and smart application. Knowledge graph, as an emerging technology, can potentially provide a way to tackle the challenges associated with oil and gas big data. This paper proposes an engineering-based method that can improve upon traditional natural language processing to construct the domain knowledge graph based on a petroleum exploration and development ontology. The exploration and development knowledge graph is constructed by assembling Sinopec’s multi-sourced heterogeneous database, and millions of nodes. The two applications based on the constructed knowledge graph are developed and validated for effectiveness and advantages in providing better knowledge services for the oil and gas industry.