Emergencies in the petroleum and petrochemical industry often occur under uncertain and complex conditions, and the scarcity of reliable data limits the usefulness of conventional decision-making tools. In this work, an improved Case-Based Reasoning (CBR) method is developed using structured descriptions of incident information. The method builds an emergency case library, brings together heterogeneous data, and evaluates similarity through a weighting scheme that combines expert input with objective measures. A dynamic adjustment step is introduced to account for shifts in on-site conditions, and the K-Nearest Neighbor (KNN) algorithm is applied to narrow down the recommended response plans. Tests on 17 emergency cases indicate that the method can maintain stable performance in demanding scenarios. These results suggest that the approach offers a practical option for supporting emergency decision-making in the petroleum and petrochemical field.