Abstract: To evaluate the reservoir quality and fluid seepage law of reservoir ZJ452 that is strongly heterogeneous in the Zhujiang Formation in Huizhou A Oilfield in the Pearl River Mouth Basin, reservoir architecture analysis was carried out and multi-level automatic clustering algorithm were used to classify and evaluate the strongly heterogeneous reservoir. Porosity, permeability, flow zone index, and reservoir quality parameters were selected for discrimination. Self-organizing mapping neural network and Bayesian discriminant flow unit intelligent classification technology were applied, from which three types of flow units in reservoir ZJ452 were recognized: Class I, Class II, and Class III. In addition, volume calculation was perform with flow unit discrimination formulas, a three-dimensional flow unit model was established, and the spatial distribution characteristics of flow units was analyzed. Results show that although the reservoir ZJ452 was strongly heterogeneity in space; Class I and Class II flow units were connected in a continuous distribution on the plane. This study pointed out the direction for tapping the potential of the highly heterogeneous reservoir ZJ452 in Huizhou A Oilfield.