Dongfang B Gas Field in Yinggehai Basin is a shallow sea gravity flow submarine fan deposit, which sand body distribution and superposition relationship are complex. The interpretation of sedimentary microfacies by manual identification is cumbersome and easy to be affected by subjective factors. A sedimentary microfacies recognition model with serial network architecture is designed based on bi-directional long short-term recurrent neural network. Taking logging data and lithology logging data as input, this model can effectively extract the morphological characteristics of logging curves of different sedimentary microfacies, and fully consider the correlation between adjacent sedimentary microfacies. The model is applied to the sedimentary microfacies identification in this area, which effectively reduces the influence of reservoir heterogeneity and artificial experienc，and improve the identification accuracy and achieves good application results.