A hybrid AVOA ‐ BP neural network with SHAP ‐based interpretation for wax deposition rate prediction

Wax deposition is an important issue affecting the efficiency and safety of crude oil transportation. Its formation rate is influenced by multiple operating parameters and exhibits strong nonlinear characteristics. Accurate prediction of wax deposition rate is therefore important for developing operating strategies and ensuring the stable operation of pipelines. In this study, a crude oil from a producing oilfield was investigated. Based on a relatively limited dataset obtained from loop experiments, a hybrid African vulture optimization algorithm (AVOA)‐backpropagation neural network model was developed for wax deposition rate prediction. The model achieved good predictive performance, with a test‐set coefficient of determination of 0.97 and a mean absolute error of 0.004 for the normalized wax deposition rate. SHAP analysis was used to quantify the contributions of input variables to model predictions. Wall temperature, oil temperature, and temperature gradient were identified as the most influential variables, indicating physically plausible trends in wax deposition behaviour. The proposed method provides support for wax deposition risk assessment under the investigated pipeline operating conditions.