New approach for prediction of asphaltene precipitation due to natural depletion by using evolutionary algorithm concept

► A neural network combine with particle swarm optimization has been presented. ► PSO–ANN model combines local and global searching ability of the ANN and PSO, respectively. ► It has improved the fitting between PSO–ANN prediction of the model and the experimental data. ► The PSO parameters have been carefully designed to optimize the ANN, avoiding premature convergence. Asphaltene precipitation affects enhanced oil recovery processes through the mechanism of wettability alteration and blockage. Asphaltene precipitation is very sensitive to the reservoir conditions and fluid properties, such as pressure, temperature and injected fluid molecular weight. In this work, the model based on a feed-forward artificial neural network (ANN) optimized by particle swarm optimization (PSO) as an artificial intelligence modeling tool to predict asphaltene precipitation due natural depletion. Particle swarm optimization (PSO) is used to decide the initial weights of the neural network. The PSO–ANN model is applied to the experimental data from one of northern Persian Gulf oil field has been used to develop this model. The predicted results from the PSO–ANN model and BP–ANN were compared to the experimental precipitation data. The average relative absolute deviation between the model predictions and the experimental data was found to be less than 4%. A comparison between the prediction of this model and the alternatives showed that the PSO–ANN model predicts asphaltene precipitation more accurately.