Atomized droplet size prediction for supersonic atomized water drainage and natural gas extraction

Abstract

In the late stage of natural gas reservoir exploitation, the energy of the gas reservoir decreases sharply, the wellbore pressure decreases, the liquid accumulation is serious, and the natural gas production decreases sharply. The influence of gas working parameters on the atomization effect and liquid carrying rate was established, and the physical model of the new atomizing nozzle was established. Through laboratory experiments, the experimental parameters of gas flow rate, liquid phase flow rate, liquid phase inlet diameter and average droplet diameter SMD were obtained. Then the BP neural network atomization model optimized by genetic algorithm is established, and the Matlab is used to train the 45 groups of data sets before the experiment. After the model training, the normalized atomization parameters are trained for sensitivity analysis. The relationship between the strength and strength of the factors affecting SMD is as follows: gas flow > liquid inlet diameter > liquid phase flow. The last 15 sets of data sets outside the training samples were tested by BP and GA-BP model, and the size of SMD was predicted. The experimental results show that the determination coefficient R2 of the established GA-BP network model to the experimental parameters is 0.979 and the goodness of fit is high; the MSE, MAE and MAPE of the predicted value of GA-BP atomization model and the experimental value are 4.471, 1.811 and 0.031 respectively, the error is small, the prediction accuracy is high, and the establishment of the model is accurate. The GA-BP model plays an important role in setting the working condition parameters and predicting the droplet size, and has a certain guiding significance for solving the problem of wellbore fluid accumulation and improving the drainage efficiency.