Long-term and short-term classified prediction on well production based on grey model and time series

Abstract

The calculation methods for predicting well production in pilot production and stable production stages are disadvantageous for insufficient precision, low efficiency, and small application scope. In order to solve these problems, a method that divides the prediction model into short-term production and long-term production was proposed. For short-term production prediction, based on the grey prediction model GM (1,1), the parameter correction on production conversion is added depending on the changes in short-term production environment, and the discrete process of the model is optimized, which makes up for the deviation of the gray derivative discrete process. For long-term production prediction, based on time series, and combing with BP neural network and the extraction results of short-term sampling eigenvalues, the cumulative production data were modeled and predicted. Using the production data of multiple wells in 2018~2020, model training, prediction and comparison were performed, and the accuracy of the results predicted by the model was analyzed. The results show that the average relative error of the prediction results of the long-term prediction model is 0. 091 4, and the average absolute error of the prediction results of the short-term prediction model is 0. 118 7. The calculation accuracy is higher than the average relative error of 0. 15 66 calculated by traditional ARIMA algorithm, which can meet the needs of actual working conditions, providing a new method for predicting of well production. © 2022 Well Testing. All rights reserved.