Integrating linear and nonlinear forecasting techniques based on grey theory and artificial intelligence to forecast shale gas monthly production in Pennsylvania and Texas of the United States

Abstract

Pennsylvania and Texas accounted for about 60% of U.S. total shale gas production. Better forecasting shale gas production in Pennsylvania and Texas can serve us to better predict U.S. shale gas production. In this work, we integrate the linear and nonlinear forecasting techniques in order to use the advantages and avoid the disadvantages of linear and nonlinear forecasting models, so as to improve forecasting accuracy. Specifically, we develop two hybrid forecasting techniques, i.e., nonlinear metabolic grey model–Autoregressive Integrated Moving Average Model (NMGM-ARIMA), and Autoregressive Integrated Moving Average Model - Artificial neural network (ARIMA-ANN). 60 samples (monthly shale gas production in Pennsylvania and Texas) are used to test these two proposed forecasting techniques and these existing single nonlinear (NMGM, and ANN) and linear (ARIMA) forecasting techniques. The results show that for samples from either Pennsylvania or Texas, the mean absolute percent error of NMGM-ARIMA (3.16%, 1.64%) is smaller than that of NMGM (4.31%, 2.98%) and ARIMA (3.53%, 2.03%), and that of ARIMA-ANN (2.06%, 1.38%) is also smaller than ARIMA (3.53%, 2.03%) and ANN (3.09%, 1.71%). The proposed hybrid NMGM-ARIMA and ARIMA-ANN can achieve more accurate forecasting effect than the single theory-based models that made them up, and can be used in forecasting other fuels. The forecasting results show growth rates of shale gas production in Pennsylvania is higher than Texas in 2017 and 2018.