Sha-2 Gas Reservoir Field Fracturing Horizontal Wells Parameter Optimization And Yield Prediction

Xinchang gas field is located in the fertile Chengdu Plain in Sichuan Province, which is in the northwest of Deyang City and in the southeastern of Mianyang City, and it covers approximately an area of 208km2. Shaximiao gas reservoir belongs to the Xinchang gas field. It can be futther divided into three sub gas reservoirs: JS1, JS2, JS3. The JS2 can be divided into four groups as JS21?JS22?JS23?JS24. JS21. The JS22 is the main gas layer,and has been in the late development stage with little development potential. JS23 and JS24 are difficult to develop, and are the focus of this paper.Xinchang gas field is a typical tight sandstone gas reservoirs.Tight sandstone gas reservoir is one of the potential resource for clean energy. Due to the extremely low permeability and high penetration resistance, the production of vertical wells in tight sandstone gas reservoirs was poor, and the production of horizontal wells is also low. In order to further stimulate the horizontal wells, hydraulic fracturing is generally applied to open multiple channels to increase oil and gas flow.In this paper, the geology, structural characteristics and reservoir properties of Xinchang gas field are collected from the on-site fractured wells.Based on scientific mathematical statistical methods, Univariate Analysis methods and Gray Association Analysis method were combined to determine which main factors will affect the fracturing effect, including the permeability, porosity, effective thickness, sand ratio and displacement and so on. On the basis of the optimization of fracturing parameters, and through orthogonal experiments, the best combination of artificial fractures was determined. According to field data, multiple linear regression analysis method and BP neural network method respectively were applied to establish prediction model, and to predict production capacity after stimulation. Error analysis was also made when actual production capacity was also compared. The results show that BP neural network method in the prediction of pressure and gas production is more efficient. According to the field data, the MATLAB software was used to establish the prediction software of the neural network structure which is suitable for the geological situation of the gas field.