涪陵焦石坝页岩气井调产分析及预测

Analysis and Prediction of Yield Adjustment of Shale Gas Well in Jiaoling Dam of Fuling

为了研究涪陵焦石坝区块页岩气井产量变动和压力预测,依据该区块典型页岩气井生产数据,采用模糊C均值聚类方法对生产数据进行自适应聚类,保证生产数据的准确性;再运用Kendall秩相关系数分析法对生产数据进行相关性分析,以确定用于递归神经网络建模的输入变量;最后分别对每一类数据进行建模,形成多模型库。预测时输入需要预测的产量以及历史数据,系统将自动识别并调用其对应的模型进行压力预测,实现不同产量下的压力预测。以涪陵焦石坝页岩气田某气井实验表明,相比传统的BP神经网络和曲线拟合预测,所提方法能够有效提高预测的准确率和稳定性。

In order to study the yield variation and pressure prediction of shale gas wells in the Jiaoshiba block of Fuling,based on the typical shale gas well production data of the block,fuzzy C-means clustering method is used to adaptively cluster the production data to ensure Accura.cy of production data;Kendall rank correlation coefficient analysis method is used to analyze the correlation of production data to deter.mine the input variables used for recurrent neural network modeling;finally,each type of data is modeled to form a multi-model library.When predicting the output and historical data that need to be predicted,the system will automatically identify and call its corresponding model for pressure prediction to achieve pressure prediction under different production rates.A gas well experiment in the shale gas field of the Jiaoling dam in Fuling shows that compared with the traditional BP neural network and curve fitting prediction,the proposed method can effectively improve the accuracy and stability of prediction.