关键词: 页岩气压裂, 工况预测, 粒子滤波(PF), 自回归滑动平均(ARMA)模型, 局部加权线性回归(LWLR)

Abstract: In order to realize the prediction of downhole conditions of shale gas fracturing, prevent and control the abnormal conditions in time, a method of building optimized LWLR algorithm based on PF and ARMA model can worked out. The method uses the ARMA model and PF to build a PF_ARMA model and the PF_ARMA model can be used to predict pressure, and the prediction results can be used as the optimization basis for the LWLR model. Finally, LWLR model of optimal pressure parameter was obtained. And a comparison was made between the prediction result by optimized LWLR model and that by the traditional model. The model was used to analyze a shale gas fracturing operation curve. The result shows that the prediction accuracy by the optimized LWLR model is higher than that by any traditional model, and that the change trend and amplitude of the data could be described more accurately by the optimized LWLR model than any traditional model.

Key words: shale gas fracturing, condition prediction, particle filter(PF), auto-regressive and moving average(ARMA) model, locally weighted linear regression(LWLR)