An Intelligent Prediction Method of Horizontal Well ROP Based on Di-GRU

Abstract

ABSTRACT At present, with the continuous development of unconventional oil and gas exploration and development, horizontal wells, as an important means of unconventional oil and gas exploitation, have attracted much attention from drilling teams and oil field companies. However, the cost of horizontal well drilling is relatively high, so more targeted construction should be carried out according to the characteristics of horizontal well drilling. Therefore, rate of penetration (ROP) prediction of horizontal wells is extremely important. The current ROP prediction models of horizontal wells mainly rely on manual experience and some mechanism models, and do not take field data and actual construction conditions into account, so the field application effect is difficult to be ideal. To solve this problem, an intelligent algorithm based ROP prediction model for horizontal Wells is proposed and a novel dual-input neural network structure is designed. Sequential data such as hook load and rotary torque are input into Gated recurrent unit’ (GRU) network, and non-sequential data such as bit type are input into back-propagation neural network(BP), which is called Dual Input GRU(Di-GRU) ROP prediction model. The results show that mean absolute percentage error and R-squared of this model are 11.8% and 0.923 respectively. The prediction model based on Di-GRU proposed in this paper has higher accuracy and better mobility, which is of guiding significance for improving the drilling efficiency of horizontal wells. INTRODUTION With the deepening of oilfield exploration and development, the development of deep and complex oil and gas reservoirs has become the focus of stabilizing oil and increasing production, which has important strategic significance. At present, horizontal wells have become the main well type for developing complex oil and gas reservoirs(Yassie et al.,2020). Compared with vertical wells, horizontal wells usually have more complex geological conditions and greater drill string friction, resulting in poor formation drill ability and difficulty in drilling the drill string. The low rate of penetration(ROP) and the small footage of a single bit bring severe challenges to drilling engineering and seriously affect the progress of exploration and development(Soares et al.,2019). Accurate ROP prediction of horizontal wells can assist the scientific allocation of resources before drilling and the accurate formulation of operation plans, which is of great significance for improving drilling efficiency and reducing costs. Horizontal well drilling engineering operation is a complex process. Formation, tools and operating parameters all play an important role in the rock breaking of the drill bit, such as formation drill ability, drill bit wear, rotary speed, mud performance, drill bit pressure, build-up rate, etc. Therefore, it is difficult toaccurately predict the ROP of horizontal wells.