Prediction of drilling leakage locations based on optimized neural networks and the standard random forest method

Abstract

Circulation loss is one of the most serious and complex hindrances for normal and safe drilling operations. Detecting the layer at which the circulation loss has occurred is important for formulating technical measures related to leakage prevention and plugging and reducing the wastage because of circulation loss as much as possible. Unfortunately, because of the lack of a general method for predicting the potential location of circulation loss during drilling, most current procedures depend on the plugging test. Therefore, the aim of this study was to use an Artificial Intelligence (AI)-based method to screen and process the historical data of 240 wells and 1029 original well loss cases in a localized area of southwestern China and to perform data mining. Using comparative analysis involving the Genetic Algorithm-Back Propagation (GA-BP) neural network and random forest optimization algorithms, we proposed an efficient real-time model for predicting leakage layer locations. For this purpose, data processing and correlation analysis were first performed using existing data to improve the effects of data mining. The well history data was then divided into training and testing sets in a 3:1 ratio. The parameter values of the BP were then corrected as per the network training error, resulting in the final output of a prediction value with a globally optimal solution. The standard random forest model is a particularly capable model that can deal with high-dimensional data without feature selection. To evaluate and confirm the generated model, the model is applied to eight oil wells in a well site in southwestern China. Empirical results demonstrate that the proposed method can satisfy the requirements of actual application to drilling and plugging operations and is able to accurately predict the locations of leakage layers.

A New Method for Determination of Optimal Borehole Drilling Location Considering Drilling Cost Minimization and Sustainable Groundwater Management

“…Ensemble learning techniques are also widely adopted across wide application areas. 24 Article ( 25 ) proposed an optimized NN based solution for predicting circulation loss. The performance of ensemble models can be further improved using an optimal weight assignment scheme.…”

A Model-Based Intelligent Adjustment Method of Toolface for Bent-Housing Motor

“…The artificial intelligence method can fit the relationship between the physical model results and the actual data in the drilling site well. A BP (back propagation) neural network, based on a GA (genetic algorithm), has a good fitting effect on nonlinear functions affected by multiple parameters, has been widely used in the drilling field, and has achieved satisfactory results in recent years, such as build-up rate prediction [11], crude oil output decline rate prediction [12], crude oil production prediction [13], overflow and leakage prediction [14][15][16], bottom hole pressure prediction [17], horizontal in situ stress and natural fracture property identification [18], oil saturation prediction [19], drilling tool structure optimization [20], optimal ROP calculation, and drilling parameter optimization [21]. The downhole toolface change value during the adjustment of the bent-housing motor toolface can be regarded as a nonlinear function affected by multiple parameters, and its influencing factors at least include well trajectory, wellbore structure, drilling assembly, drilling fluid property, bit performance, formation characteristics, the interaction between bit and formation, bent-housing motor performance, and drilling engineering parameters.…”

Well log data super-resolution based on locally linear embedding

“…Therefore, in order to enhance the production capacity of the unconventional remaining oil and gas resources, the reservoir modelling resolution should be as high as possible. Meanwhile, the high-resolution logging curves can better provide the possibility to predict the location of lost circulation [2], lithology identifies [3], and predict the productivity of heterogeneous reservoirs [4,5]. In practice, because the vertical resolution of seismic data is too low and the cost of acquisition of core and other high resolution logging data is too high, it is an inevitable choice to employ conventional well log data to enhance the vertical resolution of the final reservoir models.…”