Productivity Prediction of Fractured Horizontal Well in Shale Gas Reservoirs with Machine Learning Algorithms

Abstract

Predicting shale gas production under different geological and fracturing conditions in the fractured shale gas reservoirs is the foundation of optimizing the fracturing parameters, which is crucial to effectively exploit shale gas. We present a multi-layer perceptron (MLP) network and a long short-term memory (LSTM) network to predict shale gas production, both of which can quickly and accurately forecast gas production. The prediction performances of the networks are comprehensively evaluated and compared. The results show that the MLP network can predict shale gas production by geological and fracturing reservoir parameters. The average relative error of the MLP neural network is 2.85%, and the maximum relative error is 12.9%, which can meet the demand of engineering shale gas productivity prediction. The LSTM network can predict shale gas production through historical production under the constraints of geological and fracturing reservoir parameters. The average relative error of the LSTM neural network is 0.68%, and the maximum relative error is 3.08%, which can reliably predict shale gas production. There is a slight deviation between the predicted results of the MLP model and the true values in the first 10 days. This is because the daily production decreases rapidly during the early production stage, and the production data change greatly. The largest relative errors of LSTM in this work on the 10th, 100th, and 1000th day are 0.95%, 0.73%, and 1.85%, respectively, which are far lower than the relative errors of the MLP predictions. The research results can provide a fast and effective mean for shale gas productivity prediction.

A Physics-Informed Neural Network Approach for Surrogating a Numerical Simulation of Fractured Horizontal Well Production Prediction

“…These proxy models have been used to predict gas production and to obtain optimal fracture half-lengths and horizontal well lengths (Wang et al, 2021) [1]. Multi-layer perceptron (MLP) can be used to construct surrogate models to predict the production of gas wells with parallel, equal-length hydraulic fractures on a two-dimensional plane (Wang et al, 2021) [2]. The tree-based ensemble method was used as a proxy model of a numerical simulator of 2D discrete fracture networks.…”

“…However, the input variables of the proxy models mentioned in these studies are often oversimplified. For example, among the previously mentioned proxy models, CNN, GPR, and SVM (Wang et al, 2021) [1] require that the length of the hydraulically fractured fractures in the dataset take only a fixed number of four values; the training set of the MLP network (Wang et al, 2021) [2] contains only four variables, and these four features can only take a fixed number of four values; the tree-based ensemble model (Xue et al, 2019) [3] requires that the dataset can only have these four variables, which can only take a fixed number of three values; and the input data of the transformer (Wang et al [5]) need to be preprocessed by PCA (principal component analysis) to decrease its dimension, but the variables generated by PCA often pose challenges in terms of interpretation. Although reducing the dimensions of the input data or simplifying the input data can reduce the complexity of machine learning models and make training easier, this also decreases the performance of the proxy models.…”

Optimization of Fracturing Parameters by Modified Genetic Algorithm in Shale Gas Reservoir

“…Data-driven models, established by machine learning methods, can meet the requirements of calculation speed and accuracy for fracturing parameters optimization. Wang et al [19] predicted shale gas production with a Multi-layer Perceptron (MLP) network and Long Short-term Memory (LSTM) network. Xue et al [20] compared the performance of Multiobjective Random Forest (MORF) and Multi-output Regression Chain (MORC) in shale gas production prediction.…”

Research on Productivity Evaluation Method Based on Shale Gas Well Early Flow Stage Data

“…However, due to the many idealized assumptions in the analytical model, the analytical method cannot accurately reflect the complex flow mechanisms and development laws of shale gas in the actual formation [56][57][58]. With the continuous integration of the artificial intelligence with oil and gas field production sites, machine learning methods are gradually emphasized in the productivity evaluation of shale gas wells [59][60][61][62].…”