Long short-term memory model for predicting productivity of drilling space units

In the oil and gas industry, there is a growing demand for application of big data analytics and artificial intelligence (AI) technologies to optimize operations and reduce cost. In this study, we work on the productivity prediction, which is an important and challenging task for operators. Unlike previous studies where full field or single well analysis was conducted, we focus on more active operation units, drilling space units. Moreover, significant information is extracted from geology reports which are saved in scanned PDF files and well logs. By using the extracted information, long short-term memory (LSTM) model which could take the special-temporal changes of DSUs into consideration is employed to predict the DSUs' productivity. After rigorous validation, it is found that the accuracy of LSTM model could reach to more than 60%, which is 10% higher than a multilayer perceptron (MLP) model proposed in a previous research.