Data Management Method in the Petroleum Exploration Work Area

Abstract

Petroleum exploration is an industry that generates a large amount of data, but the datasets used are highly correlated and complex to process. To achieve intelligent management of petroleum data, we propose a multi-model framework based on deep learning networks. This framework combines the advantages of Multilayer Perceptron (MLP), Convolutional Neural Networks (CNN), and Long Short-Term Memory (LSTM) to identify hot data that are more likely to be accessed by voting. In addition, we compare the performance of three commonly used time series prediction models for spatial prediction of petroleum exploration work areas. Experiments show that the multi-model framework outperforms traditional solutions by 25.3% and exhibits a 7.0% performance improvement compared to the best-performing LSTM model in a single model. LSTM is more suitable than Least Squares Regression (LSR) and Support Vector Regression (SVR) for spatial prediction of petroleum data, and a simple offset processing of the prediction results can cover more than 90% of real scenarios.