基于CNN-LSTM算法的气井产量预测研究

Predicting Gas Well Production Based on CNN-SLTM Algorithm

气井产量预测对合理评价气井产能和制定合理的排采制度具有重要意义。基于经验模型的产量预测方式,在使用条件和环境上具有较大的局限性。论文提出一种基于CNN和LSTM的融合算法,从数据角度出发,预测气井产量。通过CNN算法提取数据空间特征,用LSTM算法提取数据的时间特征,同时,基于机理模型分析气井产量与生产参数的关系,对特征参数进行预处理,提高算法的准确率。实验结果表明,与传统的CNN算法、LSTM算法相比,具有较好的预测效果,预测日产气量与实际日产气量之间误差小于5%。

The prediction of gas well production is of great significance to evaluating production capacity of gas well and establishing reasonable drainage and production system.The method of yield prediction based on empirical model has great limitations in use conditions and environment.In this paper,a fusion algorithm based on CNN and LSTM is proposed to predict gas well production from the perspective of data.The CNN algorithm is used to extract spatial features of data,the LSTM algorithm is used to extract time features of data.Meanwhile,the relationship between gas well production and production parameters is analyzed based on the mechanism model,and the characteristic parameters are preprocessed to improve the accuracy of the algorithm.Extensive experimental results are presented to show that compared with the traditional CNN algorithm and LSTM algorithm,the performance of the proposed algorithm achieves better prediction performance on the data,and the error between the actual daily production and the predicted daily production is less than 5%.