Oil production prediction method based on LSTM

The invention provides an oil production prediction method based on LSTM. The oil production prediction method based on LSTM comprises the following steps: 1. Adopting Pearson correlation analysis toweigh the linear similarity of oil field data so as to screen out suitable input variables; 2, carrying out gradation standardization treatment on each variable; 3, adopting a neural network model ofdouble LSTM layers to carry out network train; Step 4, adopting the trained model to predict the annual oil production of a new well. According to the standardization process, the predicted variablesare reversely restored and compared with the actual values. The oil production prediction method based on LSTM has a wide prediction range and good fitting effect, and the prediction accuracy is over95%. It is very important to predict oil production more accurately and effectively for oilfield development decision and production investment. 本发明提供种基于LSTM的油田产油量预测方法，该基于LSTM的油田产油量预测方法包括：步骤1，采用Pearson相关性分析来衡量油田数据的线性相似度，以筛选出合适的输