Prediction of drilling plug operation parameters based on incremental learning and CNN-LSTM

Abstract

In order to solve the problems of low prediction accuracy and insufficient use of long time series data, and ensure the safety of drilling plug operation, a method for predicting drilling plug parameters based on incremental learning and the CNN-LSTM model is proposed. By making full use of the measured data of the drilling plug operation and using the sliding window processing method to process the data, then the CNN-LSTM deep learning model is established to predict the downhole parameters, and the reliability and generalization of the model are verified. The results show that the prediction accuracy of the machine learning model is the highest when the fixed width k = 4. According to the error results, it is concluded that CNN-LSTM model has the best processing and forecasting ability for drilling plug operation. By analyzing the prediction results of the whole well data, the validation accuracy of CNN-LSTM model reaches 99.5%, indicating that the model has high accuracy and generalization, and the model has fast calculation speed, providing a prediction method with high precision and fast calculation speed for drilling plug operation.