The Artificial Neural Network Method of Well-Test Interpretation Model Identification and Parameter Estimation

Abstract

Abstract A correct well-test interpretation model is the basement of well-test analysis, which is identified by the feature of pressure derivative curves. Automatic identification of well-test interpretation model is very important for automatic parameter estimation from the well test data. Based on the excellent performance of classification and the feature of noise insensitive, the artificial neural network (ANN) has been studied extensively in identification of well-test interpretation model. Whether the trained ANN can identify the interpretation model of the actual tested curve correctly depends on the ANN construction method. Using the derivative curve's data point series as input vector to train ANN can't ensure the actual tested curve in the same model area as it training samples. So the sample curves data point series trained ANN usually misidentify the interpretation model of the actual tested curve. Based on the pressure derivative curve's common features among different interpretation models, an ANN approach is proposed to realize the type curves matching and move the tested curves to its sample position. Using curve's binary numbers instead of data points as input vector to train ANN can guarantee the actual tested curve in the same model area as its training samples, this ensure the successfully trained ANN can identify the interpretation model of the actual tested curve correctly. The examples showing that the proposed method is more efficiently than the data point series method. In the process of interpretation model identification, the parameters can also be estimated.