Semi-Iterative Well-Test-Matching Method Uses Featured Points To Increase Accuracy, Efficiency

_ This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 221287, “A New Semi-Iterative Well-Test-Matching Method Based on Featured Points for Homogeneous Reservoirs,” by Xue Guo, SPE, Zhiming Chen, SPE, and Haifeng Zhu, China University of Petroleum, et al. The paper has not been peer reviewed. _ The challenge of balancing inversion accuracy and speed is a well-known one in well-test analysis. Two homogeneous well-test models with single- and dual-porosity media are first presented for the study. Then, a new semi-iterative model (SIM) is used for a well-test curve-matching method based on featured points of these two well-test models. This study also compares the performance of the SIM, the formula-based analytical model (FAM), and the full-iterative model (FIM). It is found that the SIM can be more efficient and accurate in inverting the parameters of homogeneous reservoirs. Methodology Errors in curve matching can lead to nonunique well-test interpretations. The points or segments featured on the double logarithmic curve of pressure and pressure derivatives, known as featured points (Fp) or featured segments (Fs), have a greater influence on curve-matching results. Setting up Fp or Fs based on the target reservoir curve can produce more-accurate results than relying solely on engineers’ experience to filter alternative results. Analytical and iterative methods are used to address each feature point. The accuracy of these methods varies depending on the specific feature point. Therefore, this section of the complete paper presents a semi-iterative well-test curve-matching method that combines analytical and iterative approaches based on Fp and Fs. The analytical solution formulas for the storage-capacity ratio and interporosity flow coefficient also have been improved. The model enabled more efficient and accurate inversion of reservoir parameters. Additionally, a semi-iterative parametric automatic inversion model is introduced. The model takes processed dimensionless time, pressure, and pressure-derivative data as inputs. First, two targeted well-test models are introduced. These theoretical models are used not only to generate the curve to be measured at the input but also as an interface to the parametric automatic inversion model to self-generate the target curve. Three parametric inversion models are presented, comprising the FAM and the FIM as a control group, and a SIM as an experimental group for controlled experiments. The theoretical well-test models and parametric-inversion models were integrated into a parametric automatic inversion model and applied to two targeted well-test models. Fig. 1 shows the technical route of this paper. Results and Discussion Integrity Diagnostic and Classification Results. This section of the complete paper evaluates data integrity using 10,000 processed random samples and assesses the prediction results for complete data. Fifteen hundred curves were randomly selected based on running moments and processed with truncation of 5, 10, 20, 60, and 80% of the end data. The diagnosis of data integrity was only 26.53% accurate. This was because the model was fault-tolerant when a small amount of data was missing. The false positive can be changed to invert the reservoir parameters based on the remaining data. Despite the fact that 17 sets of dual-porosity media proposed steady-state stratigraphic well-test curves were deemed to be single-media reservoirs because of the absence of data, this particular instance of inaccuracy represents only 1.13% of the total.