Well Performance Monitoring by Statistical Processing and Classification of Well Test Data

摘要

Production of heavy oil brings numerous operational challenges that may require arncombination of generic proven solutions with specifically tailored technologies. Thernpaper outlines a general concept of a well performance monitoring system. Its keyrnprinciple is to systematically monitor deviations from expected behavior (productionrnrates), and consequently infer the most likely deviations’ reasons as well asrnrecommendations on possible actions to eliminate existing production bottlenecks.rnStatistical processing and classification of well test data is the first step of the overallrnworkflow, and its objective is to recognize if the test results represent real productionrnvolumes, or if the results are influenced by faults of the testing equipment, such asrnmixing of fluids in well test separator. The authors describe a statistical classificationrnalgorithm, which was tailored to the problem of well test processing. It is based onrnmemory-based classification method that utilizes principles of nearest-neighbor searchrnand lazy learning.