Statistical analysis of sucker rod pumping failures in the Permian Basin
This thesis serves the research project. The Artificial Lift Energy Optimization Consortium (ALEOC), which is supported by 11 oil companies in the Permian Basin. The objectives of ALEOC are to share successes and failures in production operations between consortium members, thereby reducing present operating costs, increasing lift efficiency, extending lower-rate well producing life and increasing oil well profitability. The first step toward the goal is to analyze the recorded databases to find out the production operation history and direct the future operations, and hence this thesis. The Permian Basin is one of the largest oil production areas in the world and sucker rod pumping is the main kind of artificial lift in that area. Wasson San Andres field is one of the top old fields and among the most complex in the Permian Basin. Denver City Unit is the largest of all the units in Wasson field. This thesis has just concentrated on tracing the history of this unit. Denver City Unit is operated by Shell Oil Company, it mainly produces oil from the San Andres formation (4700 to 7300 ft. deep, averaging 5200 ft.). The productive portion of the San Andres at Denver City Unit is subdivided into First Porosity and Main Pay. Main Pay possesses the most favorable reservoirs and porosity development. The discovery well was completed on September 28, 1935. Water flood began just after its foundation in 1964, and resulted in the peak production, 150,000 BOPD, in 1975. CO2 injection began in mid-1984, and maintained the steady production thereafter. Denver City Unit Water-Alternating-Gas injection process has the advantages over both continuous CO2 injection and WAG process. Experience shown that in Denver City Unit 7-in. casing has higher artificial lift efficiency. During the 1980s, the beam pumping units were mainly API 640's and 456's. The average run time between failures was approximately 15 months. In recent years sucker rod pumping failures have decreased gradually. The data provided by 11 oil companies came from about 25,000 sucker rod pumping wells, a quarter of the total sucker rod lifted well numbers in the Permian Basin. This is a big and reliable sample group from the population of sucker rod pumping wells in the Permian Basin. The databases were first pretreated from Access files or Excel files to the generalized Excel data file; with data sorting, the data were reorganized according to their company, field, location, formation and depth. Failure frequencies for total, pump, rod, and tubing were calculated to make them more comparable. According to the sorted failure frequencies, failure frequency plots were made to make them more straightforward. Observations of the failure data and plots revealed that different companies have very different failure frequencies, which is an index of field operation efficiency, facility manipulation, underground working conditions of the sucker rod pumping equipment; there is a trend of failure frequency decrease year after year among the participated companies with a few exceptions. In this thesis Fault Tree Techniques have been successfully applied to the analysis of the sucker rod pumping system. After the system was fully understood, a big fault tree was built from top event to bottom events. The evaluation of the fault tree is in the reverse direction, from bottom to top. The statistical probability of occurrence of the events at different levels were calculated. From the analysis of the fault tree structure and Company A's data, the conclusions are: because of its OR-gate structure, sucker rod pumping system is liable to suffer failure, any component may result in complete failure of the whole system; the downhole pump has the highest probability to fail: the weakest portions of the sucker rod string are polished rod, VA rod body, and 7/8 rod box and pin. Suggestions are to get deep into the working theories of the whole system; make the whole system equal-strength during design; find out the failure causes related to operation, manufacturer, equipment working conditions, and so on. Traditional statistical techniques are applicable to all kinds of observed data. In this thesis, the necessary tools have been presented, and used the data for all the companies' total as an example to show the analysis methods. To do the complete analysis here, normal distribution, x"-distribution, and t-distribution are needed to compute their means, variances, and standard deviations. By fitting the normal (or x'- or t-) distribution to observed data, we may convert the discrete system to continuous system, and do the sampling distribution analysis. Regression analysis is used to relate the dependent variable to the independent variable(s), and to predict the future occurrence on a statistical basis. According to the sampling analysis of the failure data from the Permian Basin, a rough idea about the failure frequencies are: total is 0.66 per well per year, pump is 0.25 per well per year, rod is 0.22 per well per year, and tubing is 0.16 per well per year. Due to the incompleteness of the failure data, the main purpose of this part is to provide the necessary methodology.