Data mining/analysis

Data-Driven Production-Impact Assessment During Unplanned Facility-System Events

A data-driven approach to successfully analyze and evaluate production-fluid impact during facility system divert events is presented. The work flow effectively identifies opportunities for prompt event mitigation and system optimization.

Getty Images

A data-driven approach to successfully analyze and evaluate production-fluid impact during facility system divert events is presented. The work flow effectively identifies opportunities for prompt event mitigation and system optimization. The methodology used includes data streaming, advanced computations on high-frequency data, visualization, data mining, and rules extraction. The paper demonstrates that applied data-driven analytics led to learnings and observations that had a positive effect on the management of facility systems during divert events.


The subject field is a heavy-oil reservoir, using steam for enhanced oil recovery. The field architecture is set up so that the active wells separate the produced liquids and produced (casing) gas at the wellhead. The operating objective is to minimize the pressure in the casing gas-collection system to maximize the liquid production from the wells. The casing gas flows from each individual well through a check valve into a common network and then enters one of the two casing-gas stations in the field. The gas is cooled, which removes a large volume fraction of condensables (mostly water) from the gas stream. Next, the gas flows into the inlet of the gas-plant compressors. From the compressors, the gas will follow paths that depend on the operating conditions. During normal field operations, the gas flow goes into the plant. The plant removes hydrogen sulfide so that the gas can be incinerated in the steam generators.

When the plant or the steam generators are not available to run because of planned or unplanned events, the discharge of the compressors is redirected into the casing-gas-collection network. This mode of operation is called “divert.” Operating in divert mode results in an increase in the casing-gas-network pressure with respect to normal conditions. This high pressure on the casing of the well is known to reduce the production volume.

Data Availability

An extensive data-identification exercise was conducted for this study. The result consisted of both structured data (surface and subsurface) and unstructured data such as event notes, dates, and unique features.

Event Data. For a better understanding of this concept, we provide the following definition: An event is defined by an increase in the overall system pressure caused by a planned or unplanned system shutdown that increases the casing-gas pressure and imposes a back­pressure on the formation, thus adversely affecting production rates and, as a result, the field output.

A total of 158 events have been identified since 2009. Because they represent the most-recent and -accurate information of the facility system, for this study only the events from January 2012 through October 2013 were considered. A thorough analysis of each event, with particular concentration on duration and impact, led to the decision to include only those events longer than 1 hour.

Data Streams. Multiple sources of data at different locations in the field provided diverse and convergent information to describe the entire-system behavior and the event behavior.

Casing-Pressure Data. For the wells equipped with individual pressure transmitters, only the average daily pressure data were available. Given that most of the events last from 3 hours to more than 24 hours, these data do not have the granularity to allow extraction of observations and conclusions. However, pressure effect for the longer events was retrieved.

Reservoir-Pressure Data. The field has several pressure observation wells that allow monitoring the reservoir pressure and understanding the efficiency of the steamflood operations.

Production Data. Two main sources of production data were considered for analysis: supervisory-control-and-data-acquisition (SCADA) data, which provided fluid information with a frequency of seconds, and the “system of records” data, which provided daily production data.


  1. Work Flow. Data gathering, data processing, and development of a new visualization tool (discussed in detail in the complete paper) constituted the work flow.
  2. Event-Processing Visualization: Assurance Analysis and Validation.To validate the automatic calculations of the main attributes for the duration of events, a group exercise was performed. By use of the specific event knowledge domain, the event-window duration was adjusted in the case of a few events.
  3. Reservoir-Pressure Maps and Production Effect. In the subject field, the reservoir pressure is monitored by a number of observation wells spread across the field. Pressure surveys are run twice a year in each of the observation wells. The pressure information not only helps to understand depletion rate in the reservoir but also guides future development projects.
  4. Casing-Pressure Maps. The shutdown event leads to an undesirable pressure increase in the system, with detrimental effect on well-production capability. When operating in normal conditions, most of the wells exhibit casing pressure within a range of 5–40 psi. While in “divert” mode, the pressure increases significantly across the field, with the highest values in the southern and central part of the field (ranging between 70 and 80 psi).

Observations and Results

Pressure Profiles During Event. An event lasting more than 2 hours exhibits a well-defined pressure profile and is characterized by four regions coded A, B, C, and D, as shown in Fig. 1.

  • A very rapid (15- to 25-psi/hr) rise as soon as the plant goes into divert mode (Region A)
  • A slow rate of pressure increase (0.8 psi/hr) during the time the plant is down (Region B)
  • An immediate relatively short rapid decrease as soon as the plant starts up (Region C)
  • A decreasing pressure trend as the plant ramps up to the normal processing rate and pulls the gas out from the reservoir (Region D)
Fig. 1—Behavior of a standard pressure profile during an event.


Event Length. Event duration has a significant influence on gross production. In addition to the event pressure profile, the event duration is a critical attribute for event characterization. Clustering by the duration of each event, four types of classes were defined on the basis of event length, with events categorized within these classes; these are detailed in the complete paper.

Oil- and Gross-Production Behavior During Event. A 20–30% higher oil impact was observed vs. a 15–20% gross-production impact measured during a long event.

The lowest gross-fluid production was approximately 15–20% less than the pre-event level; however, the lowest oil production was 20–30% with respect to pre-event values. It was possible to compare the oil and gross-fluid impact only during long-enough events (longer than a day), which accounted for a total of three events. Although this is a limited data sample, the 20–30% impact was consistent with previous production models. This implies that increased casing pressure affects the overall oil cut for the field.

Pressure Amplitude in Divert Mode (Greater Than 55 psi). Pressure magnitude during an event has a significant influence on production; there was no production impact observed when ­casing-gas-system pressure was below 55 psi. When the system is in divert mode, gas accumulates in the system, thus creating a significant casing pressure that can match or exceed the reservoir pressure. This restricts the flow of liquid into the wellbore and reduces the productivity of the well. The threshold value of the casing-gas-system inlet pressure when gross-­fluid production would return to pre-event ­levels is approximately 55 psi.

Well Casing-Pressure Values Show Regional Variation. Values are higher in mature developed areas of the field during an event.

Distribution Range of Percentage-Gross-Production Impact. Effect on gross fluid production is highly variable between the first 2 to 16 hours and plateaus after 30 hours. The percentage-gross-fluid impact is defined as the ratio of the difference between the average gross production before and during the event, to gross production before the event.

This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 173993, “Data-Driven Analytics for Production-Impact Assessment During Unplanned Facility-System Events,” by Andrei S. Popa, Hugo Leon, Juan Medel, Tuan Nquyen, Steve Cassidy, and Dallas Tubbs, Chevron, prepared for the 2015 SPE Western Regional Meeting, Garden Grove, California, USA, 27–30 April. The paper has not been peer reviewed.