Anomaly Detection

summary prodx computes a variety of daily anomalies from cleansed input data as well as derived metrics a full documentation of each type can be found below basic anomalies high water cut high gor downtime rndp across choke rod pump high fluid load esp upthrust downthrust gas lift shallow injection multipoint injection lift integrity end of tubing injection flow assurance unstable flow liquid loading productivity index pi decline basic anomalies basic anomalies are trend based anomalies that can be configured using a "simple" check or a trend based "linear" check the simple check is purely a threshold based alarm configured with an absolute threshold and a check direction (increasing or decreasing) the linear check fits a straight line through the input data (for a predicted start and end value of the line) and compares an absolute and relative threshold defined by absolutetrigger = predictedend predictedstart relativetrigger = absolutetrigger / abs(predictedstart) in the default configuration, prodx computes a downtime anomaly when downtime hours is greater than 1 using a simple check rndp across choke , high water cut , and high gor are computed using a linear check with varying thresholds based on primary fluid and well type rod pump rod pump high fluid load anomalies are computed by analyzing a moving window of fluid load and rod pump spm gradients against thresholds, considering a consecutive period of concern before triggering an anomaly esp esp anomalies include upthrust and downthrust , which involves analyzing the in situ rate at pump intake against the manufacturers max and min recommended operating ranges, adjusted to that day's operating frequency downthrust when the in situ rate at intake is below the minimum recommended rate, a downthrust anomaly is triggered upthrust when the in situ rate at intake is above the maximum recommended rate, an upthrust anomaly is triggered gas lift gas lift anomalies include shallow injection, multipoint injection, flow integrity, and end of tubing injection these anomalies are determined by analyzing the injection traverse against valve opening and closing pressures in the process of bottomhole pressure computation, valves are determined to be closed, likely closed, open & flowing, open & not flowing, or open & cannot flow it is important to remember that a gas lifted system is dynamic second by second, so while the daily traverse may indicate certain valve statuses, this can change intra day shallow injection if a single unloading valve is open and flowing with an orifice installed, a shallow injection anomaly is triggered multipoint injection if multiple valves are open and flowing, a multipoint injection anomaly is triggered lift integrity if gas lift diagnostics indicate that all valves are closed or open & cannot flow with a positive injection rate, a lift integrity anomaly is triggered when this occurs, an "optimistic" injection point is assumed at the deepest depth where the difference between production and injection traverse pressures at depth are greater than 45 psi (above the intersection point, if applicable) this scenario indicates a potential stuck valve, leaking valve, hole in tubing, or a transient state in which the casing pressure is actively building prior to normal valve operation end of tubing injection if gas lift diagnostics indicate all valves are closed, there is no packer in the well, and the production and injection traverses do not intersect shallower than end of tubing, gas will be assumed to be injecting around the end of tubing and an end of tubing injection anomaly will be triggered flow assurance flow assurance anomalies are intended to provide warning that the well could be experiencing liquid loading or slug flow liquid loading for gas wells, a critical rate is computed based on the configured correlation (turner, coleman, nagoo, or xecta hybrid) when the total gas rate at end of tubing is below the critical rate, a liquid loading anomaly is triggered unstable flow using the flow properties at the surface node of the traverse, the flow pattern is predicted using the tuffp ansari model slug flow patterns trigger and unstable flow anomaly productivity index productivity index anomalies are designed to intelligently alert when abnormal pi drops are observed in the well every configured number of days, a pi vs cumulative fit is established across a segment a segment is established each time a shut in event occurs that is of a minimum duration based on the fit, a standard deviation band is established any abnormal pis outside of the low band of standard deviation trigger a pi decline anomaly the graphic below illustrates a given window and the available settings for configuration pi decline anomalies are classified using the follow generalized flow