At a mature, high-water-cut oilfield production site, the curves on the SCADA screen keep moving and the parameters keep refreshing in real time. But the question that troubles operations staff most has never been “not enough data.” It is this: the polished-rod load on a pumping unit suddenly drops 60%. Is the sensor drifting, or has the rod actually broken? The winding temperature of an electric submersible pump climbs step by step, and sand content is rising too. Are the two related? The total liquid flow rate at the metering station keeps falling. Which well has the problem, and how long before it affects the transfer station?
These questions share a common feature: an anomaly is not confined to one device. It propagates step by step along the physical flow line from wellhead to metering station to transfer station to storage tank. Looking at the trend chart of any single device in isolation, you cannot see the full story.
Mature high-water-cut oilfield: high liquid production, low oil output, faster equipment degradation
The setting is a typical high-water-cut waterflood production block with an overall water cut of 78 to 92%. Its eight production wells (six beam pumping units and two electric submersible pumps) produce 240 m³ of liquid per day, but only 32 tons of oil. Large volumes of water are lifted to the surface, carrying only small amounts of oil.
This operating condition brings several hard challenges:
- Sucker rods experience large alternating loads, and their fatigue life is only 2 to 3 years (compared with 5 to 7 years under lower-water-cut conditions), so the risk of rod failure is high
- Electric submersible pumps carry a high overload risk. Formation sand production, combined with rising emulsion viscosity, can accelerate impeller wear and reduce motor cooling
- For a well at 90% water cut, a 1% metering error means the oil-output error is amplified tenfold, making hidden production losses hard to detect
- If an anomaly on a single well cannot be identified and traced within hours, it propagates step by step along the pipeline to the transfer station and even the storage tank
In the traditional setup, each system runs separately. Wellhead data lives in system A, metering-station data in system B, and station data in system C. When something goes wrong, engineers have to move back and forth across three systems, making it difficult to quickly piece together the full causal chain.
From isolated anomalies to visible propagation: how the data gets reorganized
TDengine’s approach is to map scattered sensor tags onto physical entities such as wells, metering stations, and transfer stations, forming an Asset Tree. With this model, each data point carries business context: which asset it belongs to, where it sits in the hierarchy, and which upstream and downstream entities it relates to.
On this Asset Tree, the real-time analytics engine continuously monitors key metrics for each node. When an attribute exceeds its threshold or an abnormal pattern appears, the system does more than raise an alert. It generates an event context, packaging the anomaly’s start time, end time, duration, and related parameter trends, then pushes it to the operations console.
More importantly, when an anomaly at an upstream wellhead propagates downstream, operations staff do not need to manually search three systems. In TDengine’s analysis workbench, they can follow the upstream-downstream relationships in the Asset Tree and bring the wellhead dynamometer card, metering-station total liquid flow rate, and transfer-station inlet pressure into the same panel. The anomaly propagation path becomes clear at a glance.
Two typical cases: how cascading faults are found and traced
Case 1: sucker rod failure, from dynamometer card collapse to a 14% drop in block output
Well CY-03, a beam-pumped well in the block, produces 18.5 m³ of liquid per day. During the mid-shift of Day 3, it experienced a sucker rod failure.
Figure 1 Polished rod breakage correlation analysis
Step 1: The metering station surfaces the first anomaly, an abnormal drop in total liquid flow
The total liquid flow rate at metering station MS-01 fell from 9.2 m³/h to 7.5 m³/h, a drop of 18%, lasting more than one hour. The system raised a Major-level alert. But this alert only tells you that less liquid is arriving at the metering station. It does not tell you which well has the problem.
Step 2: Trace up the asset tree to locate the anomalous well
Starting from the alert event at MS-01, trace upstream through the Asset Tree to check the four wells it monitors. The dynamometer card anomaly flag on CY-03 has already been set to true, which means the shape of its dynamometer card has changed significantly.
Step 3: Open the dynamometer card panel to confirm the breakage signature
Open the dynamometer card scatter analysis panel for CY-03. The instantaneous polished-rod load plunged from 78 kN to 28 kN (-64%), the dynamometer card area dropped from 154 kN·m to 48 kN·m (-69%), and the fill factor fell from 0.78 to 0.31 (-60%). These are typical signs of sucker rod failure: the load drops sharply, the card collapses, and pump efficiency falls.
Step 4: Track downstream propagation to quantify the scope of impact
Overlay CY-03’s daily liquid production curve, MS-01’s total liquid flow curve, and TS-01’s inlet pressure curve on the same panel, and a clear cascade appears. Seven hours after CY-03 failed, MS-01’s total liquid flow began to drop. Five hours later, TS-01’s inlet pressure fell from 0.42 MPa to 0.36 MPa, and total flow fell from 11.2 to 9.6 m³/h. The failure of a single well, CY-03, had propagated along the pipeline to the transfer station.
The complete causal chain: wellbore wax deposition and sucker rod fatigue failure -> a sudden drop in polished-rod load and a collapsed dynamometer card -> pump efficiency falls from 68% to 22% -> metering-station total liquid flow drops 18% -> transfer-station total flow drops 14% -> the storage tank level rises at half the previous rate.
Case 2: electric submersible pump failure, 14 hours after sand content rose
ESP-02 is an electric submersible pump well producing 35 m³ of liquid per day, the highest-output well in the block. From the mid-shift of Day 5, it went through the full sequence from sand production to pump failure.
Figure 2 ESP well overload shutdown correlation analysis
Step 1: Sand content exceeds the limit, the earliest warning signal
ESP-02’s sand content rose from 0.025% to 0.08%, more than tripling. In routine monitoring, this signal is easy to overlook. The sand content “only rose a bit.” But when you look at it together with pump efficiency and motor winding temperature, the trend becomes clear.
Step 2: Winding temperature crosses the warning line, confirming the unit is abnormal
Fourteen hours after sand content rose, the motor winding temperature climbed from 92 °C to 115 °C, past the recommended operating limit of 110 °C. Motor current also rose from 38 A to 45 A. Sand grains may have restricted cooling flow, while impeller wear added to the motor load. The two causal threads meet here.
Step 3: Current jumps and voltage drops, leading to pump shutdown
On the morning shift of Day 6, the winding temperature reached 142 °C, current jumped to 65 A, and voltage fell from 400 V to 360 V. Overcurrent protection tripped, and the unit shut down. ESP-02’s daily liquid production went from 35 m³/d to zero.
Step 4: From a single-well failure to a drop in block output
Overlay ESP-02’s shutdown with the data from MS-02 and TS-01. After ESP-02 shut down, MS-02’s total liquid flow fell from 10.5 to 6.8 m³/h (-35%), TS-01’s total flow dropped to 8.4 m³/h (-22%), and the storage tank level turned from rising to falling. The block’s export volume fell by nearly a quarter.
The complete causal chain: waterflood breakthrough leads to formation sand production -> sand grains wear the impeller and may restrict cooling flow -> sand content rises, pump efficiency falls, and motor winding temperature climbs -> insulation ages and breaks down -> the unit trips on overcurrent and shuts down -> metering-station total liquid flow drops 35% -> transfer-station export volume drops 22%.
Intelligence across the oilfield gathering system: what is missing is not data but causal insight
The two cases share a common feature. The root cause of each anomaly sits on wellhead equipment, but what gets noticed first is usually downstream: abnormal total liquid flow at the metering station, a drop in inlet pressure at the transfer station, or a storage tank level that stops rising. Traditional monitoring reacts to whichever point raises an alert. The real need is to start from a downstream alert, trace upstream against the flow, and find the root cause at the source.
TDengine uses Industrial Ontology modeling to turn the physical topology of wellheads, metering stations, gathering facilities, and storage tanks into a data topology. Its real-time analytics engine turns a single threshold exceedance into an event with context. Its analysis workbench lets key parameters from upstream and downstream assets be placed in the same panel for side-by-side comparison.
Combined, the three capabilities solve the same problem: operations staff no longer flip back and forth across three systems. Instead, they see the complete causal chain on a single panel.
TDengine includes a high-performance distributed time-series database, Industrial Ontology modeling, and the Industrial Agent Runtime. It provides a full-stack solution for industrial data streams, from collection and storage to real-time analytics, visualization, event management, and root-cause analysis. To learn more about TDengine, visit the official website at www.tdengine.com and download it for free.
Try it yourself
Install and deploy TDengine
Visit the TDengine Download Center, select TDengine All-in-One, choose the deployment platform and architecture that matches your environment, and follow the guided steps to complete the installation.
Load the sample data
After first activation, on the sample data loading screen, select Oil and Gas Field Production Process Monitoring and wait for the load to finish.
Customers who have already activated the product can click the user avatar at the top right, open the admin console, select Sample Data in the left menu, then choose Oil and Gas Field Production Process Monitoring to load it. The load takes a few minutes, after which you can start exploring.
Oil and Gas Field Production Process Monitoring
