Giving Power Transformers the attention they deserve

Power transformers are ubiquitous devices of our electrical networks. They are a key component that allows the interface between networks of different voltage levels and without them, the AC networks that we know today would have not been possible. In fact, they were a deciding factor that made AC win the War of the currents.


Substation power transformers are the critical link that connects wind and solar farms to the electrical network. When transformers fail, they often cause long and costly downtimes of the entire, or part, of the farm.

Figure 1: Power Transformers are a single point of failure in wind and solar farms.

To understand the economic impact that these downtimes may have let’s examine the figure below. We can see how the yearly revenue loss caused by these downtimes changes when downtime duration varies from 2 to 8 weeks and total site capacity varies from 100MW to 500MW. For example, if a site with 300MW total power capacity has one transformer which is offline for 5 weeks, this translates to a cost of 0.75M Eur/year.

Figure 2: Yearly revenue loss due to power transformers downtime. Assuming a 24% capacity factor, energy price of 60Eur/MWh (source) and two transformers at the substation (each with 60% of the project capacity).

Yet, despite being a critical link with such economic risk, power transformers are often “forgotten”. They tend to be pieces of equipment that are underinsured, under sensorized and under monitored. As a consequence, maintenance teams can only rely on sparse analyses such as oil and electrical tests that require bringing the power transformer offline. 

We believe there is a need to improve current monitoring solutions, to reduce the risk still associated with these assets. We believe that is time to give power transformers the attention they deserve.

Putting power transformers back in the main stage

We offer an end-to-end solution to monitor power transformers that can be divided into two main parts: installation of sensing equipment and application of our predictive intelligence stack. Let's dive into each one of these in more detail!

Collecting the right data

We first need to collect the right data to allow proper monitoring. To do so, a sensing hardware solution is retrofitted to existing power transformers which enables the measurement of voltages and currents of both the primary and secondary sides collected at sampling frequencies in the kHz order.

The proposed sensing solution has a set of highly desirable characteristics: 

  • its installation procedure is simple, safe and non-invasive (measurements are made at the electric cabinets, not at the asset directly), 
  • it enables continuous real-time monitoring during normal operation and 
  • is adaptable to any transformer type since it’s independent of the voltage, core type or load of the power transformer (in fact, it is also adaptable for monitoring distribution and wind turbine transformers).
Figure 3: A diagram of the retrofitted sensing hardware solution

Once the installation is complete and data acquisition is ongoing, the transformer health monitoring can begin. This is where Jungle’s predictive intelligence stack jumps in.


Predictive intelligence applied to Power Transformers

Our normality ML models allow us to understand, at each point in time, if the assets are operating as expected, e.g. internal components temperatures and pressures (see our blog post for a deep dive about our normality models) and create alarms when deviations are detected. For a more detailed explanation of our alarms mechanism please have a look at our blog posts (part 1 and part 2) dedicated to the topic.

Figure 4: We create a unified model that extracts the general dynamics from the available assets but also allow each one to have their normality levels. 

Our normality models are very powerful and generically applicable. The same models can be employed to create a digital footprint of a power transformer enabling visibility on how each transformer sub-component should be behaving at any point in time. It is by combining the right sensing hardware solution with our state of the art predictive intelligence stack that the full monitoring potential is unleashed.

Our monitoring solution enables the unambiguous identification of more than 80% of the most common failures of power transformers as well as early detection of abnormal behaviour that precede such failures.
Figure 6: Failure types covered (in colour) by our monitoring solution amounting to >80% of the most common transformers failure types (source).


Real-world case examples

Curious about the insights our solution is able to distil? We have successfully identified many failures and below we show three examples of such findings.

Case 1 - OLTP degraded contacts

Our system detected a problem in consecutive taps of an on-load tap changer (OLTC). This triggered a maintenance inspection where the detection was also visually confirmed. The OLTC oil was changed and a later replacement was scheduled.

Figure 7: Degraded contacts show a different signature that is easily distinguishable from problem-free contacts.

Impact: Early fault detection and a greater lead time for the OLTP parts ordering was possible. Maintenance intervention scheduled to minimize cost impact.


Case 2 - Windings degradation

Our system detected a degradation located in the windings of the transformer due to abnormal phenomena (partial discharges, leakage earth currents). Three months after this was confirmed via oil analysis. The power transformer has been operating with this ongoing degradation. Our system is continuously checking any potential failure developments.


Figure 8: Excitation currents with unusual spikes (left) and an asymmetrical intrinsic signature.

Impact: Investment delayed and minimized. Maintenance intervention scheduled for a later moment to minimize cost impact.


Case 3 - Oil analysis false positive

Despite our system not detecting any abnormal behaviour, an internal failure was detected via recurring (every 3 months) oil inspections. The transformer was sent to the factory for inspection and the failure presence was not confirmed. The false positive was caused by the transformer insulating oil being contaminated.

Figure 9: A false positive given by an oil analysis was later proven wrong after a open-core inspection at the factory.

Impact: Transformer was not operating for a long period. The maintenance intervention in-factory amounted to 750 000 €. All of which could have been avoided.



Main advantages for our users

We put a spotlight on power transformers with our solution. The main advantages for our users include:

  • Continuous monitoring of power transformers for >80% of typical failures
  • State-of-the-art Machine Learning models specially tailored with your assets data
  • Actionable insights delivered via our online dashboard
  • Targetted, informed and shorter maintenance interventions
  • Eliminate false alarms (positive and false negatives) and unnecessary maintenance intervention costs
  • Increased availability and reliability 
  • Transform unplanned downtime to planned downtime
  • Minimize energy production impact
  • Maintenance cost reduction
  • Avoid catastrophic power transformer failures
  • World’s first continuous online electrical tests (online short-circuit inductance measurement)
  • No need for online DGA systems
  • Lower oil analysis frequency requirements (which require the transformer to be taken offline).
Figure 10: Jungle offers an end-to-end solution for Power Transformers monitoring
Get in touch

If you identified yourself with our mission and approach to predictive maintenance, feel free to reach out! Let us help you reduce uncertainty from your operations! Send us a message to hello@jungle.ai, and we will be in touch shortly after! It’s that easy to get started.


 


 


Keep an intelligent pulse on your assets.
start small, build big, together creating value from day one
Start Now