Digital twins are used in a variety of applications like anomaly detection, asset management, and fleet management. For the purposes of this post, a digital twin is a digital replica of a physical asset on your power grid.
- 1 What is a Digital Twin?
- 2 How can Digital Twins be used in predictive maintenance of power grid assets?
- 3 Predictive Maintenance versus Operation Count Based Maintenance
- 4 How are Digital Twins used in Preventative Maintenance?
- 5 OTELLO, empowering predictive maintenance in power girds.
- 6 An example of using Digital Twin technologies for preventative maintenance application in the power grid
- 7 Creating a Digital Twin with MATLAB and Simulink
- 8 Detect and Predict Faults Using Machine Learning
- 9 Talk to us about this and other opportunities the OTELLO platform will give you, the grid operator, to build a modern, resilient power grid.
What is a Digital Twin?
According to Wikipedia, a Digital Twin refers to a digital replica of actual physical assets (physical twin), processes, places, systems, and devices that can be used for various purposes.
Definitions of digital twin technology emphasize two important characteristics.
- Firstly, each definition emphasizes the connection between the physical model and the corresponding virtual model or virtual counterpart.
- Secondly, this connection is established by generating real-time data using sensors.
The concept of the digital twin can be compared to other concepts such as cross-reality environments or co-spaces and mirror models, which aim to, by and large, synchronise part of the physical world (e.g., an object or place) with its cyber representation (which can be an abstraction of some aspects of the physical world).
How can Digital Twins be used in predictive maintenance of power grid assets?
A digital twin is an up-to-date digital representation or model, of an actual physical asset in operation on your power grid. It reflects the current asset condition and includes relevant historical data about the asset.
Digital twins can be used to evaluate the current condition of the asset, and more importantly, predict future behavior, refine the control, or optimize operation.
Predictive maintenance evaluates the condition of equipment by performing periodic (offline) or continuous (online) equipment condition monitoring. As such, predictive maintenance would generally require making real-time or time-sensitive decisions.
Per Wikipedia, the ultimate goal of the approach is to perform maintenance at a scheduled point in time when the maintenance activity is most cost-effective and before the equipment loses performance within a threshold.
In energy production, in addition to the loss of revenue and component costs, fines can be levied for non-delivery, increasing costs even further.
Predictive Maintenance versus Operation Count Based Maintenance
Time- and/or operation count-based maintenance, is where a piece of equipment gets maintained, whether it needs it or not. Time-based maintenance is labor-intensive, ineffective in identifying problems that develop between scheduled inspections, and so is not cost-effective.
The “predictive” component of predictive maintenance, stems from the goal of predicting the future trend of the equipment’s condition. This approach uses principles of statistical process control to determine at what point in the future maintenance activities will be appropriate.
Most predictive inspections are performed while equipment is in service, thereby minimizing disruption of normal system operations. Adoption of predictive maintenance can result in substantial cost savings and higher system reliability.
Reliability-centered maintenance emphasizes the use of predictive maintenance techniques in addition to traditional preventive measures. When properly implemented, it provides companies with a tool for achieving the lowest asset net present costs for a given level of performance and risk.
How are Digital Twins used in Preventative Maintenance?
Digital twins integrate internet of things, artificial intelligence, machine learning and software analytics with spatial network graphs to create living digital simulation models that update and change as their physical counterparts change.
A digital twin continuously learns and updates itself from multiple sources to represent its near real-time status, working condition, or position.
This learning system learns from itself, using sensor data that convey various aspects of its operating condition; from human experts, such as engineers with deep and relevant industry domain knowledge; from other similar assets; from other similar fleets of assets; and from the larger systems and environment of which it may be a part.
A digital twin also integrates historical data from past asset usage to factor into its digital model. Asset history is an important feature of a digital twin is that it captures the real asset’s history.
For example, if you’re using the digital twin for fault classification, the history captured by each digital twin can be the operational data from the specific transformer and its healthy and faulty state.
In the future, you can compare the operational data from one transformer to the digital twin histories of other transformers to understand how they behaved under similar faults and the effect on the fleet’s efficiency.
As the digital twins help you understand the history of the assets, they also help you with future planning. You can use digital twins to simulate hundreds of future scenarios to see how factors such as weather, fleet size, or different operating conditions affect the performance.
This approach will help you manage your assets and optimize operations by informing your operational and maintenance staff about the expected failures in advance so they can plan for future repairs and replacements.
Digital twins can be implemented where it makes sense for your application: at edge computing nodes (like the OTELLO VectoIII®), operational technology infrastructure, or IT systems.
It is also possible to deploy Digital Twins on commercially available cloud services such as Azure IoT Hub or AWS IoT, or implement custom integration as needed through APIs and other common integration methods such as shared libraries and RESTFul calls.
OTELLO, empowering predictive maintenance in power girds.
The OTELLO VectoIII® provides the data and tools that engineers and analysts require to develop predictive maintenance strategies.
Power grids, the world over, are facing the very real challenge of ageing infrastructure. Implementing predictive maintenance strategies provide them with a way of extending the lifespan of these assets, while allowing for significant cost savings.
The OTELLO VectoIII® provides the following, essential components for a predictive maintenance strategy:
- Compatibility with, and integration of existing sensors via open source and industry-standard protocols.
- Digital and analog inputs for additional sensors.
- It provides the connections between the physical assets on the grid and the digital/virtual assets via a scalable technology for pushing grid-wide sensor data to a central data-store.
- An edge-computing capability for enriching and rating of sensor data streams, before pushing the data to the central data-store. This provides massive scalability, without the need for additional and costly, centralised processing capacity.
- A massively scalable, open-source (license-cost free) data-store technology capable of handling large scale deployments.
- An open standard API for connecting to serverless modeling applications, Machine Learning services, and other computational tools for further processing and data modeling.
An example of using Digital Twin technologies for preventative maintenance application in the power grid
The OTELLO VectoIII® is often installed close to a transformer in a sub-station or mini sub-station.
With oil pressure, oil acidity, moisture, temperature, and vibration sensors connected to a transformer, the real-time operational status of the transformer can be compared to the manufacturers’ performance specifications. This can be used for generating performance deviation events and notifications.
Combined with Fast Transient monitoring during tap changes and switching events, comprehensive digital twins can be compiled for electrical transformers, tap changers and switchgear.
This data is streamed into Digital Twin machine learning or other algorithms to generate predictive maintenance events on the transformer.
It is well accepted that electric transformer maintenance is a key component of grid resilience. Considering the number of transformers installed across a grid, the maintenance cost savings and optimizations can be considerable.
In earlier times, 20 to 30 years ago, transformers were taken off-line, checked, and maintained or repaired based on a schedule. If the transformer was found to be in good condition, the operator had wasted time, resources, and money. Digital Twin technology presents a new opportunity for aging infrastructure assets in power grids.
According to the DOE, nearly 70 percent of the transformers in the U.S., are more than 25 years old. The average age of large power transformers in the U.S. is 38-40 years. Depending on the location of the transformer in the system, one failed transformer can result in an outage and potentially millions of dollars of lost revenue for electricity users, as well as large expenses for electric utilities.
Replacement costs for LPT’s (large power transformers) can range from $1 million to $7.5 million. It is for this reason, that power grid operators, seriously consider digital twin preventative maintenance strategies.
Other applications include predictive maintenance of power generation equipment such as power generation turbines, diesel engines, and locomotives.
Further examples of industry applications:
- Wind turbines
- Commercial and industrial Buildings
- Utilities (electric, gas, water, wastewater networks)
A digital twin can be used for monitoring, diagnostics and prognostics to optimize asset performance and utilization. In this field, sensory data can be combined with historical data, human expertise, and fleet and simulation learning to improve the outcome of prognostics. Therefore, complex prognostics and intelligent maintenance system platforms can use digital twins in finding the root cause of issues and improve productivity.
Over time, the lower marginal costs of developing digital twins of power grid assets, will make it much cheaper to test, predict, and solve problems on virtual representations rather than testing on physical models and waiting for physical products to break before intervening.
Digital twin technologies leave digital traces. These traces can be used by engineers, when an asset malfunctions, to go back and check the traces of the digital twin, to diagnose where the problem occurred. These diagnoses can in the future also be used by the manufacturer of these machines, to improve their designs so that these same malfunctions will occur less often in the future.
MATLAB® and Simulink® enable you to create digital twins with toolboxes for your applications like Predictive Maintenance Toolbox™, System Identification Toolbox™, and Simscape™.
Here’s a predictive maintenance video series from MathWorks.
Data-driven methods available with MATLAB include machine learning, deep learning, neural networks, and system identification.
With MATLAB apps, you can explore these modeling methods to find the most accurate method for your application.
With MATLAB, you can define a model using data from your connected asset. You can also use Simulink to create a physics-based model using multidomain modeling tools.
Physics-based modeling with Simulink involves designing the system from first principles. You can include mechanical, hydraulic, and electrical components.
Models can also come from upstream design work that uses Model-Based Design with Simulink.
Detect and Predict Faults Using Machine Learning
Identify root cause of failures and predict time-to-failure using classification, regression, and time-series modeling techniques.
- Interactively explore and select the most important variables for estimating RUL or classifying failure modes.
- Train, compare and validate multiple predictive models with built-in functions.
- Calculate and visualize confidence intervals to quantify uncertainty in predictions.