Physical systems, such as electricity grids, are very complex and thereby difficult to model. Digital twins provide a solution.
"Digital Twins are one of the top technological trends"
Here will we discuss a literature review (2021) that was performed by researchers at Bosch Engineering. The review focused on how digital twin and "big Data" technology can be applied to complex physical systems, such as energy grids.
Without further ado, let’s dive in.
1 – The Problem
Complex physical systems are difficult to work with. Just take the image below as an example. Within this network of a few cities, there are millions of electrical systems that require power at the flip of a switch.
Optimizing and iterating upon such a large system is very difficult without the help of data.
2 – The Solution
Digital twins are digital copies of physical systems. By digitizing the physics as well as supply/demand mechanics, we can run simulations on new energy sharing policies, grid structures, and a variety of other previously untestable fields.
Digital twins are becoming widely adopted in nearly every physical industry, some of which include supply chain optimization and rocket design.
According to the literature review, there are 4 broad areas where digital twins have been applied in the Energy grid space. Let’s discuss each in turn.
2.1 – Asset Modeling
Asset modeling focuses on how energy is generated. Overall, these models have been implemented with success for small-scale systems but not at grid-scale. Developing a cohesive digital twin for such systems requires gathering standardized data from many providers, such as solar farms and natural gas power plants.
Some private companies such as ABB and Honeywell have successfully developed local asset models for their respective projects but did not attempt to scale their models beyond the organization.
One interesting paper applied Machine Learning models to optimize energy storage at local nodes, which was then distributed as needed.
Another obvious application is load i.e. demand forecasting. If we can perfectly predict the amount and location of electricity demand, we could reduce storage requirements and reduce blackouts.
Finally, let’s conclude this section with a fun fact. According to the EIA, ~66% of electricity is wasted in the production and transportation process. While the vast majority of this is lost during the generation process, even 1% of US electricity consumption equates to millions of dollars.
2.2 – Fault Modeling
Fault modeling focuses on predicting and preventing hardware failures. Fault modeling is one of the more underdeveloped fields but holds enormous potential.
Some potential causes of faults include hardware failure, natural disasters, and cyberattacks. Let’s check out some solutions…
General faults are tackled here – a variety fault types are categorized and systematically reviewed. If you’re specifically interested in the clean energy side of fault diagnostics, here’s a similar paper that focuses on renewables.
On the cybersecurity side, ANGEL is a framework where a digital twin runs along side the physical grid. When discrepancies are observed, an alarm is raised. Pretty intuitive, right?
There are also some cool discrete-time models leveraged to forecast faults. However, as noted above, other than a few specific applications this area of digital twins for energy grid optimization is pretty underdeveloped.
2.3 – Operational Modeling
Operational modeling focuses on the distribution of energy throughout the grid. By optimally storing and transporting energy, we can reduce the 66% number above. However, as with fault modeling, there are not very robust solutions.
One cool application focuses on developing net-zero energy buildings. Here’s a paper that leverages hierarchical flow charts and building information models to evaluate and optimize building design. This paper focuses on the thermal components of optimizing building design.
Another more technically challenging area is providing realtime optimization of grid load. Researchers and the Sandia National Labs have developed a realtime optimization framework for grid load, specifically for solar energy.
As you might imagine, realtime operational optimization is probably the most challenging application of digital twins – it requires harmony between hardware and software at a large scale. Currently, we don’t have very robust data to develop these models.
2.4 – Business Modeling
Finally, business modeling focuses on optimizing the economics of transferring energy between producers and consumers. Business is an especially interesting application because it can be used to test a wide range of behavior-related activities, such as tax breaks or cap and trade systems.
First, digital twins are very effective communication tools. They provide a proof of concept in previously untestable marketplaces, such as an entire city. For startups looking for funding, these models paired with visualizations can be invaluable fundraising tools.
One business model mentioned in the paper described a trading market for energy. In this system, end-users, like you and me, become temporary owners of electricity. If you have a surplus, you can sell at market prices. Testing this model would require tons of work without digital twin frameworks.
3 – Summary
And there you have it! A quick overview of digital twins and their applications in the energy space.
Further development of these systems allows for rapid prototyping and evaluation of novel frameworks. They also can enhance current operations by providing a copy of current systems for comparison.
The technology is significantly underdeveloped and provides lots of opportunity for both entrepreneurship and expansion of larger business.
Thanks for reading! I’ll be writing 26 more posts that bring academic research to the DS industry. Check out my comment for links to the main source for this post and some useful resources.