<<< Start >>>

<<< End >>>

The massive increase in computational power is rapidly introducing new possibilities that used to be nothing but fantasy. Some of them come with the buzzword artificial intelligence (AI)—basically, computers taking over (basic) manual tasks—or the closely related terms machine learning and deep learning (ML and DL).  

Traditional Geographic Information Systems (GIS) are now still heavily reliant on computational power, especially when it comes to 3D computations, map rendering, and calculating route options.  

We wanted to find a solution. So, we combined two technologies, AI and GIS, and that proved to be very promising to the utility world! Here are three useful examples we came across in our projects. 

<<< Start >>>

We wanted to find a solution. So, we combined two technologies, AI and GIS, and that proved to be very promising to the utility world.

<<< End >>>

1) Data cleansing in subterranean utility networks 

Data cleansing and data improvement are time-consuming tasks, especially when you want to get rid of missing or erroneous data of assets that sit below the surface. GIS utility data, a typically colorful collection of (historical) data of varying quality, is usually memorized by engineers rather than stored in databases. Here is where machine learning can offer unique opportunities to the utility world.

Data specialists can use machine learning algorithms to detect and ameliorate faulty data situations. For example, an algorithm can highlight which missing valve type to add when it assesses the diameter of the connecting pipes. 

<<< Start >>>

How can utilities create, improve, and sustain high-quality asset data?

The Dutch utility market is facing multiple disruptive challenges. To successfully face these challenges, we must be able to rely on high-quality asset data. Such data is not always available. The solution? Asset data management.


<<< End >

>The model then trains itself to get better by adding regional constraints or business rules. In town X, only valve type A matches the requirements, while in town Y, only valve type B is used. You can think of many other comparable situations in which the extraordinary combination of GIS and AI can help out.

2) Interpretation of utility drawings 

In the utility GIS world, many drawings exist, usually depicting parts of the network in various states (as planned, as built). Normally, the drawing data—having different purposes and quality—is manually recaptured in the utility asset registration system, instead of automatically feeding the correct data into the GIS system. Artificial intelligence can help with just that!

To prevent excavation damage, the governing bodies have been issuing national and international legislation to make the exact location of gas, water and electricity networks known to all parties doing excavation works.  

The last part of the network typically appears in millions of semi-analog connection sketches, describing where the cables and pipes are. This has resulted in huge demand for sketch digitization. 

<<< Start >>>

<<< End >>>

Here is where artificial intelligence comes into play. Complicated, hand-drawn sketches used to be handled manually, while a trained AI machine that's capable of interpreting complex data can sort the sketches automatically. By doing so, you can divide actual digitization of the sketches over juniors, who perform the simple tasks, and seniors, who take care of the more complex sketches.  

You can feed the less complex, semi-analog sketches directly into the GIS system, resulting in tasks that then only require a (sample) check before completion. 

3) Asset recognition in images

A third example is related to going through thousands and thousands of photographs—usually of the 'internal world'—of a station showing the configuration of assets. The collection of images gets scanned through image-recognition software trained to find specific assets and checks how assets are interconnected. This software also checks type attributes and reads meters on photos, such as the gas pressure. 

Image-recognition software works best when the pictures have a geotag, but it can also work when you link photos to specific objects in the GIS. Next, you'll need to derive the correct type of assets—based only on the picture—then check and correct the data available in the GIS database. 

<<< Start >>>

<<< End >>>

How to start the artificial intelligence integration process 

In any situation, you'll need to check whether AI is the best tool to apply to your problem or business case. A common mistake is to take AI for granted and spontaneously choose it as the ideal solution to fix your problem. As a matter of fact, it’s the other way around: you need to choose the right tool for the problem. But how do you determine whether AI is the right solution?  

  1. You need to have a reasonable amount of data available. A machine can't learn if it only has a limited set of example data available. The required amount of data—in most situations—depends on the complexity of the problem. Complex problems and situations require more data than simple problems.
  2. The available data must be representative of the problem you’re trying to solve. An AI model cannot recognize gas pipes in pictures if you train it with pictures of street lights.
  3. It's also important that you determine the level of repeatability or the extent to which the results can be generalized or reproduced: you need to consider the importance of pattern recognition. For example, it's possible to categorize types of gas pipes, because they possess specific characteristics. If all gas pipes had been unique, there would be no way to put them into generic categories. 

Quantity, representativeness, and repeatability of the data are simple rules of thumb to determine on a preliminary basis whether AI is a feasible tool for the situation. If these three are present, there is strong potential for applying AI to make your company’s life just a bit easier. 

<<< Start >>>

How Geographic Information Systems (GIS) can fuel the energy transition

With energy transition peaks, challenges arise as Dutch transmission and distribution (T&D) entities seek a way to cope with the rising demand. Find out how you can unlock the full potential of GIS and power the energy transition.


<<< End >>>

Three ingredients for being successful: culture, change, and mindset

In addition, culture, change, and mindset are highly important to realize AI projects at your GIS client. Usually, when people see that new technologies are being introduced, the first reactions are usually skeptical in nature.  

People are often used to a certain way of doing things and seemingly complex matters like AI tend to nurture resistance to change even more. Therefore, it takes a so-called ‘social-technological’ approach to successfully implement AI in a client organization.  

Besides showing impressive AI tools and complex data science to your client, you should also pay as much attention to how they generally feel about adopting (or accepting) new technologies and AI in particular. Be aware of their ‘AI maturity’ and approach them accordingly.  

<<< Start >>>

<<< End >>>

Finally, make sure that your AI project is something that you do together with your clients. Use co-creation techniques like design thinking to make sure everyone is on board. Or start small and simple with something easy to explain and understand. This is how you’ll significantly raise the chances of your (first) AI project. 

To fully train AI to be completely fit for these tasks can be challenging, especially with many obstacles lying ahead. However, experience shows that once trained, computers will rapidly take over the work that data engineers did previously, both in a highly qualitative manner and at a very fast pace. 

<<< Start >>>

Training artificial intelligence to completely fit these tasks can be challenging. But once trained, computers will rapidly take over data engineers' work with pace and proper quality.

<<< End >>>

Interpretation of utility drawings

We've explained many interesting trends in the world of GIS, but what does this all mean for you? 

Currently, major GIS software manufacturers are already including tools in their products to make machine learning more accessible to GIS professionals. For example, ESRI has provided several tools in their ArcGIS Pro product suite for image classification and object detection using deep learning techniques.  

Thanks to this, what used to take a lot of programming knowledge and a deep understanding of machine learning, is now available with just a couple of clicks away. 

Do you have any questions about artificial intelligence or Geographic Information Systems? Or are you looking for expert advice on this topic? Don't hesitate to contact us; we are here to help! 

Bas West

Associate Manager - Accenture Technology

Daan Robeerst


Rik Jacobs

Business & Technology Integration Senior Analyst - Accenture Technology

Subscription Center
Subscribe to Accenture Insights Subscribe to Accenture Insights