GRDF: Delivering an innovative way to ID chemical products with Google Cloud Machine Learning Engine

About GRDF

GRDF builds, operates, and maintains Europe’s largest natural gas network with 200,000 kilometers of pipelines serving 10.9 million customers across France.

Industries: Utilities
Location: France

About Devoteam

With offices in 17 countries, Devoteam G Cloud offers technology consulting services to manage transformations and help make companies truly digital.

GRDF uses agile development to rapidly prototype an app, using Google Cloud Machine Learning Engine with TensorFlow and Cloud Storage to enable technicians to identify 60 chemical products using their smartphones.

Google Cloud Results

  • Powers an app that identifies 60 different chemical products based on their packaging
  • Enables rapid prototyping and development, with the app ready to go in under two months
  • Supports experimentation with a transparent costing structure

Agile app development in less than 2 months

By maintaining the largest natural gas network in Europe comprising 200,000 kilometers of pipeline, since its creation in 2007, GRDF (Gaz Reseau Distribution France) has become a dynamic force in the French utilities landscape, serving 10.9 million customers and registering a turnover of €3.6 billion in 2017.

To position itself as a major player in the transition towards greener energy, GRDF is implementing key directives including the development of biomethane and transforming its network for greater efficiency. As part of this transition, the company is planning to go beyond its role as a distributor, most notably with services such as its smart meter Gazpar, the largest project of its kind in Europe for a gas network. It also plans to change the ways GRDF uses data in order to implement more agile, innovative ways of improving its network.

“For the moment, we are experimenting in order to understand how cloud computing can help us answer the demands of tomorrow. As the aim is to test products so their potential value is more tangible, we are looking for technological solutions that enable us to prototype something very quickly.”

Jean-Charles Jorandon, Head of Digital Innovation, GRDF

As part of this transformation, GRDF’s Digital Innovation team decided to explore how machine learning might help to solve one of its key challenges—better maintaining the safety of the network—by enabling its pipeline technicians and engineers to accurately identify chemical products. Google Cloud Platform (GCP) helped provided the answer.

“For the moment, we are experimenting in order to understand how cloud computing can help us answer the demands of tomorrow,” explains Jean-Charles Jorandon, Head of Digital Innovation at GRDF. “As the aim is to test products so their potential value is more tangible, we are looking for technological solutions that enable us to prototype something very quickly.”

Mobilizing AI rapidly

One of the main challenges for large utility companies is helping to ensure the security of their workers, as well as their networks. To maintain its pipelines, GRDF’s technical engineers often need to handle chemical products, whether to clean the pipes or fix a leak. Each chemical has an associated Safety Data Sheet (SDS), which provides instructions for safe handling. However, because these sheets are stored in paper files in technicians’ vans, they do not always read them prior to usage. GRDF looked for a way to enable its technicians to rapidly identify and access the correct SDS for the product they were using, by making it available on their smartphones. “The names of chemical products are often very complicated, so we needed to avoid having a list where the technician had to look for the right name and know exactly what it was called,” says Jean-Charles.

To do that, using Devoteam G Cloud, GRDF developed an application that uses TensorFlow and Google Cloud Machine Learning Engine to identify chemical products based solely on a photograph of its packaging. “Using AI, the app can automatically recognize a product by identifying the shape of the container or packaging, and pull up the correct information sheet for the technician,” explains Jean-Charles.

“We turned to Google Cloud Platform because the building blocks offered by Google Cloud Machine Learning Engine are immediately usable. The potential for a very rapid mobilization of those tools corresponds directly with our approach of testing and prototyping.”

Jean-Charles Jorandon, Head of Digital Innovation, GRDF

“In order for the recognition of shapes to work, you have to take thousands of photographs of the product. Every chemical product that we use was photographed from every angle, distance, and configuration. We made the photos available on Google Cloud Storage and then the Devoteam G Cloud team set up the parameters on Cloud Machine Learning Engine using TensorFlow to produce the recognition models, then integrate those models into an application.”

“We turned to Google Cloud Platform because the building blocks offered by Google Cloud Machine Learning Engine are immediately usable,” says Jean-Charles. “The potential for a very rapid mobilization of those tools corresponds directly with our approach of testing and prototyping. Also, Google has put so much research into advancing AI that we knew the technology was mature, which was a key advantage for us.”

Improving team knowledge

In addition to developing the application, implementation partner Devoteam G Cloud carried out workshops so GRDF’s teams could understand how the application worked and explore the potential of AI in transforming the company’s processes.

“Our data scientists were already familiar with machine learning and TensorFlow, but it was very interesting for them to go deeper into the technology,” says Jean-Charles. “Before the workshop, our developers didn’t know much about machine learning or AI. It was really important to us to improve our team’s understanding of Google’s products, and look at a concrete use-case.”

Innovation for the future

Thanks to Devoteam G Cloud’s agile development using GCP, the SDS smartphone app took less than two months to launch. “We implemented it in the Lyon region for a test period of three months,” says Jean-Charles. “The results are very encouraging: it successfully recognizes 60 products and enables our technicians to easily access safety instructions.” With GCP, GRDF was able to develop the prototype quickly, and with no unforeseen expenses. “The costs are very transparent, as well as being reasonable,” says Jean-Charles. “Using Google Cloud Platform means we can mobilize resources very quickly at a cost that is very reasonable within the framework of a test case.”

“In order to drive innovation, we need to be able to test very quickly, directly with the people doing the work, to develop solutions that respond to real-life challenges. We want to get solutions out there as quickly as we can, and using Google Cloud Platform, we can do that.”

Jean-Charles Jorandon, Head of Digital Innovation, GRDF

Now, GRDF hopes to explore more opportunities for future innovation. “We are looking into what else might help us tomorrow, in terms of transformation,” says Jean-Charles. “We are currently looking into voice recognition, and exploring more around big data, of course.”

“In order to drive innovation, we need to be able to test very quickly, directly with the people doing the work, to develop solutions that respond to real-life challenges. We want to get solutions out there as quickly as we can, and using Google Cloud Platform, we can do that.”

About GRDF

GRDF builds, operates, and maintains Europe’s largest natural gas network with 200,000 kilometers of pipelines serving 10.9 million customers across France.

Industries: Utilities
Location: France

About Devoteam

With offices in 17 countries, Devoteam G Cloud offers technology consulting services to manage transformations and help make companies truly digital.