Furnas fosters new predictive maintenance culture with AI and Azure Support

In the mid-1950s, Brazilian industrialization faced its initial challenges, and the threat of an energy collapse was a major driver for the construction of the country’s first hydroelectric plant: the Furnas Plant, in the state of Minas Gerais.

This is one of the founding events of Furnas, an energy company that brings together a system of 21 hydroelectric plants, two thermoelectric plants, and a wind complex. Owned or in partnership with the private sector, they also manage 34,995.13 km of transmission lines and 72 substations in the states of São Paulo, Minas Gerais, Rio de Janeiro, Espírito Santo, Paraná, Goiás, Mato Grosso, Mato Grosso do Sul, Pará, Tocantins, Rondônia, Rio Grande do Sul, Santa Catarina, Ceará, and Bahia, and in the Federal District.

Furnas is a mixed economy power generation and transmission company, a subsidiary of Eletrobras (Centrals Elétricas Brasileiras S.A.) and has become a reference in the sector in terms of suitability for the Brazilian regulatory landscape, focused on operational efficiency, process performance, and sustainability of investments, in addition to transparency in its management.

Technology as a tool to empower the energy industry

“This project was born as an initiative to make Furnas an AI benchmark for the electrical industry,” explains Ana Claudia Rodrigues, Digital Transformation Manager at Furnas. “We partnered with an ongoing R&D project with the business objective of reducing penalties related to the unavailability of transmission assets.”

In the electrical sector, following the determinations and values established by the regulatory agency, Aneel (National Electricity Agency), the company is compensated for the construction and maintenance of power transmission assets for the purpose of delivering power from one end to another. Companies are subject to penalties in the event of failures that are calculated due to the unavailability being scheduled – whether there was scheduled preventive maintenance, for example – or unscheduled – resulting from an unforeseen situation that generates fines of higher orders of magnitude.

In 2020, Furnas started this project and, with an open notice, carried out a partner selection process, in which Pontifícia Universidade Católica do Rio de Janeiro (PUC-Rio) joined as an academic arm, responsible for defining and building AI models, while Radix, a Microsoft partner company, joined as project executor, responsible for architecture, data integration, and other processes.

“Since the beginning of the project, the need has been placed for a delivery in production, in cycles, and in the form of a monetizable digital product,” adds Rodrigues. “An accelerator was needed, an architecture that gave us the possibility to accelerate deliveries.”

Managing a large volume of data for a complex and critical system

Predictive maintenance of transmission assets is already part of Furnas’ daily life, with routine maintenance and checks, and the knowledge of technicians, with technology being an ally to the procedure’s improvement.

“The project utilizes machine learning techniques for predictive maintenance of transmission system substation equipment that, in turn, are part of the national interconnected system,” explains Marcelo de Carvalho, Manager of Digital Transformation and technical coordinator of the R&D project. “The goal of the work is to be able to identify failure situations from equipment operational standards and operational data such as orders, notes, and records.”

Thus, already used to the Microsoft Cloud that would allow the insertion of that solution into the Furnas architecture, a key concern of the project was in relation to the data from several core systems of the company, which were collected and concentrated in a data lake, where they could be worked with.

The solution architecture brings together Azure Data Lake, a storage service that offers identity, management, security, and easy integration with data repositories like Azure Data Factory, a fully managed data integration service, Azure Synapse Analytics, which performs big data processing for insights generation, Azure AI, an AI service suite, and Video Analyzer, a platform with the ability to capture, record, and analyze live video and publish the results. Finally, Power BI also integrates the solution.

Having started in 2020, the deployment of the solution is expected to end in 2023. With the scheduled cycles, the Digital Transformation team has been gradually feeling the possibilities generated with the new technology, the first steps that can have a big impact on the entire industry.

“Electrical companies still lack simpler evolutions, such as data integration and standardization,” explains de Carvalho. “In the project, we are delivering a data lake that will be used in other solutions as well, a gain that is already extremely important.”

New tools fostering a new culture

Rodrigues also states that the project, in addition to generating product, has a very strong internal capacity, bringing all the knowledge inside the company – which has a path of cultural change ahead. “We are a company with mission-critical systems, and we will have to overcome a cultural issue with this project as well,” she says. “What’s going on is simulations and once the project is ready, you’ll need to prove that the process will gain speed and safety, and we need to show the accuracy of the models.”

As a traditional company of a traditional sector, fear comes a lot from the need to execute an operation critical to the functioning of the country as a whole. With this and other projects underway, Furnas has been finding ways to meet one of the project’s initial objectives and become a reference in AI in the electrical sector.