8 Artificial intelligence Uses in the Energy Sector

Share on facebook
Share on linkedin
Share on whatsapp

Artificial Intelligence (AI) has been pointed out as one of the technologies that will create disruptive innovation in many sectors, enabling automation of not only simple and repetitive processes, but of activities previously performed efficiently exclusively by human beings.

AI-based solutions have been successfully used in many sectors, aiding in disease diagnosis and prevention, equipment failure prediction in industry, the prediction of weather phenomena that can affect agriculture, the optimization of routes for the logistics sector, and in the service and interaction with consumers through voice interfaces etc.

In the coming years, AI will drive the Digital Transformation already underway in many sectors even further, bringing greater efficiency to operations and processes, creating new business models and new ways for people to relate to products and services. The Utilities sector, in particular the Electricity segment, will also be part of this transformation and can benefit from it to overcome the new challenges of the field, known as 3Ds (Distribution, Decarbonization and Digitalization).

Distribution: the popularization of solar panels enabled energy to be generated in a distributed way, no longer centralized in large power plants. Consumers are now able to produce their own electricity, no longer being totally dependent on energy distribution companies. If, on the one hand, distributed energy generation can help the electrical system meet the growing demand for energy, on the other hand, this demand brings new challenges to the management and operation of the system, which becomes more dynamic and complex with the addition of these intermittent energy sources;

Decarbonization: the impacts of global warming and the need to reduce carbon emissions should boost the arrival of electric vehicles in a market that, although still incipient in Brazil, is already walking by leaps and bounds in some countries. The growing demand for electricity and the mobility of these charges will also increase the complexity of the management and operation of the electrical system;

Digitalization: connectivity and digitalization have profoundly changed our life habits, altering the way we relate to products and services and requiring the energy sector to adapt to meet the demand of more empowered and demanding consumers. In addition, the new generation of equipment and solutions became connected, resulting in a multitude of data points about their use. However, it is necessary to adopt technologies that can actually extract knowledge from this data and produce practical results that can deliver real value for companies and their customers.

Below, we will introduce some AI uses in the electric power sector that can be and already have been applied by companies to improve their efficiency and overcome the challenges presented above.

1.  Detection of non-technical losses

Currently, there is much talk about sustainability and efficient use of resources. Electricity losses in Brazil still represent a great waste that affects both distributors and consumers.

According to ABRADEE (Associação Brasileira de Distribuidoras de Energia Elétrica or Brazilian Association of Electricity Distributors), the total electricity generated in Brazil in 2016 was 327 TWh, with a gross revenue of 216 billion reais. In the same year, global energy losses reached 13.9%, that is, around 45.4 TWh of energy was not commercialized, representing a billion-dollar loss for the sector.

Part of these losses (around 7.8%) occurred due to technical factors, such as the dissipation of energy in the form of heat (called the Joule Effect), which occurs by the very passage of the electric current through the wires that carry the energy from the generating source to consumers. However, another part of these losses (approximately 6.1%), called non-technical losses or commercial losses, were related to theft and fraud in energy consumption.

Some of these losses related to theft and fraud can be detected by applying Artificial Intelligence techniques similar to those used by other fields, such as the identification of credit card fraud. In these solutions, Machine Learning algorithms are trained with power usage data history, generating a model that can be subsequently applied to anomaly identification.

In the case of the energy sector, data may include, for instance, the monthly electricity use history (in KWh). Machine Learning algorithms are trained with data history, tracing a profile of clients. The algorithms can subsequently identify anomalies, that is, changes in this energy use profile that could potentially indicate irregularities.

In addition to the monthly energy usage, other information can be used to bring greater accuracy to the algorithm, such as connection type, phase type, billing group, voltage group, usage class, neighborhood, georeferenced coordinates etc.

Energisa, one of the largest electricity distributors in Brazil, is one of the dealerships that is using Artificial Intelligence technologies to detect irregularities. According to the company, in the four years following the implementation of the project, it has found a reduction of 3.2% in non-technical losses, energy enough to serve 2.4 million residential consumers during a month. Compared to the year before the project, there was a 370% increase in the amount of energy recovered.

2.  Breakdown of energy consumption

Electricity bills usually bring the amount of energy consumed by the unit in the period of a month. However, a bill doesn’t indicate how the energy was spent. A better understanding of how energy is consumed is a key element for clients to adopt efficient energy saving measures.

The more efficient use of electricity brings direct benefits to the consumer (through the reduction of their monthly bill), but it is also important to the electricity companies, reducing the investment required in the expansion of the generation infrastructure to meet the growing demand for electricity.

Artificial Intelligence techniques have been used to allow the disaggregation of energy consumed in an installation, identifying the share of each load in the overall consumption of a residence, business or industry.

The energy disaggregation technique is based on the use of an algorithm that has been trained to identify the changes that a particular equipment (such as an air conditioner) causes in the electrical signal of a facility (such as a residence). Different equipment cause different changes in electrical signals, so each piece of equipment has a signature pattern that is identified by the algorithm.

This technique is called NILM (Non-Intrusive Load Monitoring), since it does not require any type of change in the internal electrical installation of the consumer unit. The figure below illustrates how the different pieces of equipment can be identified from the analysis of the electrical signal of a residence.

 

Breakdown Survey Paper, Stanford

 

The identification of the electrical signature of the equipment can be made directly from the consumer unit’s figures, however, it is necessary that the measurements are made with high granularity — that is, in very short time intervals, so that the algorithms can infer the behavior of each equipment.

The implementation of smart grids, which use connected electrical meters capable of performing consumption measurement remotely and in real time, can allow energy companies to implement energy disaggregation solutions in their networks.

In addition to bringing greater engagement of residential consumers to the energy efficiency programs of the dealerships, promoting more conscious consumption of electricity, more detailed knowledge of user’s consumption profile can also generate other business opportunities for distributors and companies in the sector.

3.  Forecast of renewable energy generation

Renewable energy sources, such as wind and solar, have stood out as alternatives to meet the growing demand for energy in a clean and sustainable way. However, although they represent a great opportunity, the intermittent nature of these energy sources also brings challenges to the sector. AI techniques have been applied to mitigate these challenges and add even more value to these energy sources.

Recently, Google — through DeepMind, a subsidiary of the company that operates with Artificial Intelligence — announced that it is using Machine Learning techniques to predict the power generation capacity of the company’s wind farms located in the Midwest of the United States.

Wind turbine data history and weather forecasts were used by DeepMind to train a neural network capable of predicting the power generated throughout the day by the turbines of the plant 36 hours in advance. Based on these forecasts, it is possible to plan the use of this energy a day in advance, which makes this energy more valuable to the network.

The National Center of Atmospheric Research (NCAR) in Colorado, United States, has also been working with the use of Artificial Intelligence for wind power forecasting generated by state wind farms. The reliability of the use of energy from wind farms allowed the operators of the region to increase the share of renewable energy in their operation, with no effects on the stability and reliability of the grid.

In particular, for Brazil, which has a huge potential for wind and solar power generation, initiatives such as these can add enormous value to the electricity sector, further boosting the use of renewable energy sources.

4.  Management of distributed energy resources

Distributed power generation, carried out through solar panels or wind generators, reduces the control of dealerships over the infrastructure of the electrical system, since these pieces of equipment are distributed in the system and their energy generation has an intermittent nature, that is, it depends on environmental conditions (such as sun and wind).

In this scenario, Cleantechs (companies, usually startups, focused on clean and sustainable technologies) such as Embala and OpusOne Solutions are developing software systems called Distributed Energy Resource Management Systems (DERMS) to explore generation and storage resources in the most optimal way possible, ensuring the balance between energy supply and demand and bringing greater stability to the grid.

These systems use IoT technologies to monitor network power resources in real time and AI techniques to analyze how the capacity of Distributed Energy Resources (DER) assets and their control parameters evolve over time to predict available capacity. The modeling of the network’s behavior can then be used for the automatic and optimized control of the assets of DERs.

5.  Customer service

Digital Transformation has profoundly changed the way companies relate to their customers. In this new era, consumer experience has become a central element of business. In this context, AI techniques have been used to create new channels of customer service through interfaces based on written or spoken communication.

Natural Language Processing (NLP) is the field of Artificial intelligence related to human language processing, which allows machines to extract information from texts or even interpret speech.

NLP has been used by companies from many fields, including the energy sector, to extract important trends from customer feedback. Companies analyze data from emails, surveys, call center conversations and even comments on social networks to identify possible causes of customer dissatisfaction and implement improvements to their processes and services to better serve them.

More recently, NLP has been applied in customer service, allowing direct interaction through text (in this case, they are often called “chat bots”). Personal virtual assistants (such as Amazon’s Alexa) also allow for the creation of new voice-based service channels.

CEMIG is developing an R&D project to improve communication and relationship with its customers, applying AI, Big Data and chat bot techniques. The solution applies AI to decode messages sent by customers and create a custom response based on the user’s profile and relationship history. The system can also proactively initiates interactions motivated by specific events.

These new tools based on natural language can help industry companies improve consumer satisfaction with their services, as well as improve efficiency and reduce costs related to customer service processes.

6.  Smart recharging of electric vehicles

Smart Charging systems can optimize the use of public and private charging infrastructure, as well as minimize the impact of electric vehicles on electric system operations. Smart Charging systems are able to manage charging sessions, balancing the demand of vehicles with the availability of charging points and the supply capacity of the electrical system.

Artificial Intelligence algorithms are used to generate recharge strategies that take into account the needs of drivers, as well as the limitations of the grid and the electrical installation of charging stations. Systems can take into account the specific characteristics of vehicles (such as the capacity and current level of their batteries), the needs of drivers (their geographical location and charging preferences, for instance) and the public and private charging infrastructure available (including factors such as proximity and pricing).

Systems can plan charging sessions, managing not only the allocation of vehicles to charging points, but also the power that will be used dynamically during the charging session. The system can increase or reduce the charging power according to the instantaneous capacity of the network (or local installation), thus avoiding spikes and overloads in the electrical system.

Smart Charging can also implement a concept known as V2G (Vehicle to Grid), in which the battery bank of an electric vehicle can be used as a storage unit for the power grid. In this concept, a smart recharging system prioritizes recharging batteries in periods of excess power generation (for example, in periods when there is large generation in solar sources) and subsequently uses part of the stored energy to send power back to the power grid in periods of excess demand.

V2G is being studied as an alternative to increase the insertion of intermittent renewable energy sources without compromising grid stability. From the point of view of the operators of the electrical system, this integration with the network can help in the management of energy supply and demand, avoiding spikes and overloads of the network.

7.  Cyber security in the energy sector

The 2019 Global Risk Report pointed to cyber-attacks as one of the main risks for the coming years, also indicating a trend in the increase of attacks aimed at critical infrastructure sectors, such as electricity. This increase in the risk of attacks in the energy sector is related to the growing adoption of digital technologies, such as the Internet of Things (IoT), which add vulnerabilities to systems, increasing their exposure to attacks.

The first reported cyber-attack to cause blackout on a power grid happened in Ukraine, in December 2015. The hackers managed to compromise the systems of three power distribution companies and disrupt the supply of electricity to end consumers for several hours. Since then, other attacks and invasion attempts have been identified in different countries around the world.

In Brazil, there are still no confirmed cases of cyber-attacks that have caused power supply suspension, although there are rumors that blackouts occurred in 2005 and 2007 were caused by cyber-attacks. However, companies in the energy sector should be aware of this risk and consider investing in information security as one of their priorities.

Security policies should be adopted by companies aiming to ensure the security not only of their operation, but of their customers ‘ data. Artificial Intelligence techniques, such as anomaly detection, can help in the detection of possible invasions, so that measures can be taken to prevent further damage.

8.  Predictive maintenance and equipment fail prediction

The use of AI in predictive maintenance and equipment failure prediction has already been successfully applied in several fields. In the electricity sector, where uninterrupted maintenance of power supply is a top priority for power companies, these solutions can be particularly important.

In predictive maintenance solutions, IoT sensors and machine learning algorithms are used to continuously monitor network equipment. The algorithms identify deviations in the operation of the equipment, enabling maintenance actions that guarantee the efficient operation of the equipment or even perform equipment failure prediction. The system can assist in the diagnosis of failures and also allows the issuance of alerts to mitigate equipment stoppages.

Applications based on computer vision have also been used to perform the inspection of equipment and transmission and distribution networks. Through cameras installed in vehicles (or drones), images of the power grid are collected that are analyzed by neural networks capable of identifying problems such as cable rupture or simply tree branches reaching the power grid.

These solutions allow power companies to improve their efficiency by anticipating the maintenance of their equipment or by quickly identifying and correcting a problem that could cause damage or interruptions in the supply of electricity.

Conclusion

Artificial Intelligence is already in use in the electricity sector, allowing better operational efficiency, reduction of process costs, improvement in understanding and interaction with consumers, better management and optimization of the energy resources of the network, as well as identification and control of fraud and cyber-attack prevention. The quick evolution of AI solutions should bring even more promising results in the coming years, for these and other use cases.

Venturus develops solutions based on Artificial Intelligence, Machine Learning and Big Data for many business sectors. The institute is working on the creation of a Data Lab, a laboratory that will combine technical expertise and state-of-the-art infrastructure to meet the growing demand for data analysis related projects. To learn more about this or other projects, visit our website or contact us.

Veja mais
Industry 4.0
Pedro Almeida

Demystifying Industry 4.0

The term Industry 4.0, although it has already become popular, still brings a lot of doubt. Most people link the concept to highly technological factories
Leia mais »
Campinas / SP - Brasil

Estrada Giuseppina Vianelli di Napolli, nº 1.185
Condomínio GlobalTech Campinas
Polo II de Alta Tecnologia
CEP 13086-530 – Campinas – SP
+55 (19) 3755-8600

+55 (19) 3755-8600
contato@venturus.org.br

Bitnami