When artificial intelligence is brought up in conversation, the classic idea of a robot versus a human emerges – somewhat of an us-versus-them mentality – but artificial intelligence works at its best when it – machine learning, natural language processing, and robotics – is viewed as a partnership with the human workforce. Enter augmented intelligence, which sits at the nexus between artificial intelligence and humans, and revolves around technology helping people to complete their work more efficiently and allowing them to focus more on high-value “human-only” type activities.
Today’s utilities are faced with multiple market disruptions including the proliferation of distributed energy sources, evolving regulatory and policy changes, the increased adoption of energy efficiency products and programs, changing consumer behaviors, and an imperative to modernize their technologies and processes. Faced with these disruptions, utility executives can leverage innovative approaches such as augmented intelligence to position themselves for success.
Exploring the ‘art of the possible’ with machine learning and natural language processing
Capital Budget Planning
Utilities make investments in new equipment by upgrading existing assets, such as transformers and substations, and performing preventative maintenance — all with the goal of improving reliability of service. Current approaches to budget planning require utility engineers to try and analyze hundreds of different parameters from dozens of different data sources to identify the capital investments that will deliver the most improvement to reliability.
Utilizing outage and other operational data, maintenance history, asset types, and load patterns, as well as new nontraditional variables such as DER adoption, energy efficiency program efficacy, augmented intelligence can more systematically and consistently define the most effective ways to deploy capital in modernizing or upgrading the electric network. By analyzing patterns of prior investments that have improved reliability of the grid in the past, and factoring in diminished returns, unsupervised machine learning could reveal the most impactful near-term capital investments in the electric grid to the human workforce.
Damage Assessment and Restoration Activity
After a storm or fire event, utilities typically send crews out to assess the damage to their assets (poles, wires, transformers) and then they prioritize restoration and repair activities. The crews visually review the storm damage and phone in their reports to the storm base, and then they drive to the next area and repeat this activity manually/visually over many days.
Based on a combination of aerial, drone and satellite imagery, augmented intelligence could be used to analyze images to more quickly and reliably assess damage to the grid after a storm (downed wires, damaged poles and transformers), ultimately helping the utility to determine the priority of repairs that would restore power in the fastest, most effective manner.
Electrification of Transport
Electric vehicles (EVs) create a challenge for utilities in planning for load growth. Right now, engineers are trying to predict – with disparate datasets and manual analysis – which customers on their network will likely purchase an EV, as well as the timeframe. This information is used to predict the network load and help determine if upgrades are needed. For example, if 20 customers are all on the same circuit and they all purchase a Tesla at around the same time, the load requirements on that circuit would increase significantly, potentially causing outages.
Based on consumer behavior (e.g. social media feeds), car dealership sales, local government incentives, augmented intelligence could be used to identify a pattern for consumer and fleet vehicles (e.g. UPS, FedEx, etc.) electric car adoption. By predicting when a utility customer might purchase an EV, a utility could plan its investment in the electric grid to increase load requirements, thus ensuring no overloads (outages) on a given circuit.
Customer Engagement
When a utility’s call center receives calls from customers, a Customer Service Representative (CSR) may not know the identity of the customer, their past interactions with the utility, nor may they know what the customer might be calling about at that moment, etc. This leads to a poor customer service interaction, and a CSR who is unable to be proactive about how they serve the customer or offer them additional or new services.
Utilizing predictive data analytics, the CSR would be able to identify the customer when they call, and predict why they are calling based on past interactions or specific customer characteristics. This in turn would provide the CSR with the best information to provide a more positive customer experience. The CSR would also be well-positioned to offer additional targeted services, such as energy efficiency services.
Some utility customers (e.g., millennials) prefer not to make a phone call at all. Recent advances in natural language processing (NLP) provide customers with the opportunity to use a chatbot instead of talking to a human being. NLP technology is used to evaluate the text entered in the chat field to automatically answer simple questions (e.g., “How much is my bill this month?”) or direct the customer to a human being for more complex questions (e.g., “Why is my bill high this month?”). In an outage, NLP can also be used to analyze social media content from Twitter, Facebook, and Instagram to improve utility operations situational awareness, which will ultimately allow the utility to provide informed updates to customers.
Virtual assistants, such as Amazon’s Alexa, also stand to transform the customer experience, allowing the customer to be able to ask questions or take actions based on connections with smart home devices or utility systems. For example, “What’s my usage so far this month?”, “What’s my solar contribution to my bill so far this month?”, “Pay my bill” or “Which appliance uses the most electricity in my home?” Taking it a step further, virtual assistants could also develop new skills, providing them with the ability to assist in certain actions (e.g. running the washing machine at the least expensive time of day).
The use of robotics to streamline organizational processes
There are several applications of robotics that utilities could adopt, with all of them providing a combination of cost savings, reducing operational expenditures, and offering better customer service.
Hazardous Vegetation Detection
Vegetation near power lines can cause power outages during severe weather events. Identifying and managing vegetation is a costly and on-going battle for many utilities – when to trim, where to trim and how often.
Autonomous drones could sense and follow the power line routes and automatically identify and mark overgrown points along the route (live video feeds to the cloud, coupled with machine learning). The drone would sense overgrowth and automatically fly a 360 around the area at different altitudes to provide a full view of the extent and type of obstruction. This would provide vegetation crews with specific GPS locations of problem areas so they can systematically plan and manage their vegetation removal actions.
Further, these drones can identify diseased or damaged trees outside of the utility’s typical trim zones for targeted removal. Doing so can be helpful in ensuring that they do not fall down on the power lines and cause outages during a storm.
Regulatory Inspections
Utilities are required to inspect their transmission and distribution power assets on a regular basis for damage, potential issues and environmental impacts.
Autonomous drones could be used to perform visual automated inspections of power assets including substations, transformers, poles, towers etc. After setting an initial starting point, the drone would sense the asset type ( e.g. pole mounted transformer) and then fly a predetermined flight sequence to fully capture a real-time multi-angle view of the asset. In addition to high definition video, drones can be fitted with thermal, inductive, X-ray or magnetic inspection devices for other types of non-destructive analysis. This can be particularly helpful in the future either during nighttime or less than ideal conditions (i.e. wind) too hazardous for humans.
Right now, there are significant hurdles, particularly in the United States, as Federal Aviation Administration (FAA) regulations limit drone operations to “visual line of sight,” meaning that the pilot or an observer must always be able to see the drone while in operation. However, progress is being made to modify the FAA regulations.
Construction & Repairs
When power poles and lines are damaged by storms or accidents, repair crews are dispatched to assess and repair damage by fixing or replacing poles, transformers, and insulators. Autonomous vehicles could be used to deliver tools and replacement parts to specific job sites more cost-effectively and quickly, and have the construction trucks loaded before the start of a shift so utility crews can minimize loading times and maximize construction times. The automated loading drones can record information generated by the autonomous damage assessment drones and determine which tools and spare parts are required to restore service. Again, this minimizes the loading times, and can speed restoration activities.
The road to augmented intelligence
The utility landscape has changed dramatically over the past several years, and as such, utilities have shifted their focus to digitizing existing business processes, meaning that they are layering new technology onto existing processes. However, to become a successful Next Generation Energy company, the path forward must include the integration of augmented intelligence into business processes, enabling them to truly innovate with technology.
