Four points where AI is already helping in network management
The power grid becomes more and more complex the more renewable energy is fed into it. Where previously a small number of large power plants supplied most households with constant and regular electricity, today millions of solar systems generate variable electricity, wind turbines are connected to the grid every minute or switch off. Increasingly unpredictable weather conditions are making it even more difficult to balance electricity supply and demand. But how do you manage this chaos? The answer is increasingly: artificial intelligence.
Advertisement
The ability of AI systems to draw conclusions from large amounts of data and react to complex scenarios makes them particularly suitable for the task of keeping the power grid stable. A growing number of software companies are bringing AI products to the notoriously disruptive energy industry. In the USA, this trend has now been recognized by the US Department of Energy (DoE). Grants worth three billion US dollars are now expected to flow to various smart grid projects, which also include AI-related initiatives.
Interest in AI in the energy sector is growing. Some observers are already speculating about the possibility of a fully automated power grid in which, theoretically, people would no longer be needed – at least for everyday decisions. But the prospect is still a long way off. For now, the promise lies in AI’s potential to help people by providing real-time insights for better network management. The following four examples are intended to show where and how technology is already changing the work of electricity network operators.
1. Faster and better decision making
The power grid is often described as the most complex machine ever built. Because the network is so large, it is impossible for a single person to fully grasp everything that is happening in it at any given time, let alone predict what will happen later. Feng Qiu, a scientist at Argonne National Laboratory, a U.S. government-funded research institute, explains that AI helps the power grid in three ways: The technology helps operators understand current conditions, make better decisions and predict potential problems.
Qiu has spent years researching how machine learning could improve network operations. In 2019, his team partnered with Midcontinent Independent System Operator (MISO), an electric grid operator serving 15 U.S. states and parts of Canada. A machine learning model was tested that is intended to optimize daily planning work for a network that is comparable in size to that of MISO.
Already, grid operators like MISO perform complex mathematical calculations every day to predict the next day’s electricity needs and find the most cost-effective solution for distributing that energy. Qiu’s team’s machine learning model has shown that these calculations can be performed 12 times faster than without AI – reducing the time needed for a run from nearly 10 minutes to 60 seconds. Considering that system operators perform these calculations multiple times a day, the time savings could be significant.
Currently, Qiu’s team is also developing a power outage prediction model that takes into account factors such as weather, geography and even the income levels of different neighborhoods. It can then use this data to reveal patterns, such as the likelihood of longer and more frequent power outages in low-income areas with poor infrastructure. Better forecasts could help prevent power outages, respond more quickly to accidents and disaster scenarios, and minimize customer suffering when such problems do occur.
2. Tailored approach for each home
Efforts to integrate AI are not limited to research laboratories. Lunar Energy, a battery and grid technology startup, uses AI software to help its customers optimize their energy consumption and save money. “You have a network of millions of devices and you have to create a system that can take all the data and make the right decision,” says Sam Wevers, head of software at Lunar Energy. “Not just for each individual customer, but also for the network. This is where the power of AI and machine learning comes into play.”
Lunar Energy’s Gridshare software leverages data from tens of thousands of homes to provide information about energy use to charge electric vehicles, run dishwashers and air conditioners, and many more. Combined with weather data, this flows into a model that creates personalized forecasts about the energy needs of individual houses.
As an example, Wevers describes a scenario in which two houses on a street have solar modules of the same size, but one house has a tall tree in the garden that provides shade in the afternoon, so its modules produce slightly less energy. These are the types of details no utility could manually track at the household level, but AI makes it possible to automatically perform these types of calculations at scale.
Services like Gridshare are primarily focused on helping individual customers save money and energy. But overall, they also provide utilities with clearer behavioral patterns that help them improve energy planning. Capturing such nuances is critical to the network’s responsiveness.
3. Bringing electric cars together with the grid
Although crucial to the transition to low-emission transport systems, electric vehicles pose a real challenge to the electricity grid. John Taggart, co-founder and CTO of grid controller WeaveGrid, says the introduction of electric cars significantly increases energy demand. “The last time that [die Energieversorgungsunternehmen] “We had to deal with such growth when air conditioning came along.”
The introduction of electric vehicles is also concentrated in certain cities or even just districts, which can lead to overloading the local power grid. To reduce this problem, the San Francisco-based company is working with utilities, automakers and charging companies to collect and analyze data about electric vehicle charging.
By studying charging patterns and their duration, WeaveGrid determines optimal charging times and provides customers with recommendations via SMS or app notification on when to charge their vehicles. In some cases, customers even grant companies full control over automatically charging or discharging batteries based on network demand, in exchange for financial incentives such as vouchers. This makes the cars themselves a valuable source of energy storage for the power grid. Major utilities such as PG&E, DTE and Xcel Energy are involved in the program.
DTE Energy, a Detroit-based utility serving southern Michigan, has worked with WeaveGrid to improve grid planning. According to its own information, the company was able to identify 20,000 households with electric vehicles in its service area and uses this data to calculate long-term load forecasts.
4. Detect disasters before they happen
Several utilities have also begun integrating AI into critical operations, particularly in the inspection and management of physical infrastructure such as transmission lines and transformers. For example, poorly managed vegetation is a leading cause of power outages, as branches can fall onto power lines or cause fires. Until now, the lines have usually been inspected manually, but given the large size of the networks, this can take months.
PG&E, which serves Northern and Central California, uses machine learning to speed up such inspections. By analyzing photos taken by drones and helicopters, areas where trees need to be trimmed are identified. Defective systems that need to be repaired are also discovered. Some companies are going even further and using AI to assess climate risks.
In October, Rhizome, a startup based in Washington, DC, launched an AI system that combines utilities’ historical data on energy asset performance with global climate models. It is intended to help predict the likelihood of grid failures as a result of extreme weather events such as snowstorms or forest fires.
There are dozens of areas that network operators could address to improve their resilience, but there is a lack of time and resources to address them all at once, says Mish Thadani, co-founder and CEO. With software like his, utilities could now make smarter decisions about which projects should and should not be prioritized.
What’s next for network operators?
If AI is capable of making all of these decisions quickly and reliably, would it be possible to simply let it take over network operations and send the human operators home? Experts say: No. There are still major hurdles to overcome before the power grid can be fully automated. The most serious problem is security. Researcher Qiu explains that there are currently strict protocols and controls in place to avoid errors in critical decisions – e.g. B. when it comes to the question of how to react to possible failures or errors in the devices.
“The power grid has to follow very strict physical laws,” says Qiu. While AI is excellent at improving mathematical calculations, it is not yet ready to take into account the operational limitations and special cases that arise in the real world. This poses too great a risk for network operators whose main focus is reliability. Because: A wrong decision at the wrong time could lead to massive power outages.
Data protection is another problem. Jeremy Renshaw, management technical expert at the U.S. electric power company’s Electric Power Research Institute, says it’s crucial to anonymize customer data so that sensitive information – such as: B. the times of day when people stay at home are protected. AI models also carry the risk of introducing biases into the model that could disadvantage minorities. In the past, poor neighborhoods were often the last to get power restored after blackouts, Renshaw says. Models trained on this data could continue to assign them a lower priority.
To counteract this potential bias, the importance of employee training when introducing AI into companies is increasing, says Renshaw. So that employees understand which tasks are suitable for the technology and which are not. “You could drive in a screw with a hammer, but a screwdriver will probably work much better,” he says.
(bsc)
To home page