Zapped: The grid is on life support. Can AI fix it?

Everything you wanted to know about our fragile, ageing, and increasingly unreliable grid (and how AI may be our only hope).

grid.jpg

pxFuel

America's electric system is long overdue for an overhaul. With a 2021 American Society of Civil Engineering report finding that 70% of T&D lines are over 25 years old, it's no shock that large, sustained outages are occurring with increased frequency throughout the country. 

Last year, major outages in California and Texas were both triggered by extreme weather events, causing local power demand to exceed supply. With climate change fueling extreme weather events, plant and city managers are increasingly turning to AI technologies to predict energy consumption levels days in advance, mitigating the potential of power outage incidents and increasing overall power grid reliability.

To understand the problems facing the current power grid, which by one conception constitutes the largest and most complicated machine in the world, I reached out to Steve Kwan, Director of Product Management at Beyond Limits, which develops industrial AI for growth in a variety of industries. 

GN: Can you  explain some of the challenges  facing grid managers in the U.S. currently? 

Steve Kwan: The ageing electrical grid is being put to the test as it stretches to support a much higher volume of users than it was designed for in far more strenuous circumstances. Historically, power has flowed from large generators, such as power plants, out to the consumers. Under normal conditions, the weather does not play a large part in the decisions that grid managers have to make. 

However, the explosion of distributed energy resources such as rooftop solar, commercial solar, and wind installations are creating many more scenarios that grid managers must take into consideration when deciding on the best way to operate the grid. Now, with rooftop solar, power can flow "backwards" in certain situations. This means that the cost of generation can be affected greatly by the weather and sun conditions. 

These additional parameters significantly influence the grid manager's decisions on how best to operate the grid. As a result, there's a clear need for a more resilient electrical grid, especially as climate change experts indicate that global warming will cause  an uptick in natural disasters in the coming years. The call for carbon neutrality and net-zero emissions has also bolstered efforts to transition towards green energy, and as new regulations trickle out, the industry will have to adjust to changes in standards that could dramatically alter operations.  

GN: Within the grid management ecosystem, where are we seeing penetration of AI and ML technologies?  What's the use case here? 

Steve Kwan: AI and ML technologies are superior in their ability to learn from past behaviors and predict future needs. One of the most critical elements of grid management is the need to ensure that power is efficiently routed and available at the lowest cost. As such, it's always a game of checking that there is a sufficient supply of power to meet the demand. 

However, both power supply and demand are constantly changing. Power demand fluctuates throughout the day due to human needs and activities. Both supply and demand are influenced by weather conditions. The proliferation of electric vehicles and IoT devices is also changing power consumption patterns. On the supply side, changes in the way we produce power are making it much more difficult to match supply with demand. From decreasing fossil fuel generation to increasing renewables and climate change, grid management has become a headache both in real-time and when planning for the future.  

AI and ML technologies are entering the grid ecosystem because of their ability to learn from usage patterns and provide accurate predictions of future needs, making this technology the perfect solution for grid management. AI can analyze the droves of data being produced by factories and accurately predict when there will be an abundance of energy supply to charge batteries versus when to drain them. 

Artificially intelligent time horizon analyses will lead to dramatic improvements, allowing manufacturers to see the full picture of their operations. AI can be used to sift through all data points, including past weather conditions, the network's current state, and potential fault points. As such, power providers are using AI and ML technologies to optimize their dispatch and planning activities, predict market demands, plan for future business expansion, and more. Consumers are using these technologies to plan and optimize their power consumption behaviors to take advantage of excess power and pricing. 

As an example, consumers are deciding when they should charge their electric vehicle based on time-of-use pricing, and in some cases, they can even support grid demand by flowing power from residential batteries back onto the grid. 

GN: When you're talking about energy, you're obviously also talking about environmental concerns. What improvements can AI leverage when it comes to waste and carbon emissions? 

Steve Kwan: Today, the largest source of carbon emissions is from burning fossil fuels. Fortunately, AI technologies can be leveraged to reduce these emissions in multiple ways. AI can reduce energy waste by helping plant and city managers predict energy consumption levels for days in advance. With this foresight, producers can plan their operations for the anticipated demands and precisely generate energy. AI can also help grid managers know when to store energy or cut off into microgrids to achieve a constant energy flow and meet consumption demand. This not only saves energy but also relieves stress by putting restrictions and limitations on the grid that take line capacity and congestion into consideration.  

Along with predicting consumption levels, AI technologies are reducing carbon emissions by helping the industry transition to renewable energy sources like solar and wind. By anticipating weather conditions using historical data and impending weather patterns, these solutions can predict the availability of naturally-generated energy. This provides managers with the ability to forecast the mix of renewable energy and traditional fossil fuel energy that is required to satisfy the anticipated demand while maximizing the use of renewable energy. This also makes renewable energy sources more appealing in the energy sector, as this previously unpredictable source of power is becoming increasingly reliable.  

GN: Are there any concerns about these new AI technologies or that adoption might be happening too quickly?

Steve Kwan: As with all technologies, numeric AI technologies must be verified and spot-checked, as oftentimes, these are "black box" applications heavily dependent upon the implementation and data quality of the training data set. When human knowledge and expertise are embedded into numeric AI technologies, they are dramatically improved, which can significantly alter outcomes and generate better predictions and recommendations. 

GN: What does the future hold for AI in grid management?  

Steve Kwan: With continued improvements to AI and ML technologies, we will be able to increase the size of the data set, frequency, speed, and accuracy of model predictions. Together, this would provide better toolsets for grid operators to optimally match supply with demand in unprecedented fashions. This creates the opportunity for optimal grid and generation investments in areas that would provide the most benefits. For example, AI and ML technologies can be used to optimize the placement and installation of renewables such as solar and wind farms. Better modeling of the grid will help operators route power most efficiently and allow grid operators to lower the total cost for the consumers by minimizing congestion. The more accurate the models are, the better grid operators will be able to plan out their operations and put mitigations in place in anticipation of impending problems.

Show Comments