There is a storm coming. How can you prepare?
When facing a storm, machine learning can be the difference between storm damage and storm preparedness. As utilities head into storm season, machine learning can help them better position themselves to handle storms of all kinds coming their way.
The result is a better-prepared organization that minimizes damage during storms! In addition, your power restoration process will be more efficient, which means safer and happier customers.
Leveraging machine learning can help you do all this and more. For example, with Storm Impact Analytics, you can optimize restoration and move beyond the forecasted weather to accurately anticipate the impact on your customers and your operations.
Let’s dig deeper into the challenge behind preparing for extreme weather, how utility companies specifically use machine learning, and what it can mean for you.
The challenges of extreme weather & utilities
Facing extreme weather is one of the most difficult challenges that a utility company will face. Unfortunately, the reality is that weather is challenging to predict. Many variables, such as rainfall, wind speed and direction, lightning, storm surges, and storm size, come into play.
A utility company’s operations plan is built on historical weather data, which can be limited since it does not consider all the variables. Also, climate change means that data collected from historical storms is limited when predicting the type of weather that may be happening today.
Another challenge that utilities face is aging infrastructure. The average age of an installed base is at least 40 years old, with most grids being over 50 years old. Managing storm damage becomes even more difficult when you couple unpredictable weather with aging infrastructure or natural disasters like wildfires or earthquakes.
Of course, you must consider the financial implications. The Department of Energy estimates that power outages cost businesses more than $150 billion annually. Utilities are already running on tight budgets, and every move must be made to efficiently and prudently utilize your resources.
On top of storm damage being unpredictable, it is also challenging to find the resources necessary to fix storm-related damage quickly while continuing normal operations in between storms.
As you know, infrastructure hardening is being required after a storm or significant outage. Decisions need to be made promptly and confidently, whether it’s deciding how many crews to deploy or whether mutual assistance is required.
Traditionally, storm preparedness relied upon the historical knowledge of employees. For example, how were storms handled in the past? However, changing customer expectations and older employees retiring means that that historical knowledge may not be accessible.
There are also risks to your utility’s customer satisfaction, regulatory compliance, and public image. Unfortunately, these factors are a daily reality for your utility.
Many utilities have responded by focusing efforts on infrastructure hardening. However, that alone is not enough. Machine learning and infrastructure hardening can go hand-in-hand to address all of these challenges.
What is machine learning?
Whether you call it machine learning or artificial intelligence, it studies how algorithms can learn from and make decisions using large data sets, allowing machines to self-improve. The more data the device has, the more it learns and the better its outputs will be.
In storm damage case management, machine learning takes the form of advanced analytics tools with predictive capabilities. Machine learning allows utilities to use real-time measurements from sensors or simulations to determine when, where and how extreme weather will affect a utility company’s infrastructure.
Using machine learning, storm forecasts, and data from the utility’s smart meters and sensors become more accurate and meaningful than ever before. As a result, operator response time is shortened because every second counts in a storm situation.
Machine learning’s power is based on its ability to adapt and learn from new data. For example, machine learning allows storm forecasts, measurements, or sensor data to be continuously fed into the system for an indefinite period.
The more information that a machine-learning system receives, the better it can forecast storm damage and make decisions in real-time.
It’s all about the data
Machine learning relies on large datasets to be able to make accurate predictions. That data must be correct and specific to your utility company. Each provider has a different infrastructure, geographic location, weather patterns, level of staff, and more. Therefore, each storm is different for every utility company.
Therefore, having that data is vital to use machine learning successfully.
What kind of data is needed? Specifically, you need:
- historical outage data
- historical weather information
- your overhead distribution system data
- location, types, and amount of vegetation
- maintenance practices
All of these pieces together will determine how a storm, whether it’s a tornado, hurricane, or severe thunderstorm, will impact your infrastructure. For example, knowing your tree trimming schedule will significantly lower the risk of fallen branches so that you can deploy resources elsewhere.
Utilities must understand the weather patterns in their area and how they affect their infrastructure before using machine learning for storm management. They also need to predict when storms will hit, where they are likely to strike, and what level of damage can be expected.
From data to a predictive model
What is a predictive model? It is a model that predicts the future behavior of an event.
How does this work? It starts with historical storm data and other relevant information, which is fed into a machine-learning algorithm. The output will be an intelligent prediction of the storm’s intensity at different points along its path.
Utilities must first figure out how reliable their predictions are. The ability of machine learning to validate the accuracy of a model before putting it into use is one of its major advantages. Cross-validation methods provide one crucial advantage of machine learning: the capacity to validate the quality of a model before deploying it.
Cross-validation divides the data into subsets and, depending on the type of cross-validation technique, uses each subset of data once or a few times as the validation set and trains on the remaining subsets.
Artificial Intelligence meets Human Intelligence.
As noted, artificial intelligence is critical for storm preparation and response for modern utilities. With intelligent and reliable forecasts, it will bolster your understanding and allow you to do what you do best – make prudent and well-thought-out decisions.
Storm Impact Analytics provides all of the artificial intelligence you need, along with human intelligence. With its rating as the top forecaster for ten years, combined with insights from meteorological experts, your utility can confidently face the storm, minimize its damage and keep your customers safe.
Learn more about how to add this tool to your arsenal to face whatever storm may come your way.