Every year, utilities the world over face challenging weather conditions, often finding themselves either underprepared for outages or deploying resources in inefficient ways.
In the modern world of 24 hour news, digitisation and always-on media, utilities are perhaps better positioned to respond to extreme weather than has previously been the case.
Numerical Weather Prediction (i.e., weather models) skills have improved significantly over the past twenty years, and recent advances in computational processing have brought cloud-resolving weather models into the mainstream, which results in a significant bump in skill with regards to predicting small-scale weather phenomena that impacts outages around the world.
In spite of being warned of its impending arrival however, the recent 'Beast from the East' storm across the UK caused widespread service disruption and negative reporting across mainstream press.
How could utilities have planned earlier for such disruption? What were traditional forecasting mechanisms missing?
With more accurate IT-driven data solutions, The Weather Company predicted such disruption as early as in January 31st, something that traditional forecasting methods would not be in a position to do.
“The real difference is that as a starting point we use a number of different models to get the predictive data. The typical news channel will not invest in the technology that will get them that 3-6 week schedule. For the Beast from the East, we used a number of in-house developed tools that monitor stratospheric and tropospheric interactions,” says Dr. Michael Ventrice, Meteorological Scientist at the Weather Company.
He explains, “in the winter, there’s a vortex that spins over the north pole every year, and some years it breaks down much earlier - this is called a sudden stratospheric warming event. We saw a potential for this four weeks in advance, based on our statistical models, making us aware of the storm ahead of its breakdown.”
Utilities, however, were unable to prepare. In London alone, an estimated 20,000 customers were left without water. Depending on the duration of their outage, customers could be entitled to up to £150 of compensation.
It is smaller events, however, such as storms and flooding that perhaps pose the greatest loss to utilities.
Combined with ever increasing customer expectations for service reliability, this is forcing utilities to look more carefully at their preparation and response to weather events and limit downtime when outages occur.
Simply knowing what’s coming may not be enough, so how could actionable predictive analytics re-shape outage responses?
Network response to weather incidents
In its white paper, “Utility Outage Prediction 2.0”, The Weather Company and its research affiliate Zpryme unveiled that weather-related outages cost the US economy between $18-33bn each year.
In addition to this, 89% of utilities agree that adverse weather affects their operational decisions.
Despite this, many will rely on guesswork, or on unsophisticated public reports from local news channels.
Brandon Hertell, Offering Manager - Meteorologist at The Weather Company, says “in the current paradigm in which utilities currently exist, they are reactive rather than proactive. They don’t have their own internal meteorological knowledge or the analytical power to relate the weather forecast to exactly what is going to be impacted, how it will be impacted or where it will be impacted.”
Typically, network operators respond to weather-related incidents by mobilising teams of responders to assess, report on and repair potential damages or interferences in service.
This has lead to mass disruption and in many cases, avoidable delays and costs in fixing problems.
Further research in the white paper indicates 77% of utilities that frequently mobilise find themselves regularly overestimating the amount of materials needed.
Hertell has seen this first hand, saying, “we worked with a large investor-owned utility in the US which had a huge problem with over mobilisation. We related that cost to be about $8m a year from their operational budget.”
According to Hertell, this cost typically accumulates from two errors - mobilising for storms and outages that never happen, or over mobilising for storms that don’t need so much resources.
“In one instance with that same utility, there was bad weather predicted on a holiday Monday. The utility, being proactive and cautious, knew they wouldn’t be able to get people in, so they began mobilising. That mobilisation alone cost $3m over three and a half days, just to cover all possibilities,” explains Hertell. “Had they used our predictive technology, they would have known that Friday, Saturday and Sunday were blue sky days and mobilised more intelligently.”
Although weather analytics are valuable on your worst day of the year, the model is also useful on smaller, more frequent weather events that happen season on season.
Hertell estimates that mobilising decisions for more routine weather events can cost anything from $10,000 to $50,000 a day. Over a year that adds up, and can cost just as much as those single, extreme events.
Actionable data and outage prediction
Despite this evident business case, adoption of these more intelligent weather analytics is not yet widespread.
Research has shown 65% of utilities are relying only on local news and updates, and 43% operating only in a reactive manner. The Weather Company has, as such, created a predictive model which combines big data and machine learning to provide utilities with safe and efficient outage management processes.
Other research conducted by the Weather Company indicates utilities could save between $3-6m per year with more intelligent weather prediction and their Outage Prediction application.
“In the current world, there are many long-term employees sitting at utilities with a wealth of knowledge regarding how utilities respond to these instances. We don’t want to discredit that - what we want to do is integrate that into the tool, using data to reinforce their decisions,” says Hertell, “These employees can look at the tool and see where, if at all, the differences lie between what is remembered and what is predicted. It is a way to augment the internal knowledge of utility employees to help support them to make better decisions.”
By using a data-driven, analytical solution, utilities can make evidence based and unbiased decisions as to how to prepare for an upcoming event.
As technology continues becoming more intelligent, Ventrice expects these insights will become more accurate and actionable. He says, “there are two ways I can see long-range forecasting of extreme weather events moving forward. On the one hand, there’s new technology which helps with better assessment upcoming weather conditions. The other, would be neural networks and deeper machine learning - taking an increasingly statistical approach to weather prediction.”
Perhaps the largest barrier in the adoption of this new technology is the necessary shift in mindset.
Hertell says, “there has always been this stigma around weather. Some utilities think that there’s little to be done about it, so why bother. Others think that if the weather forecast is wrong, they’ll be making a bad decision on a predictive tool. We won’t get everything right - any machine learning model will be wrong some percentage of the time. Even so, you can still make better decisions based on a slightly imperfect forecast.”
For Hertell, the more analytical, skilled approach to forecasting brings a multitude of benefits.
By pinpointing outage points and de-risking the traditional utility wildcard of weather budgeting, not only is downtime and outage reduced and cost saved but customer satisfaction is markedly increased. “There are many other downstream events that happen that are positive when you start to build models around storm mobilisation systems. You can improve your storm mobilisation process, budgeting processes, operational processes, your knowledge of storm hardening.”
For more on this topic register for the Engerati webinar, “How to use weather solutions to minimise network power outages,” featuring specialists Michael Ventrice and Brandon Hertell.