Together, IBM and the University of Michigan Solar Car Team push the boundaries of what is possible. IBM’s technology informs Michigan’s Strategy Division in the realms of data-driven forecasting and optimization. This partnership yields more effective real-time decision-making based on accurate predictions. The end result? A faster, more energy-efficient race.
In a solar race, solar energy intake is crucial. It dictates a large part of race strategy, affecting everything from speed to charging locations. When you depend on the sun for fuel, anything that gets between the solar array and the sun is a serious threat. Think of a sports car: no matter how sleek or well-designed a sports car is, if it can’t get to the gas station, it will never drive. For solar cars, weather poses that threat, so UM Solar’s strategy depends heavily on reading weather and reacting accordingly. That’s why IBM’s support is so important to the team.
IBM’s contribution was particularly notable during the 2016 American Solar Challenge, a solar race largely characterized by rain and clouds, rather than actual sun.
On the third day of the ASC, Michigan’s radar showed something threatening up ahead on the race route: a California-sized mass of clouds. Michigan found itself on the very edge of another storm, too, a fast-approaching one filled with thunder and lightning. The team could not risk having solar car Aurum get caught up in a storm like this; not only was it dangerous, it would mean slowing down and less access to solar energy, which would add hours onto the team’s lapsed time. Strategy had to take quick, decisive action. It had to determine when and how much to speed up in order to escape the storm.
That’s where IBM comes in.
IBM’s machine learning algorithms consolidate data from multiple sources—local weather stations, the team’s pyranometer, sensor networks, satellite observations, and more—into one coherent and readable 2D interface. The synthesis of all this data—data concerning cloud count and height, wind speed and direction, sun position, and weather event forecasts—provides Strategy with the clear picture it needs of both current and projected future solar radiation intake. This comprehensive view drives effective decision-making that cuts precious time from the team’s route.
When faced with that threatening storm on race Day 3, Strategy consulted IBM’s weather models and ran simulations; it was because of IBM that the University of Michigan succeeded in dodging the storm.
With the help of IBM’s cognitive computing technology, Michigan timed its break precisely; while the rest of the field struggled through the storm, Aurum charged ahead under bright sun—it was “the perfect strategic situation,” Operations Director Jonathan Cha says. He ascribes this success to IBM’s contribution. “The IBM model was accurate enough that we could pull off a stunt like this.”
To expand upon this, driver and Engineering Director Clayton Dailey steps back and considers the bigger picture: IBM’s importance to the team’s performance over the duration of the entire race. “Building a fast car is one thing, but being able to predict the weather is another. This is one of the key aspects that separates the good teams from the great teams,” Clayton explains. Weather can cause a fast, well-designed car to underperform relative to its potential. In the end, strategy—human decision-making—can make or break the race. IBM takes much of the guesswork out of that decision-making, allowing for more sound, informed decisions. “Without IBM’s weather models, there is no doubt that we would not have won by as large of a margin as we did.”
To illustrate why this is the case, consider the Day 3 storm. Without access to such a precise and reliable model, Michigan would have been less equipped to dodge the storm, and may have become caught in it. This would have forced Aurum to slow down and waste time in the clouds, causing it to expend valuable solar energy. All other teams experienced this, which is a large reason why Aurum won by a record 11-hour lead—and as the only car to finish the race solely on solar power.
The reach of IBM’s influence extends even further than short-term predictions; another way IBM technology played a hand in creating this 11-hour, purely sun-derived lead was with its long-term predictions.
How does it work? It’s similar to the way IBM’s technology facilitates strategic decision-making in the short-term. Yet again, instead of leaving Michigan to sift through a data dump, IBM allows direct access to the most relevant information, harnessing the power of machine learning to combine disparate data sets into one massive model. A cognitive computer takes data from various sources, like other weather prediction models and historical forecasts, and collates them into a coherent whole that provides long-term forecasts. Long-term forecasts are useful when planning hours or even days in advance, and keeps Michigan prepared. For example, the Strategy team uses this long-term forecast when choosing prime charging locations at the end of every day.
A good charging location is very open so that when the sun rises or sets at a low angle, nothing blocks the sun rays’ trajectories to the car’s solar array. Minimizing potential shading is key, and that means finding a location with the least amount of clouds and shade as possible, as well as an ideal temperature range—(too hot and the array’s efficiency falls).
While Aurum does charge on the move, stationary charging via pointing the array at the sun at the start and end of each day accounts for a significant portion of the car’s total charge. Michigan’s IBM-enabled long-range forecast capability, then, is crucial to overall race strategy.
Thanks to IBM’s cutting-edge technology, the University of Michigan Solar Car Team has continued to outperform its competition. IBM’s innovative approach to forecasting taps into the power of cognitive computing and machine learning, which opens a whole new realm of possibility—in both the short and long-term. It is in this realm of possibility that UM Solar now operates. Thank you, IBM.