CARLOS F. COIMBRA knew from the outset that he would have to crack the code of clouds. As an engineering professor new to the University of California’s campus in Merced, he led a successful drive to get 15 percent of the school’s power from an array of solar panels.
But clouds, wandering and capricious, had foiled his efforts on two occasions by casting sudden shadows, forcing the school to rely on conventional power instead. To neutralize the clouds, he would have to track them.
So Professor Coimbra, a Brazilian-born expert in fluid mechanics with a flair for computer modeling, tried a new kind of forecast. The campus would make better use of sun power if he could figure out exactly when a puffy drifter would arrive overhead. He wrote a computer algorithm to project how clouds move and change shape as they move across the sky — one of the most complex and chaotic phenomena on earth, influenced by an endless set of variables.
Now, six years later, Professor Coimbra, 44, and his collaborator, Jan P. Kleissl, 37, have created a forecasting engine that they say is 20 to 40 percent more accurate than the model in common use. Weather, energy and power grid experts said that the innovation could accelerate the adoption of renewable energy, save billions of dollars in energy costs and help turn cloud-watching from an idle pastime into a vital and profitable part of the weather forecast.
“I can’t tell you what’s going to happen at 4:23 p.m. on a Sunday,” said Professor Coimbra, whose forecasts extend to seven days but with decreasing accuracy. “But I can tell you what will happen between noon and 6 today.“
Potential cost savings are drawing the interest of companies that build and operate solar-power plants, as well as utilities and grid operators. Each bets a bottom dollar on when the sun will come out tomorrow. A fine-tuned forecast makes it easier for utilities and grid operators to use the sporadic power of sun and wind when they are available, giving renewable energy a reliability close to that of a fossil-fuel or nuclear power plant.
Furthermore, it could help utilities predict exactly when homeowners will turn on their air-conditioners in the summer, which could reduce the power grid’s need for backup power plants.
As it saves money in energy markets, the technology could also shake up the world of weather forecasting by providing greater resolution. Such data could give airports a firmer window of when storms will arrive and leave, resulting in fewer flight delays.
It could tell farmers when to expect the first frost, or when a rainstorm will hit, reducing the need to pump water for irrigation. A precise prediction could guide the maneuvers of forest firefighters, project the path of bioterror attack or pinpoint the path of a tornado.
Melinda C. Marquis, the renewable energy project manager at the government’s Earth System Research Laboratory in Boulder, Colo., speculated that this technology, born to serve renewable energy, might end up changing our relationship with weather. “Any improvement that we make here for renewable energy will be very, very important, in some cases more important, for other sectors of the economy,” she said.
But the forecasts are likely to find their first application at solar and wind farms. Currently, the caprice of weather makes electricity more expensive for producers and consumers of utility-scale renewable power. Traditional weather forecasts aren’t accurate enough to predict when the sun will poke through the clouds on a partly overcast day and make mistakes in estimating the length and strength of high winds.
To compensate, some solar and wind farms maintain large, expensive banks of backup batteries to store surplus energy and release it when needed. Grid operators scramble to buy power on the spot market when weather-related energy sources fall short, buying power for 10 to 100 times more than they would if they bought a day ahead, according to Manajit Sengupta, a scientist who leads solar-forecasting efforts at the National Renewable Energy Laboratory. A perfect forecast for wind, if it represented 20 percent of the power supply, would save $1.6 billion to $4.1 billion a year, according to several studies.
The spikes and troughs of wind and sun, known in the power industry as “ramp events,” cast another shadow for plant operators. Abrupt changes can damage equipment and cause backup batteries to fill or drain too quickly, shortening their life spans, and, with them, the useful life of the plant.
The test sites that Professor Coimbra and Professor Kleissl use are small affairs, made up of a few small, spindly weather instruments that reside on rooftops and alongside solar arrays at facilities in Washington State and throughout California, from Davis down to San Diego.
The most important instrument is a fisheye camera that points at the sky, taking photos of five square miles every 30 seconds. This device, engineered by Professor Kleissl, tracks cloud speed and creates a forecast for the next three to 20 minutes.
For longer time frames, Professor Coimbra’s number crunching takes over. A self-learning computer algorithm, it digests dozens of measurements — solar irradiance, wind speed, satellite images, soil moisture — and sorts out which are relevant and which are noise. Such modeling has been used to predict the behavior of other complicated systems, like the interaction of molecules or the swings of the stock market, but it has never been used to create hyperlocal forecasts.
In the two years since Professor Coimbra moved to University of California, San Diego and started a partnership with Professor Kleissl, their team of graduate and postdoctoral students has grown from seven to 25 as research money comes in.
The California Public Utilities Commission issued a $1.5 million grant to improve forecasts of San Diego’s ever-changing coastal fog, and the California Energy Commission gave $1 million to help Professor Coimbra develop an accurate model of the thick, ground-clinging fog of the Central Valley.
The two researchers are awaiting approval of projects to model the weather above the military’s large solar installations in and near the Mojave Desert, where the North American monsoon can create violent summertime storms.
So far, however, the only commercial application is at a 22-acre field of kiawe trees in Hawaii, at the site of a proposed five-megawatt power plant in Kalaeloa, Oahu. The builder, Scatec Solar North America, hopes that an accurate profile of the “solar resource” will compensate for the island’s vulnerabilities: bright sun, sudden squalls and a small, outdated power grid that cannot handle large spikes in the power supply. An accurate forecast would eliminate the cost of backup batteries and make it easier to find investors, the chief executive, Luigi Resta, said.
Back in California, the California Independent System Operator, which manages the state power grid, is about to compare the forecasts of the University of California, San Diego, against its own, a model based on temperature, time of day and historical records of energy use that has not been tweaked in decades. If the San Diego model is more accurate — say, correctly guessing that a region uses 900 megawatts of electricity on a hot summer day instead of 1,000 — “then that’s some money” that will be saved by ratepayers, said Jim Blatchford, the smart grid solutions manager.
Professor Coimbra’s prognostications could even give racecar drivers the inside track. A private company, Forecast Energy, has hired him to create a model that could, for example, inform a driver’s team during inclement weather exactly how wet the track will be for the next 10 laps — data that could determine the crucial decision to use slower, safer wet-track tires or faster, dry-track tires to win the race.
How solar forecasting will be used as it matures is anyone’s guess, but it could be handy for anyone who needs a more exacting snapshot of the weather — like Professor Coimbra himself, who keeps a careful eye on distant thunderheads as he rides his motorcycle along Interstate 15. As he does so, he ponders the sort of question that perplexes farmers, pilots, backyard grillers, and anyone else whose fortunes depend on the whimsical schedule of clouds.
“Is the rain going to get there before I get to Point B?” he asked. “Or can I go 50 miles more?”