The term “Intermittency” is often used when utility-industry experts discuss the drawbacks of renewable-energy supplies. Because the wind doesn’t always blow and the sun doesn’t always shine in parallel with our demand for electricity, utilities and merchant power generators still must keep backup supplies on hand. Equipment makers and software experts are developing tools to address at least one kind of intermittency problem, and the result could be more reliable and efficient wind and solar resources.


“Intermittency” actually encompasses several concepts. The most obvious involves the chunks of time when solar and wind resources simply aren’t available—from dusk until dawn and when the wind doesn’t blow. A different kind of intermittency is more difficult for power planners to predict: a wispy cloud passes through an otherwise clear sky or a sudden gust momentarily speeds up a wind turbine’s production. Such fluctuations, also called “variability,” can wreak havoc on transmission systems, which must respond almost instantly by either increasing or reducing electricity from other sources.


In some regions, wind-farm owners already are paying a price for this variability. Some utilities are charging these power generators for some of the cost to maintain backup “spinning” reserves (i.e., power plants that continue to operate offline at a low level, even when their electricity isn’t required, so they are available for rapid startup should a variable resource drop out). Wind experts armed with big data hope to provide producers with the predictability they need to better understand wind’s availability. They have also developed small-scale storage technology to help smooth out such quick ups and downs.


IBM recently made public a program it calls “HyRef” (hybrid renewable energy forecast). “Hybrid” is because the offering brings together data from local weather forecasts, along with inputs from sensors at each individual turbine that measure temperature, wind speed and humidity. Using this information, IBM’s wind predictors have boosted integration of wind energy at China’s 670-megawatt (MW) Zhangbei wind farm by 10 percent.


“The fact that you bring different data points together brings a lot of benefit,” said Rolf Gibbels, IBM’s leader in Global Power Generation Solutions. 


Sophisticated modeling provides predictions as far as a month in advance or in 15-minute increments. 


The Chinese installation is called a “demonstration” project because this was the first real implementation of IBM’s hybrid approach. However, at 670 MW, the Zhangbei wind farm would rank among the largest such facilities in the United States; therefore, news of HyRef’s impact is piquing interest among U.S. owners, Gibbels said. And an investment in such predictive capabilities only becomes more valuable over time, as the system begins to use its growing backlog of site-specific historical information to improve the accuracy of ongoing forecasts.


Turbine manufacturers also see great value in smarter wind production. These companies have made great strides in reducing costs and increasing the output of their products. Now, they are beginning to recognize the value accurate forecasting can bring to developers whose project costs can run into the billions of dollars.


GE has launched a new equipment line that is so smart the company has labeled its offerings “Brilliant.” Similar to IBM’s HyRef effort, the Brilliant turbines incorporate networked sensors that communicate wind speed and operating status to centralized processing software that amalgamates the data into power-production forecasts ranging from 15 minutes to 45 minutes ahead.


The twist GE adds to this forecasting capability is onboard battery storage at each turbine, with capacity ranging from 50 to 150 kilowatt-hours. As forecasts begin to predict variable wind patterns, the batteries can step in to provide short-term electricity supplies for more consistent turbine performance.


“You put all that together, and we think we will actually decrease the cost of the energy from wind turbines,” said Keith Longtin, GE’s general manager for wind products.


As with the IBM offering, the algorithms behind GE’s power forecasts can become smarter over time through the “industrial Internet,” which Longtin describes as “turbines talking to turbines, wind farms talking to wind farms, machines talking to machines.” This kind of software-driven automation will likely become more important as the U.S. portfolio of solar and wind resources continues to grow. Otherwise, the need for matching fossil-fuel-based spinning reserves to back up green resources could make our grid more carbon intensive, even as renewable generation capacity grows.