Big data offers big opportunities for wind energy
Armed with the tools of computer science, Lei Yang aims to tame the world's inherent randomness
Lei Yang wants to predict how much the wind will blow. It sounds like an impossible task, but predicting the impossible is at the heart of Yang’s research, which uses stochastic optimization and data mining techniques – approaches designed to solve problems with random variables – to tame the randomness in the field of renewable energy.
Lei Yang joined the College in 2015 as part of a University-wide effort to hire faculty with expertise in big data.
Yang, an assistant professor of computer science and engineering, focuses his current research on cyber-physical systems, with a particular emphasis on the smart grid. He’s interested in wind because while wind power is a cheap, renewable, eco-friendly energy source, its inherent unpredictability makes it hard to integrate into the power system.
Power companies engage in a delicate balancing act, day in and day out, to balance energy demand and energy supply. The grid currently operates on a 60 hertz frequency – too much energy or not enough can result in not only economic inefficiencies but also blackouts and brownouts.
“In the power system, the most important thing is to maintain the balance between demand and supply,” Yang said. “As we have more and more renewable energy integrated into the system, all these things will make the system more random. To integrate them into the power system, we need to know their output.”
Integrating renewables into the power grid
While renewable energy sources such as wind and solar are able to generate significant amounts of power, the imperative of grid stability makes it challenging to integrate renewables into the power mix. To mitigate against the uncertainty of a sudden increase or decrease in wind energy, power companies rely heavily on backups.
“Although we have a lot of renewable energy, you may see that your bill is still high,” Yang said. “This is one problem we want to solve – to make it more predictable and controllable so that overall we can make our environment cleaner and at the same time we can lower our bill.”
Integrating wind, and other variable renewable sources such as solar power, requires developing algorithms that can improve the accuracy of wind forecasting, relative to existing techniques.
Yang’s research relies on massive data sets from wind farms. He uses data from a large wind farm, with around 300 wind turbines, where the output of each turbine is measured every 10 minutes. Additional sensors collect data about wind direction, wind speed, temperature, humidity and other environmental factors. Yang sifts through these data using stochastic optimization and data mining techniques to identify patterns.
“Based on historic data, we forecast the renewable generation,” Yang said. “Also we find that since we have multiple windfarms and they are spatially correlated, we have spatio-temporal correlation. My current research is trying to characterize the spatio-temporal correlation structure such that we can better forecast the renewable energy generation.”
How does the smart grid benefit you?
By incorporating sensors throughout the grid, consumers can see lower prices and better control their own energy usage.
Only about 5% of our power supply currently comes from wind, but the Department of Energy wants to raise that number to 20% by 2030.
Better wind forecasting techniques allow energy companies to integrate more wind into the power supply by reducing uncertainly about how much power the wind will generate.
The current electric grid is 99.97% reliable, yet outages still cost at least $150 billion each year.
Balancing energy supply and demand is critical to preventing blackouts. Sensors deployed throughout the power grid provide more accurate, timely information about power generation and consumption, enabling better balance for optimal efficiency.
A trial program conducted by NV Energy saved participants $1.5 million over two years by shifting electricity use to more cost-efficient, off-peak hours.
43% of homes in the U.S. have a smart meter installed. Combined with smart pricing, which currently reaches only a fraction of consumers, smart meters enable you to monitor your energy usage and lower your bill.
Yang is also interested in optimizing the randomness in the power grid from the demand side. A smart grid relies heavily on smart meters that can send a lot of information back to the grid about power use. In search of an ever-balanced power grid, an oversupply on the generation side can be dealt with by incentivizing power use, likely through cheaper energy prices when supply is high. A home equipped with smart sensors and a power management device could optimize energy use by charging an electric vehicle or running the dishwasher when the grid signals prices are low.
However, sending that information about demand back to the grid raises privacy problems. The energy use profile of a particular home can deliver a lot of information about when people are working, sleeping, cooking or using particular devices. Consumers who are concerned about privacy are reluctant to let utilities access that kind of personal data. So Yang is working on developing a battery-based solution that could work in tandem with smart meters to charge and discharge electricity, cancelling out the signals coming from a particular energy user and sending a flat line back to the power company.
“Your behavior is random, so it’s a stochastic optimization problem,” Yang said. “Everything is stochastic. I like everything I’m working on. It’s very exciting and interesting.”