A greener future requires, in part, more large-scale energy storage systems. The most common system is pumped-storage hydropower, in which excess energy is stored by pumping water to a higher elevation and releasing it to drive a turbine when energy demand rises. Pumped storage currently accounts for 96 percent of utility-scale energy storage in the United States.
But what if pumped storage could be combined with a desalination technology like reverse-osmosis water purification—in which water is pushed through a membrane under pressure to separate out particulates and produce potable water—to get more benefit from the entire system? Researchers from Cornell University recently used machine-learning models to understand how such a system could operate.
The researchers’ models focused on combined pumped-storage hydropower and reverse-osmosis systems that use seawater, because large volumes of water are necessary to make the system practical. In the combined system, the water pumped to a higher elevation for storage can also be fed into the water-purification component, which uses the resulting water pressure for reverse osmosis. A combined system could supply both the electricity and drinking water needed for a coastal community, while saving on the costs needed to build both systems separately.
“There’s a potential for a 16 percent decrease in the break-even time if you do a combined system,” said Maha Haji, assistant professor of engineering at Cornell and one of the study’s authors. “So the sharing of infrastructure can already reduce a number of costs without even considering the operations.”
One of the optimal operating conditions that favored producing more potable water could meet the electricity demands of 1.66 million people and the freshwater demands of 11.6 million people.
Creating an accurate model is important because researchers eventually want to build a real-world demonstration project. Construction would be expensive, so the models can develop the best possible design and optimize its operations before breaking ground.
While the pumped-storage component is relatively straightforward to model, modeling reverse osmosis is much more challenging. In pumped-storage hydropower, the energy output is simply dependent on the turbine efficiency. But many factors determine the potable water output of a reverse-osmosis system, including the pressure, flow rate, water salinity, and type of reverse-osmosis membranes that are used.
Existing techniques for modeling reverse osmosis don’t capture the full complexity, Haji said. To get better results, researchers used a blend of existing models and learning from experimental data with different reverse-osmosis membranes. Haji and graduate student Matthew Haefner used neural networks, a type of machine-learning technique that can be trained on experimental data, to extract insights from the data. Using this combined method, they were able to calculate the expected output flow rate from inputs like pressure, salinity, and input flow rate.
The researchers were able to use the model to find optimal operating conditions for the system, such as how much electricity and how much freshwater should be generated to maximize benefits. Haji said there is a trade-off between how much of each is produced from the system, because water used for one cannot be used for the other. They found that strongly favoring one operation over the other, regardless of which one produced more, was more cost effective than splitting up resources evenly. One of the optimal operating conditions that favored producing more potable water, for example, could meet the electricity demands of 1.66 million people and the fresh water demands of 11.6 million people.
Haji said the model was also useful for determining the optimal operations to reduce waste-treatment costs from the reverse-osmosis process. Reverse osmosis leaves behind a salty brine that can be released back into the ocean after being mixed with seawater. But laws govern the level of salinity that’s allowed, so researchers used the model to calculate how much the pumped-storage hydro component needs to operate so the salty brine can mix with enough seawater to reach an acceptable salinity level.
The work needed to fully model a combined pumped-storage hydropower and reverse-osmosis system is not yet complete. Haji said future work will focus on the pretreatment process for water purification, which involves removing major contaminants from the water. Current pretreatment methods all operate at low pressure, but reducing the pressure before the seawater reaches the reverse-osmosis system would be highly inefficient because one of the major benefits of the combined system is the pressurized water for reverse osmosis. Haji is interested in developing methods to perform pretreatment at high pressures, which could make or break the overall feasibility of the project. If it’s successful, the next step might be looking at potential sites for demonstration builds.
Source: Spectrum
