When the wind isn't spinning turbine blades and the sun isn't warming solar panels, how do we keep the lights on and the heat pumping?
The key to uninterrupted power is to store extra energy we produce when generation capacity is high so we can use it later, when we need it. However, renewable energy is weather-dependent, which creates challenges for managing it that we don't see with traditional energy from fossil fuels and biomass. And the best ways to harness the recent advances in energy storage technology are largely a mystery.
The good news is that a game-changing model for smoothing the peaks and dips in renewable energy demand has just landed.
Over the past 4 years, the National Research Council of Canada (NRC) and a group of international scientists have created a set of computer simulation models for electrical and thermal energy storage systems. It is part of the Energy Storage Technical Collaboration Programme, sponsored by the International Energy Agency (IEA) and led by Germany's Fraunhofer UMSICHT. This project allows utilities and other groups to simulate various scenarios and optimize usage.
According to Darren Jang, Project Manager and Systems Engineer at the NRC's ;Energy Mining and Environment Research Centre, these models are critical to assessing and designing the best-fitting solution for any combination of energy.
"Our Canadian and international partners bring vast expertise to the table that is advancing the safe, reliable and cost-effective integration of energy storage," he says. "Together, we bring the right tools, talents and creativity to bear on this challenge."
The Canadian partners include the NRC's Aerospace Research Centre and Advanced Clean Energy Program at the Energy Mining and Environment Research Centre as well as Carleton University's Sustainable Building Energy Systems group, the University of Calgary's Department of Electrical and Software Engineering, the Wind Energy Institute of Canada (WEICan) and the Office of Energy Research and Development at National Resources Canada. International collaborators come from many countries, among them Germany, Switzerland, Denmark, the United Kingdom, Austria, South Korea and Portugal.
This multinational team has made great strides toward crafting scientifically proven models and model descriptions for energy storage devices using data provided by clients as input parameters for simulations. The models are also open source, allowing users to license the source code and design documents or content. "The NRC's early and committed participation in the project was very important and taking on the responsibility as subtask leader was crucial. At the same time, the commitment also meant that an above-average number of committed participants came from Canada in particular," mentioned Professor Christian Doetsch, Task Manager of the IEA's Task 32.
An unconventional model
While Jang's project focuses on electrical and thermal energy storage in general, the NRC has developed a novel AI-driven model that can be adapted to other storage technologies. It uses machine learning techniques to develop customized models of complex storage systems based on operational data.
A recent collaboration with WEICan successfully demonstrated the possibilities. At the institute's 10-megawatt wind farm, which uses a Tesla Powerpack 2 to store standby power from wind turbines, they had no way of predicting the impact of varying loads on their energy storage system. And the collected operational data needed to make the model work effectively were also limited.
The project team met this challenge by developing a predictive model for state-of-charge using available data from the energy storage system management controller. They also developed the model training software. "Machine learning is front and centre on everyone's mind, and rather than relying on established software for training, we developed our own from scratch," says Alexander Crain, a sustainable aviation researcher from the NRC's Flight Research Laboratory. "Given the open-source ambitions of the project, we wanted to ensure that both the model and the software used in training were clear to researchers unfamiliar with the field."
Jang adds that the team applied machine learning techniques to develop an accurate model of the storage system by training it with whatever operational data were available. "Once the model has learned a storage system's behaviours, it can fill in the blanks and predict future scenarios." In ongoing research, the team will continue to improve the model's reliability and performance and compare real results with the simulations.
In another collaboration with Carleton University, the team developed models for a seasonal thermal energy storage solution. This work involved storing as much heat over the summer as possible in underground tanks, then taking it out in the colder months to heat homes and water.
Now that the model has been tested and validated, its data-driven modelling techniques can be expanded into other areas. Building on the successful collaboration between the Energy Mining and Environment and Aerospace research centres, Jang and Crain have partnered in a new virtual battery lab project with WEICan and McMaster University to advance these methodologies. This project is funded by the Office of Energy Research and Development.
They are also working with the University of Waterloo on developing a related AI-based diagnostic tool for high-voltage insulation failures or degradation in modern grid or electrified transportation applications.
In the long run, the benefits of well-managed energy storage are massive. They include economic, reliability and environmental improvements. Electricity storage will help the utility grid operate more efficiently, reduce the likelihood of partial power interruptions during peak demand and allow for more renewable resources to be built and used. And this should keep the lights burning brightly.