Spotlight on a student: Kevin Yang
Getting wind of change to come
Kevin Yang is a budding data scientist with a passion for renewable energy. He was able to bring the two interests together earlier this year via the Melbourne Energy Institute’s Zero Emission Energy Laboratory (ZEE Lab) internship program.
On a placement at Meridian Energy’s Mount Mercer wind farm, Kevin delved into on-site data and developed algorithms to improve power generation forecasting, helping the clean energy generator make the most of its potential in our energy mix.
We asked Kevin to tell us more about his experience as a ZEE Lab intern, and his plans for a number-crunching, clean energy future.
Can you tell us about your ZEE Lab internship project and your role in it?
The goal of my ZEE Lab internship project was to build and improve models for wind farm power forecasting. I was placed as a research assistant with another ZEE Lab intern, Erdi Gao, and together we worked on building and exploring approaches to forecasting for Meridian Energy’s Mount Mercer wind farm in Western Victoria.
Forecasting is important for renewable energy generators, because it’s how they are able to participate in the National Electricity Market. Every five minutes, generators bid how much energy they can supply to the mix, and at what price. The Australian Energy Market Operator decides financial settlement and operational dispatch based on these bids, which are essentially generation forecasts.
Previous work through an ARENA-funded project with the Melbourne Energy Institute developed an advanced method to provide 5-minute-interval forecasts for wind power using energy data at the farm level – this was a big step up from existing methods, which had made forecasts based on all wind and solar data from across the entire National Electricity Market.
But on-site data can provide an even closer picture – 5-second-interval data is recorded at both the farm level and for each individual wind turbine, providing a potential fine-scale basis for more accurate forecasting, helping wind farms participate in the market and contribute more to our energy mix. The question is, how do you sift through that much information and make sense of it?
Erdi and I built new algorithms that could efficiently process years’ worth of 5-second-interval wind turbine data for all 64 wind turbines at Mount Mercer. We utilised various time-series clustering approaches, developed features based on physical farm attributes, and implemented different machine-learning algorithms to improve the accuracy of forecasts.
Through our work, we learned a lot about the complexity of wind power forecasting. Nonetheless, we did achieve our goal of improving wind power generation forecasting. Future researchers working on this problem can learn from our findings related to the physical and temporal scales of wind propagation to improve their forecasts.
Why did you want to be part of ZEE Lab? What have you learned from the experience?
I was interested in the ZEE Lab because it aligned with my previous work in sustainable energy, and it provided me with an opportunity to work with a real-world high-resolution dataset. There are very few internship opportunities that allow you to both work on an applied problem as well as being under the guidance of world-class researchers in the field. From this experience, I have learned about the challenges of wrangling energy data, the complications and intricacies of the energy market and forecasting systems, as well as different time-series forecasting algorithms.
What did you study to get here?
I am currently enrolled in a Master of Computer Science, and previously completed an undergraduate degree with a major in Mathematics at the University of Melbourne. The data science and computing skills developed through these studies, coupled with my experience as a consultant in renewable energy, made me a great fit for this research project.
What’s the bigger picture? How will your work contribute to the transition to a clean energy system?
I am very passionate about the transition to a cleaner energy system and hope to contribute to this using my data science capabilities. This could involve improving forecasting systems for renewable generators of electricity, helping them reduce charges related to ancillary services. There are currently great opportunities in the market utilising both large-scale renewable generation and battery storage technologies. The optimisation problems related to both power generation and market pricing are very exciting, and I am particularly keen to help solve them. This would enable both cleaner electricity generation as well as better economic outcomes for renewable generators. In turn, this provides more incentives for organisations to fund renewable generation and accelerates Australia’s transition to a zero-emission economy.
What do you want to do next?
I will continue to develop my data science capabilities and hope to find roles in the energy industry working on various forecasting problems. On the side, I will continue to develop software projects making use of publicly available energy data.
Further information
Kevin is happy to answer questions related to his experience with ZEE Lab, and can be contacted at kevinsinghoi.yang@unimelb.edu.au.