Dr Christoph Bergmeir

TIME SERIES FORECASTING

By Ms Nancy Van Nieuwenhove | 8th April 2020

Dr Christoph Bergmeir

Research Fellow, Dept. of Data Science and AI, Faculty of Information Technology

Research interests: time series forecasting, machine learning, data science, artificial intelligence, power generation forecasting for renewable energies

Christopher Bergmeir

Christoph has always been interested in computer sciences, since early age. He became interested in forecasting and time series during the first year of his PhD. “I did my undergrad in computer science at the University of Ulm (Germany) and did a research exchange at University of Alcala de Henares (Spain). I really liked Spain, so I went back there to do a master in soft computing and intelligent systems, followed by a PhD in computer sciences at the University of Granada. My PhD stipend allowed me to do research abroad and as Monash University is known for its forecasting expertise, I contacted Prof Rob Hyndman, visited the team at Monash Clayton, to end up with a job in the Faculty of IT”.

Christoph started at Monash in October 2014 as a Research Fellow, and started as a Lecturer in Data Science in January 2018. Temporarily, he is not using his teaching hat as he is on a 3 years research fellowship with the Discovery Early Career Research Award (DECRA), working on time series forecasting. Christoph is also the external engagement coordinator of the department of Data Science and AI. “When there are opportunities in collaboration with industry, my role is to involve other members of the Faculty, connect the right people together”.

Christoph is currently working on various applications of forecasting; like retail forecasting. “We would normally produce sales forecasts based on repetitive patterns and past data. For example, if you think about weekly consuming patterns, on Monday people might buy more than on Tuesday, or if you want to know how the Christmas sales would be, you would probably look at the last years. In March 2020, with COVID-19, people have been buying a lot more and very different items than usual, resulting in shops having emptied shelfs for things like toilet paper, pasta or flour, because of the unexpectedness of the situation and the time needed to manage the change.”  

Christoph’s interest in the energy sector came naturally as in this sector lie some of the main applications of time series forecasting. He is also currently working on energy demand forecasting and renewable energy production forecasting. Energy demand and energy production are time series. A time series is something you measure many times, usually with a constant time interval, like hourly, daily, weekly, monthly demand, production, etc”.

If we look at hourly energy demand, “we can see that people are consuming more energy in the morning when they get up and get ready for work, it slows down when they leave to work and it peaks back up when they come back home, adjusting their home temperature, cooking, etc. The clear power demand curve varies daily, weekly and seasonally. Now with COVID-19, as people are working more from home and especially as winter is coming, that energy demand will change. Power producers will be able to deal with this, but households may face higher energy bills”.

If we look at energy production, “we can see a difference between fossil fuel and renewables production. If you run a fossil fuel power plant, you can react to a higher demand by burning more fossil fuel. With renewables, we depend on the day/night pattern, on weather conditions and the storage capabilities. We cannot just produce more energy on the spot, it depends on how much sun is shining, how much wind is blowing. And if you know that in five minutes there will be a shortage in wind or sun energy, then you can switch to another power source. Here, we usually use minutely data, to do the 5-minute-ahead forecasts. If a power plant contributes to network instability by producing more/less energy than their forecast, they get heavily fined. They have to stabilise the network, usually by paying others to produce or consume larger amounts of energy on short notice.”

Sustainably integrating wind and solar power into the national electricity grid requires precise forecasting of power output from wind and solar farms. Wind farms and solar farms can submit their own forecast to the grid and that major change is currently happening. Christoph is part of an ARENA-funded project, led by Advisian Digital in conjunction with the Faculty of Information Technology and the Monash Business School, in collaboration with Palisade Energy. “The project aims to provide semi-scheduled wind and solar power generators with more accurate and reliable five-minute ahead self-forecasting tools. By improving the accuracy of five-minute look ahead forecasts into the National Electricity Market, our solution could enable the delivery of more secure and reliable electricity to the grid."

Christoph is also one of the recipients of the RACE for 2030 CRC funding. He anticipates many interesting opportunities for Data Science, Machine Learning and AI. "Applications will be around power supply forecasting of renewable energy, demand forecasting, time series applications in network stability simulations, satellite image processing, and more. These will help to address the most challenging problems in the transition of our energy network towards decentralized energy production with renewable energy".

Christoph is also working on a project with Honeywell Ltd about energy saving in Lecture theaters“The idea is if the AC is on the whole day in lecture theatres, but there is no lecture all of the time, there is a waste of energy. It turns out to be a forecasting problem. We have implemented the method and will trial this in a near future”.

In 2020, Christoph wants to see more accurate forecasts widely used and different methodology used. If Monash wants to get to Net Zero emissions by 2030 and have this RACE for 2030 happening, two projects he is part of, he says we need to be prepared. “Ten years seems very far away, but you can very easily break it down year by year and you know what you need to do this year to keep on tracks”.

To go further: