Encouraging collaborative forecasting: data privacy and monetization

Carla Gonçalves is currently a postdoctoral researcher at INESC TEC, working on renewable energy forecasting. Involved since the beginning of the project, her work contributes to advancing the state of the art on data privacy and data markets.

There is work using geographically distributed data produced and collected by energy companies, in order to improve renewable energy sources (RES) forecast. However, such work has assumed that data could be gathered centrally and used without restrictions. This is not in line with the current practice where data is collected by multiple data owners that, due to business competitive factors and personal data protection concerns, might be unwilling to share their data, despite the many potential benefits.


For agents to perform collaborative forecasting models, two important properties are required: that data privacy is preserved, and that data owners are not allowed to free-ride on others and are compensated for the data they contribute (data monetization).

Current work has limitations when it comes to RES forecasting. Collaborative learning focuses mainly on data split by observations; it does not work for geographical data where each data owner is looking at different features. The case is more challenging when data privacy is a concern: collaborative learning on data split by observations requires only exchanging coefficients, but for data split by features the literature requires exchanging data.

Encouraging collaborative forecasting: data privacy and monetization – Credit: Carla Gonçalves

Data markets have also received increasing attention in the literature but with a focus on the end-consumer, not on data markets for the purpose of collaborative learning.
Within the Smart4RES project, Carla contribution is in advancing the state-of-the-art, especially with regard to data privacy and data markets, so that different agents can cooperate in RES forecast.
You can find Carla’s publications in the Smart4RES Resources Center.

More about Carla Gonçalves

On her background
Carla graduated in Applied Mathematics (University of Porto). From 2015 up to 2019, she has been involved in a wide range of energy forecast consulting collaborations between INESC TEC and the industry — a large part of these collaborations was developed in parallel with her Ph.D. studies, which started in 2016.
Since 2019, she has been focused on the Smart4RES project. She has completed the Ph.D. in Applied Mathematics in 2021, under the motto of renewable energy forecasting, and she is currently a postdoctoral researcher at INESC TEC.

On her research interests
Statistics applied to renewable energy forecasting; privacy-preserving federated learning; data monetization.

On choosing Smart4RES for her research
“Smart4RES project was a great opportunity to continue my research on RES forecast.”

Where she sees herself in 5 years
“Not yet sure, but hopefully contributing to challenging problems in the RES forecast.”

On Mentors with an impact
“Two mentors have marked the course and quality of my research work: Ricardo Bessa (INESC TEC) and Pierre Pinson (then at DTU).
The experience of Ricardo Bessa in academic and industrial research, as well as his creativity and critical opinions, were essential to identify research gaps and overcome the difficulties that have arisen throughout my research activities.
Pierre Pinson is a reference on forecasting and he received me in his research group at DTU for 6 months. His keen eye and his tough criticisms helped hone my research, in particular, translating what I was doing into writing.”