1536. De-biasing data-driven models in energy markets
Invited abstract in session MB-61: Advances in behavioral decision analysis 1, stream Behavioural OR.
Monday, 10:30-12:00Room: Maurice Keyworth G.31
Authors (first author is the speaker)
| 1. | Fahad Mehmood
|
| 2. | Hussain Syed Kazmi
|
| ESAT - ELECTA, Electrical Energy and Computer Architectures, KU Leuven |
Abstract
Data-driven models based on artificial intelligence and machine learning are increasingly being used in energy markets, especially for forecasting demand and ensuring supply security. However, the inherent balck-box nature of such models make them prone to number of biases and limits their explainability. This means that such data-driven models can make extremely inaccurate predictions in the worst-case scenario. In this study, we show the biased predictions of widely used models in energy markets. Using domain-informed counterfactuals for three case studies, we explore potential ways to de-bias the models and recommend possible solutions.
Keywords
- Artificial Intelligence
- OR in Energy
- Behavioural OR
Status: accepted
Back to the list of papers