09-15, 10:50–11:20 (Europe/Amsterdam), Bar
Join us as we explore the complexities of balancing the electricity grid amidst the rise of renewable energy sources. We’ll discover the challenges in forecasting electricity consumption from diverse industrial resources and the modelling techniques employed by Sympower to achieve accurate forecasts. Gain insights into the trade-offs involved in aggregating data at different hierarchical levels in time series forecasting.
The shift to renewable energy sources presents a major challenge for the electricity grid: solar and wind facilities are constantly varying in power output, making it harder to keep the supply and demand in balance. This creates a need for demand response: strategic activation or deactivation of large industrial resources to balance the electricity grid. Reliable demand response requires an accurate forecast of industrial electricity consumption, to get a clear understanding of which resources can be controlled at what time.
In this talk we will discuss the challenges faced when forecasting electricity consumption from industrial resources from different kinds of industries such as furnaces, greenhouses or paper mills. We’ll discuss the different modelling approaches for predicting time series including regression, forecasting and deep learning, and we will discuss the suitability of each in different scenarios. Using the forecasting of electricity consumption of industrial resources as an example, we show how we make our forecasts at Sympower to help balance the electricity grid.
Finally we will discuss a trade-off in forecasting: Trends and seasonality often only emerge at aggregate levels, making forecasting at the aggregate level easier. On the other hand, business often requires precision-level insights. Aggregate data is inherently less noisy since the errors tend to cancel out, but also might fail to capture lower-level details. We will discuss the considerations to make when forecasting at different aggregated levels in time or across groups, and what you could do to forecast consistently across different aggregate levels..
KEY TAKEAWAYS
- Gain insights into selecting the most suitable modelling technique for your forecasting need
- Understand the challenges posed by the evolving electricity grid and the significance of demand response
- Explore the trade-offs involved in aggregating data at different hierarchical or temporal levels in time series forecasting
No previous knowledge expected
Rik is a machine learning engineer with a strong foundation in electrical engineering and a specialization in leveraging electricity data for smart use cases. With previous experience at Eneco, he has focused on delivering automated home energy insights to large group of customers. Currently, Rik is dedicated to constructing a scalable forecasting model for a sustainable electricity grid, combining his passion for data science and sustainable solutions. He thrives on creating value and generating insights from raw data, demonstrating his proficiency in building robust and scalable data pipelines using Spark and Python.