Electric Nation: deep learning for the smart grid

The Electric Nation project is helping power network operators prepare, as electric vehicle ownership in the UK doubles year-on-year. To understand how and where electric vehicles are being used, we built event detection algorithms using LSTM neural networks in TensorFlow/Keras to detect the power signatures of electric vehicles charging from substation time-series data. In this talk we will discuss the challenges of dataset generation, model building and deployment in a real-world machine learning project and share some best practices. We will also look at how active demand management is shaping the future of the smart grid and is predicted to enable savings of at least £2.2bn by 2050.

Required audience experience

An understanding of the basic concepts of machine learning would be useful.

Objective of the talk

• Learn how to detect events in time-series data using LSTM neural networks built in Keras
• Understand the techniques available to make effective use of a dataset of limited size, and control over-fitting
• Deployment of Tensorflow models at scale
• Learn about smart grids and active demand management