At USIO we are passionate about solving one of the world’s most pressing issues - how to change the way we use energy and how to use more renewable energy when it's there, to make sure we only demand energy when we really need it.

Icon data sources

Data Sources

From smart meters, apps (e.g. weather, traffic, geo-location), consumer profiles, home devices (IoT), traffic data
At Usio we receive incredibly large volumes of data - with 10 second intervals from every household, from our app, from the energy markets and more.
Icon machine learning

Machine Learning

Keras, LSTM, Clustering, Tensor Flow, Cloud Computing, Spark ML, Signal Processing, Neural Network, Deep Learning, Bayesian Inference
The data is feeding into the USIO machine learning models, which are primarily using recurrent neural networks
Icon energy forecast

Energy Forecast

Energy Demand, Consumer Lifestyle Patterns, Disaggregation, Smart Buying, Smart Notification, Intelligent Renewable Allocation
Usio is then able to forecast energy demand and balance the energy system – conventional energy sources (gas) as well as renewables (solar, wind).

The USIO data science team is extending the art of the possible in energy demand forecasting by developing state-of-the-art machine learning based models of consumer behaviour. We gear our energy tariffs to meet the demands of our customer’s varied lifestyles by taking advantage or recent changes in market regulations. This is not a far-fetched goal – the USIO team already have live machine learning models that cater to our business and customer needs, while simultaneously progressing towards increasingly accurate models step by step.

Meet our problem solvers

Jaye_Cribb_Profile

Jaye

“I have been working on improving our AI platform in order to train and test multiple models in unison”

Zuzana

“I’m currently working on using wavelet decomposition and dynamic time warping to improve segmentation of our customers.

From this plot, we can identify consumer patterns over different frequencies, such as daily and weekly patterns. It also offers a glimpse into how strong these patterns are.”

Zuzzana_M_Profile
Oleg_Lenive_Profile

Oleg

“Most recently I have been setting up experiments to compare the prediction accuracy of different models using historic energy consumption data.”

Ready to switch? Get a quote now