Load Forecasting by Dexter - Video

Dexter offers custom-made electricity load forecasting for the short- and long-term trading cycle. The focus lies on day-ahead and intra-day, forecasting multiple load categories:
Wind and solar generation
Profiled customers such as households
Telemetry customers such as industrial plants
Water companies and
Prosumers, consumers with a PV installation

The forecasts are highly accurate because of the large amounts of data being collected. For example, Dexter processes more than a handful of numerical weather models. In the intra-day range, these models are augmented with nowcasting techniques by applying deep learning to satellite imagery.

When processing weather models, the challenge is to process all these models and find the best input parameter for a given forecasting application. In some cases, Dexter has developed a machine learning model on top that combines weather parameters to further boost accuracy.

When choosing the right model, our team looks at the application and the available data. From a case-to-case, physical and advanced machine learning models are benchmarked to find the best model for the job.

Customers choose Dexter's load forecasting for three reasons:

The first is the accuracy leading to lower imbalance costs
The second is we are more affordable. For our customers, buying and processing weather data, building data pipelines, and operationalizing forecasting models is too costly
To give the user more assurance our forecasts come with a software tool showing correlation with input parameters.