Automatic Energy Forecasting via API

Imbalance costs resulting from a difference in energy consumption forecast and reality are bleeding the energy industry dry. With margins under pressure, any energy player wants to get their system imbalances under control or turn them into a trading opportunity.

 

Companies look at their CIO to make this happen, but traditional approaches are not easy to integrate with IT systems. Energy forecasting requires experts to engineer predictive models and tune them for practical use. Created models need to be validated, retrained and maintained. This means a lot of human work that is impossible to automate. Or is it?

 

The process of feature engineering and model building traditionally requires specific skills and engineering time. This process is costly and not scalable because of the expertise and time needed. The need for human intervention also makes large scale forecasting systems virtually impossible.

 

TIM – Tangent Information modeller was created to solve this problem. It is a large scale automatic forecasting engine that builds models and generates forecast automatically and communicates via an API.

 

TIM’s Web Service oriented (REST and SOAP) architecture is designed to make life for system developers easy.  By a simple API call, an engineering task gets automated. Models are generated in seconds with zero degrees of freedom.

 

 

The important feature of TIMs  architecture is scalability. The highest computational complexity lies on the shoulders of TIM Engine units. If your computational demands grow, additional TIM Engine “workers” can be easily added.

 

The usual method flow is to obtain a training result from TIM Web Service and then use it to predict data. Both training and prediction are asynchronous methods, which are initiated with respective API requests. Status and results of these methods can also be retrieved periodically to see the current job state.

 

TIM Web Service also offers a combined training-prediction request. The result of this combined request is a generated prediction model and predicted values. The usual flow of the training and prediction operations can be seen in the following pictures below.

 

The API based approach to energy forecasting makes modelling scalable and automated. The beauty is in the simplicity. You send a table to TIM, you get a model back. You send a model and data to TIM, you get a forecast.

 

Watch a Demo of TIM in action

View the video below to see TIM in action. It shows TIM’s place in your data workflow, the model generation process and the resulting forecasts.