VELCO renewable energy

Data science and extreme automation key to renewable energy

How weather forecasting, data science and machine learning tackled VELCO's renewables intermittency challenge
Published: Thu 11 Jan 2018

Vermont Electric Power Company (VELCO) had a classic problem. They had wind farms operating in the northern part of the state, yet the major population centres are in the south. They ran into congestion problems, causing the wind energy to be curtailed. They had to find a solution.

Specialised weather forecasting

VELCO, the high voltage grid operator owned by the 18 DSOs across the state, called on Chandu Visweswariah and his team of data scientists at IBM Research to set up the Vermont Energy Analytics Centre.

The primary focus was on weather forecasting, or as Visweswariah calls it, “Hyperlocal, high resolution specialised weather forecasting”.

In other words, Visweswariah was bypassing traditional weather forecasting and begun trying to “track every cloud and every gust of wind”.

This level of specificity gained unprecedented single digit accuracies - 6% for solar and 9% for wind.

Extreme automation

With the enhanced weather forecasts they then created new energy demand forecasts. They then took the forecasts of energy production and energy demand, combined them with grid models and used a stochastic engine.

The engine uses what Visweswariah describes as “extreme automation”, allowing self-learning systems to thrive.

Data science and decision-making

The outcome was to help VELCO make better decisions, such as how to mitigate congestion, how to optimise storage, how to predict the probability of peaks and how to avoid peak charges.

Following the success of the Vermont Energy Analytics Centre, Visweswariah’s Utopus Insights was born as a spin out of IBM Research.

“I believe that if you take all of the data available both within the utility and exogenous data (such as weather data) it can be converted into the insight necessarily to make renewable energy much more predictable.”