A retail electric providers (REP) in Texas had to estimate the amount of energy needed to provide their customers with competitive market prices. To address this issue, a more accurate load prediction model based on machine learning and historical energy load data was developed.
Texas’ energy market is deregulated, meaning in most parts of the state consumers have the ability to choose their retail electric provider (REP) on an open market.
To set competitive market prices, REP’s must estimate the amount of energy they need to provide their customers for any given hour on any given day. Energy is traded in hourly segments on the Nymex and REP’s have traditionally relied on decades-old algorithms to determine the correct amount of power to buy for their customers. Since these algorithms generally yield conservative estimates, REP’s tend to purchase more energy per hour than required, resulting in tens of thousands of dollars in wasted overhead to the REP on an annual basis.
thinkBridge is developing a far more accurate load prediction model to help REP’s avoid the lost revenue associated with the traditional method of purchasing energy. Based on REST/HTTP-based EDI, we pull 15-year historical energy load data and analyze it using our Machine Learning service for Analytics and Prediction (BigML).
Forecasting models based on our Cloud Services and Big Data competencies are then produced for the REPs. Initial test runs have found our model to be a much more accurate forecasting tool for purchasing the correct amount of energy on the Nymex.
Our model provides REPs with a much greater ROI compared to the software traditionally used to estimate energy needs. Additionally, the depth and accuracy of our data allows us to more precisely forecast metrics like consumer load patterns and revenue generated per customer.
Read more about how we work using our accelerators, context expertise, and global delivery.
.svg){kind=small}
.webp)
