IndiaPowerTrading.info's New AI-powered Day-ahead Power Price Forecast Model Delivers 95% Accuracy

NEW DELHI - June 6, 2016 - PRLog -- IndiaPowerTrading.info ("IPT.info"), a fast emerging data and analytics solutions provider in the Indian power market, today announced the launch of its new price forecasting model for the day-ahead power market. Built upon the existing multivariate linear regression model, the new AI (Artificial Intelligence) model is based on the Support Vector Machine ("SVM") algorithm.

The Econometrics and Data Sciences ("EDS") team of IndiaPowerTrading.info reported that the new SVM model improved the price forecast accuracy by an average of 3% across time blocks during the period of January to April 2016. The accuracy of RTC price forecast improved by 5% over the last period.

The new model uses an algorithm based on Artificial Intelligent technique usually applied for classification, and takes into account more external parameters making it more dynamic and robust. The SVM model has resulted in improvement of price forecast accuracy to around 93-95% across all regions on RTC basis.

Speaking about the importance of this improved model and its business application, IndiaPowerTrading.info Energy Economist Chandan Kumar said, "After continuous market interaction, we realized that price forecast can make business sense only if its accuracy is over 95%. Keeping this mind, we made changes to our initial model of multivariate linear regression and come up with this updated version of day-ahead market price forecast model. This model is more granular, consisting of highly relevant input parameters, and technically more dynamic and robust. This model has consistently clocked an accuracy of over 95% when back tested on last 4 months data. With this updated version, we are confident to generate highly accurate price forecast for day-ahead spot market on a consistent basis."

Sharing details of the technical aspects of the new SVM model, Junior Data Scientist Vivek Mishra stated, "Using the power of Artificial Intelligence we have refined our day-ahead power market price model to achieve even higher accuracy benchmarks. This new model uses state-of-the-art algorithms in machine learning and is much more adaptable to the market behaviour. In order to fully appreciate the market sentiment we have included several external parameters which are stored in our databases; these external parameters aid the model to predict sudden shift in prices. We will continue our research and development on these models to raise the accuracy benchmark bars even higher by streaming in more external factors and further optimising  the current techniques."

Results of the new forecast model are published by everyday for the next day's trading session. Subscribers of IndiaPowerTrading.info can access these daily forecast reports on the website http://indiapowertrading.info/ as well as via email.