BHUBANESWAR: A software developed by a faculty member of Sambalpur University Institute of Information Technology (SUIIT) has been helping the TP Western Odisha Distribution Limited (TPWODL) to bring down power cuts, fluctuations and manage electricity in Western Odisha efficiently.
Under the Odisha University Innovation and Incentivization Plan (OURIIP)-2020 Scheme by Odisha State Higher Education Council, 34-year-old assistant professor (computer science and engineering) of SUIIT Sibarama Panigrahi has created a load forecasting software for getting two days ahead electricity load trend data for the entire region.
Panigrahi, also an engineer who specialises in machine learning, said the software is assisting TPWODL in knowing how much electricity is required on a daily basis for the region and accordingly, it can either buy or sell excess electricity.According to reports, the region requires at least 1,800 MW of electricity daily which is sourced from Odisha Power Transmission Corporation Ltd.
He added that since the region is an industrial hub with rural and urban consumers, electricity load forecasting for Western Odisha had remained challenging so far. “Traditionally, investigators depended on physical and statistical based forecasting models that used unit consumption and load density from buildings, industries, etc. and geographic distribution of consumers to predict the electricity load. But in the current times as these structures are being changed or upgraded frequently, these models are neither reliable nor efficient,” said Panigrahi who is a BPUT and VSSUT alumni.
Realising the need for a better forecasting model, Panigrahi as a research fellow under the OURIIP scheme proposed the council to work in this direction. He decided to use machine learning (data driven) based models that can approximate any nonlinear function to a desired level of accuracy, for the purpose. Subsequently, TPWODL signed an MoU with the SUIIT in February this year to implement the project that would forecast the next two days of electricity load (with a measuring interval of 15 minutes) for better energy planning.
“After studying the electricity load data of the region for the years 2019, 2020, 2021, we used a model called machine learning stacking regressor that uses algorithms to correlate multiple input variables like weather conditions and data such as meter point loading, to construct a forecast model. We used 192 stacking regressor models (1 model per 15 minutes) to build the software which turned out to be more effective than those used by the discom,” he informed.
Since August this year, TPWODL has been using the software to predict two days ahead electricity needs of the region. As a result, the discom officials said, they are able to maintain a balance between actual power consumption and daily power drawal, reduce deviation of electricity frequency besides, ensure power scheduling.
Currently, Panigrahi and his team are working on further improving the efficiency of the software by considering external factors like holiday effects, temperature and humidity of an area. The team has also been approached by discoms in other regions of Odisha to build a similar software.