Last year, when the country started battling the global pandemic COVID-19, Andhra Pradesh’s power transmission company – Transmission Corporation of Andhra Pradesh (APTRANSCO) adopted a new way to reduce the rising power costs in the state.
80 percent of the electricity sector’s cost is from the open market, so it is important to accurately predict the next day’s electricity demand to control costs. For the past 20 years, APTRANS relied on manual methods to predict the next day’s power demand. In this method, load dispatch operators used to estimate demand based on factors such as season, weather, and holidays. Much depended on the demand of the same day. Energy Secretary of Andhra Pradesh Srikanth Nagulapalli says’ Suppose there was a demand of 9,000 MW at 6 pm on a particular day. With a maximum accuracy of 90 percent, this demand can be predicted to remain in the same quantity the next day. ‘
Officials at Andhra Pradesh State Load Dispatch Center (SLDC) knew that a digital format could increase this accuracy of electricity consumption forecasts, thereby reducing the cost of power purchase from the open market and reducing tariffs. . Forecasting the demand of electricity is a major task for SLDC, which has to prepare everyday for the purchase or sale of electricity. More or less electricity is taken from the national grid when load and demand mismatch, leading to heavy fines as well as occasional power supplies. SLDC in Andhra Pradesh decided to leverage technology to solve this problem.
He started making a list of all the factors in the state that would affect the energy consumption in the state, such as sunrise and sunset time, rainfall and humidity. The data for the last three years was collected on all these parameters to create a predictive format based on Artificial Intelligence (AI) and Machine Learning (ML), including the demand scenario during lockdown and after unlocking. Google’s open source platform – Tensorflow was used to create this format, which gave a clear estimate of the next day’s power demand in megawatts every 15 minutes.
It becomes important to forecast the next day, because they have an idea of how much power should be generated in coal plants and how much power should be purchased through the open market. This 24-hour forecast is also important because the day is divided into 15-minute periods with 96 segments.
If this forecast is made accurately for each segment, as is done by this AI-ML format, then the correct amount of power required for the next day can be sent every 15 minutes from each generating station, thereby The cost of purchasing power is low and wastage is reduced. Nagulapalli explains, “If the forecast is not known, our plan will go awry and the cost will go up.” APTRANSCO has brought several versions of this format since last year. The most advanced format is Dove 6, which has an accuracy of 97 percent. As a result, this state, which consumes Rs 30,000 crore annually, can save Rs 2 lakh to Rs 3 crore every day by purchasing and transmitting the right amount of electricity.
Nagulapalli says work is also underway to enable the state to use this format for other forecasts a day earlier, including wind power, solar power and its use for market price forecasts. Since 25 percent of the state’s electricity demand is met from renewable sources, the forecast format for wind energy and solar energy becomes necessary. Nagulapalli estimates that this format will also be used for price forecasting, which plays an important role for companies generating and distributing electricity, as they rely on this information to formulate their bidding strategies. . If an electricity producer has an accurate forecast of prices, it can devise a bidding strategy to maximize its profit.