VIJAYAWADA: Leveraging technology, the state power utilities have developed and adopted demand forecast and energy forecast models which are said to be helping them save an estimated Rs 3 crore per day.
The state load dispatch centre (APSLDC) of APTRANSCO is now in the process of developing frequency forecast and dispatch forecast models so as to make the grid management more efficient and thereby optimising the costs.
The officials said forecasting helps plan the quantum of energy that should be dispatched from each generating units, plan market purchases and find alternatives in case of shortfall in demand/availability.
They added that though manual forecast helps in projection of necessary information, it is not accurate and leads to overdrawl and underdrawl of power from national grid, both of which will involve high power purchase costs in the market, heavy penalties and other procedural difficulties.
“The APSLDC has developed two models for demand forecast and energy forecast so far. The mean absolute percentage error (MAPE) of energy and demand forecast models in manual method is 2.84 per cent and 3.89 per cent. However, the models developed using machine learning and artificial intelligence helped bring down the errors to 1.14 per cent and 2.16 per cent. In terms of monetary benefit, an estimated Rs 3 crore per day is being saved,” a senior official explained.
For the record, the SLDC, which is responsible for integrated operations of entire power system including scheduling and dispatch of power and grid management, requires forecasting to be prepared for the next day’s energy (either conventional or renewable) consumption. The major objective of forecasting, which is done for 15 minute interval i.e. 96 intervals a day, is minimising power purchase costs.
Similarly, wind and solar energy forecast has a MAPE of 44 per cent in manual method and the officials are in the process of developing a model for the same using MI and AL. “A frequency forecast model, a dispatch model, a market model are also being planned. Forecasting with minimum error to meet the grid demand is the objective,” the official added.
APSLDC also intends to transfer the model to start-up companies willing to commercialise the product.