Yashar Ishan Agha - Department of Computer Engineering, University of Bojnord, Bojnord, Iran
Vahid Kiani - Department of Computer Engineering, University of Bojnord, Bojnord, Iran

One of the challenges faced by power distributioncompanies is the prediction of average current in orderto enable proper planning for sudden increases anddecreases that occur in the sinusoidal current signal.This planning can involve reducing production orstrengthening electrical transformers and otherequipment before reaching their limits, resulting in costsavings in terms of repairs, minimizing industrialequipment failures, and ultimately benefiting thecompany. Recently, in line with the smart gridinitiative, data loggers have been installed in city-levelpower substations to transmit information such asvoltage and current. With this data, which spans onemonth, we have developed a deep learning model usingLong Short-Term Memory (LSTM) networks to predictthe average current for the upcoming week. Through acomparative analysis, we have demonstrated thesuperior performance of our LSTM model incomparison to other neural networks, including MLPand GRU.

deep learning; average current forecasting;power distribution; long short-term memory