کنفرانس مهندسی برق ایران

صفحه اصلی / سی و سومین کنفرانس بین المللی مهندسی برق

Application of Statistical Techniques and Machine Learning in Forecasting Distribution Network Load: A Real Case Study on the Iranian Power System

نویسندگان :

Hossein Jafari¹ Mohammad Sadegh Sepasian² Fatemeh Teimori³

1- دانشگاه شهید بهشتی 2- دانشگاه شهید بهشتی 3- دانشگاه شهید بهشتی

کلمات کلیدی :

Electricity demand forecasting،Short-term load forecasting،Statistical time series analysis،Machine learning،Power system management

چکیده :

Abstract— Accurate electricity demand forecasting is essential for effective and reliable management of power system resources, especially in minimizing forecasting errors, and managing random demands to increase economic efficiency. The study aims to develop an efficient and reliable short-term load forecasting model to reduce significant residential losses in Iran. Especially those caused by summer power outages related to increased demand. The research utilizes statistical time series analysis along with machine learning methods to reduce forecasting errors. It focuses on key variables, such as national consumption, while excluding the effects of temperature and holidays, broadening the variable range to improve forecasting precision. The study emphasizes the influence of rapid demand fluctuations and environmental factors on the stability of forecasting models, advocating for a variety of forecasting methodologies. A comparison is performed between statistical analysis and machine learning methodologies to determine the most effective strategies for various forecasting periods. The findings reveal that machine learning algorithms surpass traditional statistical methods, emphasizing their efficacy in addressing complicated demand forecasting challenges.