A Study on the development of long-term hybrid electrical load forecasting model based on MLP and statistics using massive actual data considering field applications

Various studies have conducted to efficiently operate the distribution system and establish distribution system plans. In recent years, research has been highlighted on how to interpret and operate distribution systems from a predictive perspective, especially by forecasting loads. However, it is import to properly forecast not only short-term loads, but also mid/long-term loads considering actual application, as most of distribution system planning related works require those. In this work, therefore, we propose a reliable hybrid load forecasting model for distribution lines that combines Multi-Layered Perceptron and statistical techniques to solve the average convergence problem that could occur when only machine learning method is applied to forecast mid/long-term load. Also, the proposed model is developed and verified by using actual massive load data from more than 10,000 distribution lines provided by KEPCO. The verification results confirm that the proposed method has an average accuracy of 89.56% based on Normalised Root Mean Squared Error for the total 10,832 distribution lines in South Korea.