Open AccessPeak load forecasting based on multi‐source data and day‐to‐day topological network
Author(s) -
Zeng Pan,
Jin Min
Publication year - 2018
Publication title -
iet generation, transmission and distribution
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.92
H-Index - 110
eISSN - 1751-8695
pISSN - 1751-8687
DOI - 10.1049/iet-gtd.2017.0201
Subject(s) - electrical load , computer science , artificial neural network , similarity (geometry) , support vector machine , data mining , feed forward , variation (astronomy) , topology (electrical circuits) , artificial intelligence , engineering , electrical engineering , physics , voltage , control engineering , astrophysics , image (mathematics)
Electric load variation results from a variety of factors. This study discusses the inner mechanism of load variation and presents a load forecasting method based on multi‐source data and day‐to‐day topological network. Multi‐source data including weather data, seasonal attributes, calendar attributes, holiday attributes and load growth rate are considered to be the dominant factors that cause the load variation. Initially, a day‐to‐day topological network is generated that reflects the similarity between any 2 days. Additionally, the random walk with restart (RWR) algorithm is applied to the topological network to construct the training set. The support vector regression model is then adopted. With case studies using the data of Guangzhou, the authors obtain a decrease of the mean absolute percentage errors of 19.5 and 25.7% over the SV machine method and a neural network ensemble model. Their research reveals three points: (i) weather elements, especially temperature, air pressure and water vapour pressure, have a dominant influence on load variation besides historical load; (ii) multi‐source data improves the accuracy, meanwhile the applying of RWR algorithm maximises its effect; and (iii) since some of the factors that cause a constant error remains unknown, they can use the feedforward correction to decrease the affects.