Open AccessA novel short‐term load forecasting approach based on kernel extreme learning machine: A provincial case in China
Author(s) -
Zhao Xinyue,
Wang Jianxiao,
Zhang Tiance,
Cui Da,
Li Gengyin,
Zhou Ming
Publication year - 2022
Publication title -
iet renewable power generation
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.005
H-Index - 76
eISSN - 1752-1424
pISSN - 1752-1416
DOI - 10.1049/rpg2.12373
Subject(s) - extreme learning machine , overfitting , computer science , artificial intelligence , machine learning , term (time) , kernel (algebra) , dimensionality reduction , data mining , artificial neural network , mathematics , physics , quantum mechanics , combinatorics
With the rapid development of re‐electrification, traditional load forecasting faces a significant increase of influencing factors. Existing literature focuses on examining the influencing factors related to load profiles in order to improve the prediction accuracy. However, a large number of redundant features may lead to the overfitting of the forecasting engine. To enhance the performance of extreme learning machine (ELM) under massive data scale, this paper presents a kernel extreme learning machine (KELM) based method which can be used for short‐term load prediction. First, a feature dimensionality reduction is performed using a kernelized principal component analysis, which aims to eliminate redundant input vectors. Then, the hyperparameters of KELM are optimized to improve the prediction accuracy and generalization. Case studies based on a province‐level power system in China demonstrate that the presented method can significantly improve the accuracy of load forecasting by 3.14% in contrast to traditional ELM.