z-logo
open-access-imgOpen Access
EMLP: short-term gas load forecasting based on ensemble multilayer perceptron with adaptive weight correction
Author(s) -
Fengyong Li,
Meng Sun
Publication year - 2021
Publication title -
mathematical biosciences and engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.451
H-Index - 45
eISSN - 1551-0018
pISSN - 1547-1063
DOI - 10.3934/mbe.2021082
Subject(s) - computer science , perceptron , term (time) , multilayer perceptron , set (abstract data type) , ensemble forecasting , artificial neural network , data mining , data set , stability (learning theory) , artificial intelligence , machine learning , programming language , physics , quantum mechanics
This paper tackles a recent challenge in smart city that how to improve the accuracy of short-term natural gas load forecasting. Existing works on natural gas forecasting mostly reply on a combined forecasting model by simply integrating several single-forecasting models. However, due to the existence of redundant single-forecasting models, these works may not attain a higher prediction accuracy. To address the problem, we design a new natural gas load forecasting scheme based on ensemble multilayer perceptron (EMLP) with adaptive weight correction. Our method firstly normalizes multi-source data as original data set, which is further segmented by a window model. Then, the abnormal data is removed and subsequently interpolated to form a complete normalized data set. Furthermore, we integrate a series of multilayer perceptron (MLP) network to construct an ensemble forecasting model. An adaptive weight correction function is introduced to dynamically modify the weight of the previous predicted result. Since the correction function can match well the volatility characteristics of load data, the prediction accuracy is significantly improved. Extensive experiments demonstrate that our method outperforms existing state-of-the-art load forecasting schemes in terms of the prediction accuracy and stability.

The content you want is available to Zendy users.

Already have an account? Click here to sign in.
Having issues? You can contact us here