Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
January 11, 2019EnergiesOpen Access

Daily Natural Gas Load Forecasting Based on a Hybrid Deep Learning Model

View Full Paper
Ask AI
Bookmark
Share

Authors

NWNan WeiCLChangjun LiJDJiehao Duan

Discussion

Loading...

Member takes

Overview

Modeling study demonstrates superior natural gas load prediction using a hybrid neural framework, indicating improved operational accuracy for energy systems.

Key Points

  • To develop a hybrid deep learning model combining a novel feature-selection algorithm with recurrent neural networks to improve the accuracy of daily natural gas load forecasting.
  • Designed a hybrid architecture pairing Principal Component Correlation Analysis (PCCA) for redundant feature elimination with a Long Short-Term Memory (LSTM) network.
  • Trained and evaluated the model incorporating 14 weather variables using historical gas load datasets from Xi’an, China, and Athens, Greece.
  • Benchmarked forecasting accuracy against standard LSTM, PCA-LSTM, backpropagation neural networks (BPNN), and support vector regression (SVR).
  • The PCCA–LSTM hybrid achieved the lowest mean absolute percentage error (MAPE) among evaluated models, recording 3.22% for Xi’an and 7.29% for Athens.
  • PCCA successfully eliminated redundant eigenspace components while preserving critical feature correlations, outperforming conventional PCA.

Cite This Study

Wei et al. (2019) studied this question.

synapsesocial.com/papers/6a1ff619f35583189204bddfhttps://doi.org/10.3390/en12020218
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched one closely related paper. Consider it for comparative context:

  1. 1Deep Learning for solar power forecasting — An approach using AutoEncoder and LSTM Neural Networks2016 · 621 citations