Robust-mv-M-LSTM-CI: Robust Energy Consumption Forecasting in Commercial Buildings during the COVID-19 Pandemic

Tan Ngoc Dinh (), Gokul Sidarth Thirunavukkarasu, Mehdi Seyedmahmoudian, Saad Mekhilef and Alex Stojcevski
Additional contact information
Tan Ngoc Dinh: School of Science, Computing and Engineering Technologies, Swinburne University of Technology, Hawthorn, VIC 3122, Australia
Gokul Sidarth Thirunavukkarasu: School of Science, Computing and Engineering Technologies, Swinburne University of Technology, Hawthorn, VIC 3122, Australia
Mehdi Seyedmahmoudian: School of Science, Computing and Engineering Technologies, Swinburne University of Technology, Hawthorn, VIC 3122, Australia
Saad Mekhilef: School of Science, Computing and Engineering Technologies, Swinburne University of Technology, Hawthorn, VIC 3122, Australia
Alex Stojcevski: School of Science, Computing and Engineering Technologies, Swinburne University of Technology, Hawthorn, VIC 3122, Australia

Sustainability, 2024, vol. 16, issue 15, 1-21

Abstract: The digitalization of the global landscape of electricity consumption, combined with the impact of the pandemic and the implementation of lockdown measures, has required the development of a precise forecast of energy consumption to optimize the management of energy resources, particularly in pandemic contexts. To address this, this research introduces a novel forecasting model, the robust multivariate multilayered long- and short-term memory model with knowledge injection ( Robust - mv - M - LSTM - CI ), to improve the accuracy of forecasting models under uncertain conditions. This innovative model extends the capabilities of mv - M - LSTM - CI by incorporating an additional branch to extract energy consumption from adversarial noise. The experiment results show that Robust - mv - M - LSTM - CI demonstrates substantial improvements over mv - M - LSTM - CI and other models with adversarial training: multivariate multilayered long short-term memory (adv-M-LSTM), long short-term memory (adv-LSTM), bidirectional long short-term memory (adv-Bi-LSTM), and linear regression (adv-LR). The maximum noise level from the adversarial examples is 0.005. On average, across three datasets, the proposed model improves about 24.01% in mean percentage absolute error (MPAE), 18.43% in normalized root mean square error (NRMSE), and 8.53% in R 2 over mv - M - LSTM - CI . In addition, the proposed model outperforms “adv-” models with MPAE improvements ranging from 35.74% to 89.80% across the datasets. In terms of NRMSE, improvements range from 36.76% to 80.00%. Furthermore, Robust - mv - M - LSTM - CI achieves remarkable improvements in the R 2 score, ranging from 17.35% to 119.63%. The results indicate that the proposed model enhances overall accuracy while effectively mitigating the potential reduction in accuracy often associated with adversarial training models. By incorporating adversarial noise and COVID-19 case data, the proposed model demonstrates improved accuracy and robustness in forecasting energy consumption under uncertain conditions. This enhanced predictive capability will enable energy managers and policymakers to better anticipate and respond to fluctuations in energy demand during pandemics, ensuring more resilient and efficient energy systems.

Keywords: commercial building; building energy consumption forecasting; LSTM; COVID-19 pandemic; uncertain data (search for similar items in EconPapers)
JEL-codes: O13 Q Q0 Q2 Q3 Q5 Q56 (search for similar items in EconPapers)
Date: 2024
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.mdpi.com/2071-1050/16/15/6699/pdf (application/pdf)
https://www.mdpi.com/2071-1050/16/15/6699/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:gam:jsusta:v:16:y:2024:i:15:p:6699-:d:1450247

Access Statistics for this article

Sustainability is currently edited by Ms. Alexandra Wu

More articles in Sustainability from MDPI
Bibliographic data for series maintained by MDPI Indexing Manager ().