Comparative analysis of deep learning and classical time series methods to forecast natural gas demand during COVID-19 pandemic

Energy Sources Part B Economics Planning and Policy, Journal Year: 2023, Volume and Issue: 18(1)

Published: Aug. 1, 2023

ABSTRACTThe lockdown measures implemented to contain the COVID-19 pandemic have had a considerable effect on consumption of natural gas, which is closely linked economic growth countries. Accurately forecasting gas demand critical for making informed decisions in unprecedented and unexpected situations. This study aims compare artificial learning-based algorithms classical statistical time series models predicting during pandemic, using Turkey as case study. Common prediction methods, including Autoregressive Integrated Moving Average (ARIMA), Nonlinear Autoregression Neural Network (NARNN), Support Vector Regression (SVR), Long Short-Term Memory (LSTM), were utilized this purpose. The impact was analyzed by 2-year data since its onset. Root mean square error (RMSE), correlation coefficient (R), absolute (MAE) criteria used performance evaluation metrics select best model. results confirmed that deep-learning-based LSTM model provided better accuracy than time-series benchmark models, with lowest RMSE (9.442) highest R (0.997) values test dataset. Furthermore, validated analysis Diebold-Mariano Nemenyi tests.KEYWORDS: pandemicnatural consumptionenergytime-seriesdeep learningprediction Disclosure statementNo potential conflict interest reported author(s).Data availability statementThe support findings are available from corresponding author upon reasonable request.Additional informationFundingNo funding declare.

Language: Английский

---

LSTM and Bi-LSTM Models For Identifying Natural Disasters Reports From Social Media

Journal of Electronics Electromedical Engineering and Medical Informatics, Journal Year: 2023, Volume and Issue: 5(4)

Published: Sept. 5, 2023

Natural disaster events are occurrences that cause significant losses, primarily resulting in environmental and property damage the worst cases, even loss of life. In some cases natural disasters, social media has been utilized as fastest information bridge to inform many people, especially through platforms like Twitter. To provide accurate categorization information, field text mining can be leveraged. This study implements a combination word2vec LSTM methods Bi-LSTM determine which method is most for use case news related events. The utility lies its feature extraction method, transforming textual data into vector form processing classification stage. On other hand, used techniques categorize vectorized from process. experimental results show an accuracy 70.67% 72.17% Bi-LSTM. indicates improvement 1.5% achieved by combining methods. research identifying comparative performance each + Bi-LSTM, best-performing process classifying earthquake disasters. also offers insights various parameters present word2vec, LSTM, researchers determine.

Language: Английский

---

24-Step Short-term Power Load Forecasting Model Utilizing KOA-BiTCN-BiGRU-Attentions

Published: July 4, 2024

With the global objectives of achieving a ‘carbon peak’ and neutrality’, along with implementation carbon reduction policies, China’s industrial structure has undergone significant adjustments, resulting in constraints on high-energy consumption high-emission industries while promoting rapid growth green industries. Consequently, these changes have led to an increasingly complex power system presented new challenges for electricity demand forecasting. To address this issue, study proposes 24-step multivariate time series short-term load forecasting algorithm model based KNN data imputation BiTCN bidirectional temporal convolutional networks combined BiGRU gated recurrent units attention mechanism. The Kepler adaptive optimization (KOA) is employed hyperparameter effectively enhance prediction accuracy. Furthermore, using real from wind farm Xinjiang as example, paper predicts January 1st December 30th 2019. Experimental results demonstrate that our comprehensive exhibits lower errors superior performance compared traditional methods, thus holding great value practical applications.

Language: Английский

---

An Online Probability Density Load Forecasting Against Concept Drift Under Anomalous Events

IEEE Transactions on Industrial Informatics, Journal Year: 2023, Volume and Issue: 20(4), P. 5241 - 5252

Published: Nov. 28, 2023

Concept drift (i.e., the data pattern to be learned can change over time) is becoming more common in power loads due volatile external environment, posing huge challenges on load forecasting. The phenomenon could exacerbated by anomalous events, such as unprecedented coronavirus disease 2019 (COVID-19) pandemic. This article focuses handling problem of concept probabilistic prediction under abnormal events. To address this challenge, we propose a novel online model for probability density forecasting, which utilizes least absolute shrinkage and selection operator combined with quantile regression long short-term memory network base learner capture time dependencies Continuous ranked score integrated kernel estimation developed monitor performance model. Two modules, buffering tuning module, will timely adjust parameters according adapt new concepts data. Data from COVID-19 period verify effectiveness proposed dealing drift.

Language: Английский

---

Power load combination forecasting system based on longitudinal data selection

Engineering Applications of Artificial Intelligence, Journal Year: 2023, Volume and Issue: 130, P. 107629 - 107629

Published: Dec. 13, 2023

Language: Английский

---