Short- and Medium-Term Electricity Consumption Forecasting Using Prophet and GRU DOI Open Access
Nam-Rye Son,

Yoonjeong Shin

Sustainability, Journal Year: 2023, Volume and Issue: 15(22), P. 15860 - 15860

Published: Nov. 11, 2023

Electricity consumption forecasting plays a crucial role in improving energy efficiency, ensuring stable power supply, reducing costs, optimizing facility management, and promoting environmental conservation. Accurate predictions help optimize system operations, reduce wastage, cut decrease carbon emissions. Consequently, the research on electricity algorithms is thriving. However, to overcome challenges like data imbalances, quality issues, seasonal variations, event handling, recent models employ various approaches, including probability statistics, machine learning, deep learning. This study proposes short- medium-term prediction algorithm by combining GRU model suitable for long-term Prophet seasonality handling. (1) The preprocessed propose first step handling prediction. (2) In second step, seven multivariate are experimented with using GRU. Specifically, consist of six meteorological residuals between predicted from proposed Step 1 observed data. These utilized predict at 15 min intervals. (3) short-term (2 days 7 days) (15 30 scenarios. approach outperforms both models, errors offering valuable insights into patterns.

Language: Английский

A hybrid prediction interval model for short-term electric load forecast using Holt-Winters and Gate Recurrent Unit DOI
Xin He, Wenlu Zhao, Zhijun Gao

et al.

Sustainable Energy Grids and Networks, Journal Year: 2024, Volume and Issue: 38, P. 101343 - 101343

Published: March 12, 2024

Language: Английский

Citations

11
Comparing ARIMA and various deep learning models for long-term water quality index forecasting in Dez River, Iran DOI
Amir Reza R. Niknam, Maryam Sabaghzadeh,

Ali Barzkar

et al.

Environmental Science and Pollution Research, Journal Year: 2024, Volume and Issue: unknown

Published: Feb. 14, 2024

Language: Английский

Citations

9
Few-Shot Load Forecasting Under Data Scarcity in Smart Grids: A Meta-Learning Approach DOI Creative Commons
Georgios Tsoumplekas, Christos L. Athanasiadis, Dimitrios I. Doukas

et al.

Energies, Journal Year: 2025, Volume and Issue: 18(3), P. 742 - 742

Published: Feb. 6, 2025

Despite the rapid expansion of smart grids and large volumes data at individual consumer level, there are still various cases where adequate collection to train accurate load forecasting models is challenging or even impossible. This paper proposes adapting an established Model-Agnostic Meta-Learning algorithm for short-term in context few-shot learning. Specifically, proposed method can rapidly adapt generalize within any unknown time series arbitrary length using only minimal training samples. In this context, meta-learning model learns optimal set initial parameters a base-level learner recurrent neural network. The evaluated dataset historical consumption from real-world consumers. examined series’ short length, it produces forecasts outperforming transfer learning task-specific machine methods by 12.5%. To enhance robustness fairness during evaluation, novel metric, mean average log percentage error, that alleviates bias introduced commonly used MAPE metric. Finally, studies evaluate model’s under different hyperparameters lengths also conducted, demonstrating approach consistently outperforms all other models.

Language: Английский

Citations

1
Dual-layered deep learning and optimization algorithm for electric vehicles charging infrastructure planning DOI
Bishoy E. Sedhom, Abdelfattah A. Eladl, Pierluigi Siano

et al.

International Journal of Electrical Power & Energy Systems, Journal Year: 2025, Volume and Issue: 166, P. 110545 - 110545

Published: Feb. 20, 2025

Language: Английский

Citations

1
Encoder–Decoder Based LSTM and GRU Architectures for Stocks and Cryptocurrency Prediction DOI Open Access
Joy Dip Das, Ruppa K. Thulasiram, Christopher J. Henry

et al.

Journal of risk and financial management, Journal Year: 2024, Volume and Issue: 17(5), P. 200 - 200

Published: May 12, 2024

This work addresses the intricate task of predicting prices diverse financial assets, including stocks, indices, and cryptocurrencies, each exhibiting distinct characteristics behaviors under varied market conditions. To tackle challenge effectively, novel encoder–decoder architectures, AE-LSTM AE-GRU, integrating principle with LSTM GRU, are designed. The experimentation involves multiple activation functions hyperparameter tuning. With extensive enhancements applied to AE-LSTM, proposed AE-GRU architecture still demonstrates significant superiority in forecasting annual volatile assets from sectors mentioned above. Thus, emerges as a superior choice for price prediction across fluctuating scenarios by extracting important non-linear features data retaining long-term context past observations.

Language: Английский

Citations

6
A novel damped conformable fractional grey Bernoulli model and its applications in energy prediction with uncertainties DOI
Nailu Li,

Eto Sultanan Razia,

haonan ba

et al.

Applied Mathematical Modelling, Journal Year: 2024, Volume and Issue: 130, P. 94 - 118

Published: March 6, 2024

Language: Английский

Citations

5
Reconstructing historical forest fire risk in the non-satellite era using the improved forest fire danger index and long short-term memory deep learning-a case study in Sichuan Province, southwestern China DOI Creative Commons

Yuwen Peng,

Huiyi Su, Min Sun

et al.

Forest Ecosystems, Journal Year: 2024, Volume and Issue: 11, P. 100170 - 100170

Published: Jan. 1, 2024

Historical forest fire risk databases are vital for evaluating the effectiveness of past management approaches, enhancing warnings and emergency response capabilities, accurately budgeting potential carbon emissions resulting from fires. However, due to unavailability spatial information technology, such extremely difficult build reliably completely in non-satellite era. This study presented an improved reconstruction framework that integrates a deep learning-based time series prediction model interpolation address challenge Sichuan Province, southwestern China. First, danger index (FFDI) was by supplementing slope aspect information. We compared performances three models, namely, autoregressive integrated moving average (ARIMA), Prophet long short-term memory (LSTM) predicting modified (MFFDI). The best-performing used retrace MFFDI individual stations 1941 1970. Following this, Anusplin method map distributions at five-year intervals, which were then subjected weighted overlay with distance-to-river layer generate maps reconstructing database. results revealed LSTM as most accurate fitting historical MFFDI, determination coefficient (R2) 0.709, mean square error (MSE) 0.047, validation R2 MSE 0.508 0.11, respectively. Independent predicted indicated 5 out 7 events located fire-prone areas, is higher than determined original FFDI (2 7). proves indicates high level reliability proposed this study.

Language: Английский

Citations

4
The Development Trend of New Energy Electric Vehicles Based on Correlation Model DOI
Dan Chen, N. Xu,

Siqi Bo

et al.

Sustainable civil infrastructures, Journal Year: 2025, Volume and Issue: unknown, P. 317 - 331

Published: Jan. 1, 2025

Language: Английский

Citations

0
COMPARING FORECASTS OF AGRICULTURAL SECTOR EXPORT VALUES USING SARIMA AND LONG SHORT-TERM MEMORY MODELS DOI Creative Commons

Aleytha Ilahnugrah Kurnadipare,

Sri Amaliya,

Khairil Anwar Notodiputro

et al.

BAREKENG JURNAL ILMU MATEMATIKA DAN TERAPAN, Journal Year: 2025, Volume and Issue: 19(1), P. 385 - 396

Published: Jan. 13, 2025

Indonesia's agricultural sector plays a crucial role in the national economy, with significant export potential and supporting livelihoods of majority population. As part government's vision to make Indonesia world's food barn by 2045, increasing volume value product exports is primary focus, making forecasting essential for strategic decision-making. Sequential data analysis an important approach analyzing collected over specific period. This study aims compare two popular methods sector, namely Seasonal AutoRegressive Integrated Moving Average (SARIMA) model Long Short-Term Memory (LSTM) model. Monthly from West Java Province January 2013 February 2024 were used as dataset. The best SARIMA generated was (1,1,1)(0,1,1)12, while optimal parameters LSTM neuron: 50, dropout rate: 0.3, number layers: 2, activation function: relu, epochs: 500, batch size: 64, optimizer: Adam, learning 0.01. Evaluation performed using Root Mean Squared Error (RMSE) method measure accuracy both models sector. results showed that outperformed model, lower RMSE (SARIMA: 4182.133 LSTM: 1939.02). research provides valuable insights decision-makers planning strategies future. With this comparison, it expected provide better guidance selecting suitable characteristics data.

Language: Английский

Citations

0
Time series forecasting via integrating a filtering method: an application to electricity consumption DOI
Felipe Leite Coelho da Silva, Josiane da Silva Cordeiro, Kleyton da Costa

et al.

Computational Statistics, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 13, 2025

Language: Английский

Citations

0