Dense Skip Attention Based Deep Learning for Day-Ahead Electricity Price Forecasting

IEEE Transactions on Power Systems, Journal Year: 2022, Volume and Issue: 38(5), P. 4308 - 4327

Published: Nov. 4, 2022

The forecasting of the day-ahead electricity price (DAEP) has become more interest to decision makers in liberalized market, as it can help optimize bidding strategies and maximize profits with gradual market expansion. Deep learning (DL) is a promising method for its strong nonlinear approximation capabilities. However, challenging traditional DL models obtain high precision DAEP, due internal temporal feature-wise variabilities. To address issue, this paper proposes dense skip attention based model. In model, tackle variability, mechanism proposed efficiently assign learnable weights on features training. terms drop-connected structure advanced residual unshared convolutional neural network (ARUCNN) gate recurrent units (GRUs) further proposed. structure, ARUCNN developed by embedding activations deal short-term dependencies degradation while GRUs addressing long-term ones, they are integrated via drop connection reduce overfitting. Through validating real DAEP data markets Sweden, Denmark, Norway Finland, results verify our approach outperforms existing methods deterministic interval DAEP.

Language: Английский

---

A Seasonal Autoregressive Integrated Moving Average with Exogenous Factors (SARIMAX) Forecasting Model-Based Time Series Approach

Inventions, Journal Year: 2022, Volume and Issue: 7(4), P. 94 - 94

Published: Oct. 16, 2022

Time series modeling is an effective approach for studying and analyzing the future performance of power sector based on historical data. This study proposes a forecasting framework that applies seasonal autoregressive integrated moving average with exogenous factors (SARIMAX) model to forecast long-term electricity (electricity consumption, generation, peak load, installed capacity). In this study, was used aforementioned in Saudi Arabia 30 years from 2021 2050. The data were inputted into collected at quarterly intervals across 40-year period (1980−2020). SARIMAX technique time influencing factors, which helps reduce error values improve overall accuracy, even case close input output dataset lengths. experimental findings indicated has promising terms categorization consideration, as it significantly improved accuracy compared simpler average-based techniques. Furthermore, capable coping different-sized sequential datasets. Finally, aims help address issue lack planning analyses intermittency, provides reliable technique, prerequisite modern energy systems.

Language: Английский

---

Solar Radiation Forecasting by Pearson Correlation Using LSTM Neural Network and ANFIS Method: Application in the West-Central Jordan

Future Internet, Journal Year: 2022, Volume and Issue: 14(3), P. 79 - 79

Published: March 5, 2022

Solar energy is one of the most important renewable energies, with many advantages over other sources. Many parameters affect electricity generation from solar plants. This paper aims to study influence these on predicting radiation and electric produced in Salt-Jordan region (Middle East) using long short-term memory (LSTM) Adaptive Network-based Fuzzy Inference System (ANFIS) models. The data relating 24 meteorological for nearly past five years were downloaded MeteoBleu database. results show that varies according season. forecasting ANFIS provides better when parameter correlation high (i.e., Pearson Correlation Coefficient PCC between 0.95 1). In comparison, LSTM neural network shows low (PCC range 0.5–0.8). obtained RMSE 0.04 0.8 depending season used parameters; new influencing are also investigated.

Language: Английский

---

Short-Term Load Forecasting Based on the CEEMDAN-Sample Entropy-BPNN-Transformer

Energies, Journal Year: 2022, Volume and Issue: 15(10), P. 3659 - 3659

Published: May 17, 2022

Aiming at the problem that power load data are stochastic and it is difficult to obtain accurate forecasting results by a single algorithm, in this paper, combined method for short-term was proposed based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN)-sample entropy (SE), BP neural network (BPNN), Transformer model. Firstly, were decomposed into several subsequences obvious complexity differences using CEEMDAN-SE. Then, BPNN model used forecast low high complexity, respectively. Finally, of each subsequence superimposed final result. The simulation taken from our six models dataset certain area Spain. showed MAPE CEEMDAN-SE-BPNN-Transformer 1.1317%, while RMSE 304.40, which better than selected models.

Language: Английский

---

Prediction of strip section shape for hot-rolled based on mechanism fusion data model

Applied Soft Computing, Journal Year: 2023, Volume and Issue: 146, P. 110670 - 110670

Published: Aug. 3, 2023

Language: Английский

---

Short-Term Load Forecasting Using EMD with Feature Selection and TCN-Based Deep Learning Model

Energies, Journal Year: 2022, Volume and Issue: 15(19), P. 7170 - 7170

Published: Sept. 29, 2022

Short-term load forecasting (STLF) has a significant role in reliable operation and efficient scheduling of power systems. However, it is still major challenge to accurately predict due social natural factors, such as temperature, humidity, holidays weekends, etc. Therefore, very important for the feature selection extraction input data improve accuracy STLF. In this paper, novel hybrid model based on empirical mode decomposition (EMD), one-dimensional convolutional neural network (1D-CNN), temporal (TCN), self-attention mechanism (SAM), long short-term memory (LSTM) proposed fully decompose mine in-depth features forecasting. Firstly, original sequence was decomposed into number sub-series by EMD, Pearson correlation coefficient method (PCC) applied analyzing between with data. Secondly, achieve relationships series external factors during an hour scale correlations among these points, strategy 1D-CNN TCN comprehensively refine extraction. The SAM introduced further enhance key information. Finally, matrix fed According experimental results employing North American New England Control Area (ISO-NE-CA) dataset, more accurate than 1D-CNN, LSTM, TCN, 1D-CNN–LSTM, TCN–LSTM models. outperforms 21.88%, 51.62%, 36.44%, 42.75%, 16.67% 40.48%, respectively, terms mean absolute percentage error.

Language: Английский

---

Systematic review of content analysis algorithms based on deep neural networks

Multimedia Tools and Applications, Journal Year: 2022, Volume and Issue: 82(12), P. 17879 - 17903

Published: Oct. 24, 2022

Language: Английский

---

Enhanced neighborhood node graph neural networks for load forecasting in smart grid

International Journal of Machine Learning and Cybernetics, Journal Year: 2023, Volume and Issue: 15(1), P. 129 - 148

Published: March 21, 2023

Language: Английский

---

Smart city landscape design for achieving net-zero emissions: Digital twin modeling

Sustainable Energy Technologies and Assessments, Journal Year: 2024, Volume and Issue: 63, P. 103659 - 103659

Published: Feb. 14, 2024

Language: Английский

---

Efficiency evaluation of electricity distribution companies: Integrating data envelopment analysis and machine learning for a holistic analysis

Engineering Applications of Artificial Intelligence, Journal Year: 2024, Volume and Issue: 133, P. 108636 - 108636

Published: May 23, 2024

Evaluating the efficiency of electricity distribution companies (EDCs) accurately is one most important issues for regulators and policy makers. This research combines results data envelopment analysis (DEA) corrected ordinary least squares (COLS) with machine learning techniques to evaluate a set EDCs in period 2011–2020. We propose three-stage process. First, each year, scores are measured using DEA COLS methods. Then, this study applies support vector regression (SVR), powerful technique, estimate efficient frontier calculate EDCs. The efficiencies generated by DEA, COLS, SVR not same used construct fuzzy triangular numbers. Finally, considered as criteria technique order performance similarity ideal solution (TOPSIS), final ranks obtained TOPSIS (FTOPSIS) method. In addition, C-means clustering (FCM) algorithm, clustered discussed. show that there increasing decreasing trends selected 2011–2022. some act poor situation their should be improved.

Language: Английский

---