A new paradigm based on Wasserstein Generative Adversarial Network and time-series graph for integrated energy system forecasting

Energy Conversion and Management, Journal Year: 2025, Volume and Issue: 326, P. 119484 - 119484

Published: Jan. 13, 2025

Language: Английский

---

Short-term electric load forecasting based on series decomposition and Meta-Informer algorithm

Electric Power Systems Research, Journal Year: 2025, Volume and Issue: 243, P. 111478 - 111478

Published: Feb. 8, 2025

Language: Английский

---

Mixed-Frequency Grey Prediction Model with Fractional Lags for Electricity Demand and Estimation of Coal Power Phase-Out Scale

Energy, Journal Year: 2025, Volume and Issue: unknown, P. 135442 - 135442

Published: March 1, 2025

Language: Английский

---

Enhancing short-term net load forecasting with additive neural decomposition and Weibull Attention

Energy, Journal Year: 2025, Volume and Issue: unknown, P. 135486 - 135486

Published: March 1, 2025

Language: Английский

---

Dual-Modal Cross-Attention Integrated Model for Airport Terminal Cooling Load Prediction Using Variational Mode Decomposition

Journal of Building Engineering, Journal Year: 2025, Volume and Issue: unknown, P. 112344 - 112344

Published: March 1, 2025

Language: Английский

---

A TSFLinear model for wind power prediction with feature decomposition-clustering

Renewable Energy, Journal Year: 2025, Volume and Issue: unknown, P. 123142 - 123142

Published: April 1, 2025

Language: Английский

---

Integrated multi-energy load prediction system with multi-scale temporal channel features fusion

Measurement, Journal Year: 2025, Volume and Issue: unknown, P. 117559 - 117559

Published: April 1, 2025

Language: Английский

---

A combined prediction model with multi-module integration for short-term power load forecasting

Highlights in Science Engineering and Technology, Journal Year: 2025, Volume and Issue: 138, P. 181 - 194

Published: May 11, 2025

The existing power load forecasting algorithms are constrained by preprocessing limitations and insufficient prediction accuracy. Temporal Convolutional Network (TCN), Bi-directional LSTM (BiLSTM), Multi-Head Attention (MHA) for high-precision, real-time were used in this paper to propose a hybrid model, which combined Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) address these issues. ICEEMDAN algorithm is initially employed multi-layer decomposition enhance data smoothness, thereby enabling the subsequent model more effectively capture essential features. To overcome BiLSTM's inability long-term dependencies extended sequences, incorporates TCN module MH-Attention mechanism. improves local feature extraction through convolution, while mechanism enables focus on most critical features prediction, enhancing learning efficiency. validated using an actual plant Quanzhou City, China, dataset of obtained from real measurements. Experimental results demonstrate R2 0.99802, RMSE 0.00996, MAE 0.00694, showcasing exceptional Compared alternative models, 1.5%-3.4%, decreases 56.2%-80.6%, reduced 58.3%-84.1%. These validate model's superiority. proposed combinatorial framework integrates advantages decomposition, attention mechanisms, allowing in-depth exploration temporal patterns data.

Language: Английский

---

Short-Term Power Load Forecasting Using Adaptive Mode Decomposition and Improved Least Squares Support Vector Machine

Energies, Journal Year: 2025, Volume and Issue: 18(10), P. 2491 - 2491

Published: May 12, 2025

Accurate power load forecasting is crucial for ensuring grid stability, optimizing economic dispatch, and facilitating renewable energy integration in modern smart grids. However, real often disturbed by the inherent non-stationarity multi-factor coupling effects. To address this problem, a novel hybrid framework based on adaptive mode decomposition (AMD) improved least squares support vector machine (ILSSVM) proposed effective short-term forecasting. First, AMD utilized to obtain multiple components of signal. In AMD, minimum loss used adjust parameter adaptively, which can effectively decrease risk generating spurious modes losing critical components. Then, ILSSVM presented predict different components, separately. Different frequency features are extracted using combination kernel structure, achieve balance learning capacity generalization each unique component. Further, an optimized genetic algorithm deployed optimize model parameters integrating simulated annealing improve accuracy. The dataset collected from Guangxi region China test framework. Extensive experiments carried out results demonstrate that our achieves MAPE 1.78%, outperforms some other advanced models.

Language: Английский

---

Short-Term Load Forecasting Based on Similar Day Theory and BWO-VMD

Energies, Journal Year: 2025, Volume and Issue: 18(9), P. 2358 - 2358

Published: May 6, 2025

Short-term power load forecasting at the regional level is essential for maintaining grid stability and optimizing generation, consumption, maintenance scheduling. Considering temporal, periodic, nonlinear characteristics of load, a novel short-term method proposed in this paper. First, Random Forest importance ranking applied to select similar days weighted eigenspace coordinate system established measure similarity. The daily sequence then decomposed into high-, medium-, low-frequency components using Variational Mode Decomposition (VMD). high-frequency component predicted day averaging method, while neural networks are employed medium components, leveraging historical similar-day data, respectively. This multi-faceted approach enhances accuracy granularity pattern analysis. final forecast obtained by summing predictions these components. case study demonstrates that model outperforms LSTM, GRU, CNN, TCN Transformer, with an RMSE 660.54 MW MAPE 7.81%, also exhibiting fast computational speed low CPU usage.

Language: Английский

---