Short-term photovoltaic power forecasting with feature extraction and attention mechanisms

Renewable Energy, Journal Year: 2024, Volume and Issue: 226, P. 120437 - 120437

Published: April 1, 2024

Language: Английский

---

Short-term Interval Prediction Strategy of Photovoltaic Power Based on Meteorological Reconstruction with Spatiotemporal Correlation and Multi-factor Interval Constraints

Renewable Energy, Journal Year: 2024, Volume and Issue: 237, P. 121834 - 121834

Published: Nov. 6, 2024

Language: Английский

---

Ultra-Short-Term Distributed Photovoltaic Power Probabilistic Forecasting Method Based on Federated Learning and Joint Probability Distribution Modeling

Energies, Journal Year: 2025, Volume and Issue: 18(1), P. 197 - 197

Published: Jan. 5, 2025

The accurate probabilistic forecasting of ultra-short-term power generation from distributed photovoltaic (DPV) systems is great significance for optimizing electricity markets and managing energy on the user side. Existing methods regarding cluster information sharing tend to easily trigger issues data privacy leakage during sharing, or they suffer insufficient while protecting privacy, leading suboptimal performance. To address these issues, this paper proposes a privacy-preserving deep federated learning method DPV systems. Firstly, collaborative feature framework established. For central server, among clients realized through interaction global models features avoiding direct raw ensure security client privacy. local clients, Transformer autoencoder used as model extract temporal features, which are combined with form spatiotemporal correlation thereby deeply exploring correlations between different stations improving accuracy forecasting. Subsequently, joint probability distribution values errors constructed, patterns finely studied based dependencies enhance Finally, effectiveness proposed was validated real datasets.

Language: Английский

---

Multi-Objective Optimal Scheduling of Microgrids Based on Improved Particle Swarm Algorithm

Energies, Journal Year: 2024, Volume and Issue: 17(7), P. 1760 - 1760

Published: April 7, 2024

Microgrid optimization scheduling, as a crucial part of smart grid optimization, plays significant role in reducing energy consumption and environmental pollution. The development goals microgrids not only aim to meet the basic demands electricity supply but also enhance economic benefits protection. In this regard, multi-objective scheduling model for grid-connected mode is proposed, which comprehensively considers operational costs protection microgrid systems. This incorporates improvements traditional particle swarm (PSO) algorithm by considering inertia factors adaptive mutation, it utilizes improved solve model. Simulation results demonstrate that can effectively reduce users pollution, promoting optimized operation verifying superior performance PSO algorithm. After algorithmic improvements, optimal total cost achieved was CNY 836.23, representing decrease from pre-improvement value 850.

Language: Английский

---

Ultra-short-term distributed PV power forecasting for virtual power plant considering data-scarce scenarios

Applied Energy, Journal Year: 2024, Volume and Issue: 373, P. 123890 - 123890

Published: July 17, 2024

Language: Английский

---

Probabilistic Forecasting of Provincial Regional Wind Power Considering Spatio-Temporal Features

Energies, Journal Year: 2025, Volume and Issue: 18(3), P. 652 - 652

Published: Jan. 30, 2025

Accurate prediction of regional wind power generation intervals is an effective support tool for the economic and stable operation provincial grid. However, it involves a large amount high-dimensional meteorological historical information related to massive stations in province. In this paper, lightweight model developed directly obtain probabilistic predictions form intervals. Firstly, input features are formed through fused image method geographic as well aggregation strategy, which avoids extensive tedious data processing process prior modeling traditional approach. Then, order effectively consider spatial distribution characteristics temporal power, parallel network architecture convolutional neural (CNN) long short-term memory (LSTM) designed. Meanwhile, efficient channel attention (ECA) mechanism improved quantile regression-based loss function introduced training generate The case study shows that proposed paper improves interval performance by at least 12.3% reduces deterministic root mean square error (RMSE) 19.4% relative benchmark model.

Language: Английский

---

Enhancing PV power forecasting accuracy through nonlinear weather correction based on multi-task learning

Applied Energy, Journal Year: 2025, Volume and Issue: 386, P. 125525 - 125525

Published: Feb. 20, 2025

Language: Английский

---

Ultra-Short-Term Forecasting of Large Distributed Solar PV Fleets Using Sparse Smart Inverter Data

IEEE Transactions on Sustainable Energy, Journal Year: 2024, Volume and Issue: 15(3), P. 1968 - 1980

Published: April 17, 2024

Ultra-short-term power forecasting for distributed solar photovoltaic (PV) generation is a largely unaddressed, highly challenging problem due to the prohibitive real-time data collection and processing requirements sheer number of PV units. In this paper, we propose an innovative idea output large fleet units using limited sparsely selected set units, referred as pilot We develop two-stage method address problem. planning stage, use K-medoids clustering algorithm select installation remote monitoring infrastructure. operation devise deep learning framework integrating Long Short-Term Memory, Graph Convolutional Network, Multilayer Perceptron capture spatio-temporal patterns between other forecast outputs all in from few only. Case study results show that our proposed outperforms baseline methods individual well whole fleet, time resolution not dependent on weather data.

Language: Английский

---

A SVM based demand response capacity prediction model considering internal factors under composite program

Energy, Journal Year: 2024, Volume and Issue: 300, P. 131460 - 131460

Published: April 28, 2024

Language: Английский

---

Learning Coupled Meteorological Characteristics Aids Short-Term Photovoltaic Interval Prediction Methods

Energies, Journal Year: 2025, Volume and Issue: 18(2), P. 308 - 308

Published: Jan. 12, 2025

In response to the challenges posed by renewable energy integration, this study introduces a hybrid Attention-TCN-LSTM model for short-term photovoltaic (PV) power forecasting. The LSTM captures sequence characteristics of PV output, which are then combined with meteorological features extracted Attention-TCN module. leverages strengths TCN, LSTM, and self-attention mechanism enhance prediction accuracy construct reliable intervals. Aiming optimize both performance efficiency, PSO algorithm is used hyperparameter optimization. Ablation studies comparisons other models confirm effectiveness, robustness proposed model. This approach contributes improved offering more stable supply. Future work will focus on incorporating intelligent systems autonomous risk management real-time control dynamic output fluctuations.

Language: Английский

---