Forecasting Solar Photovoltaic Power Production: A Comprehensive Review and Innovative Data-Driven Modeling Framework

Energies, Год журнала: 2024, Номер 17(16), С. 4145 - 4145

Опубликована: Авг. 20, 2024

The intermittent and stochastic nature of Renewable Energy Sources (RESs) necessitates accurate power production prediction for effective scheduling grid management. This paper presents a comprehensive review conducted with reference to pioneering, comprehensive, data-driven framework proposed solar Photovoltaic (PV) generation prediction. systematic integrating comprises three main phases carried out by seven modules addressing numerous practical difficulties the task: phase I handles aspects related data acquisition (module 1) manipulation 2) in preparation development scheme; II tackles associated model 3) assessment its accuracy 4), including quantification uncertainty 5); III evolves towards enhancing incorporating context change detection 6) incremental learning when new become available 7). adeptly addresses all facets PV prediction, bridging existing gaps offering solution inherent challenges. By seamlessly these elements, our approach stands as robust versatile tool precision real-world applications.

Язык: Английский

---

Short-term photovoltaic power prediction with CPO-BILSTM based on quadratic decomposition

Electric Power Systems Research, Год журнала: 2025, Номер 243, С. 111511 - 111511

Опубликована: Фев. 12, 2025

Язык: Английский

---

Optimization of Bi-LSTM Photovoltaic Power Prediction Based on Improved Snow Ablation Optimization Algorithm

Energies, Год журнала: 2024, Номер 17(17), С. 4434 - 4434

Опубликована: Сен. 4, 2024

To enhance the stability of photovoltaic power grid integration and improve prediction accuracy, a method based on an improved snow ablation optimization algorithm (Good Point Vibration Snow Ablation Optimizer, GVSAO) Bi-directional Long Short-Term Memory (Bi-LSTM) network is proposed. Weather data divided into three typical categories using K-means clustering, normalization performed minmax method. The key structural parameters Bi-LSTM, such as feature dimension at each time step number hidden units in LSTM layer, are optimized Good strategy. A model constructed GVSAO-Bi-LSTM, test functions selected to analyze evaluate model. research results show that average absolute percentage error GVSAO-Bi-LSTM under sunny, cloudy, rainy weather conditions 4.75%, 5.41%, 14.37%, respectively. Compared with other methods, this more accurate, verifying its effectiveness.

Язык: Английский

---

Study of Wind Power Prediction in ELM Based on Improved SSA

IEEJ Transactions on Electrical and Electronic Engineering, Год журнала: 2025, Номер unknown

Опубликована: Янв. 13, 2025

This paper proposes a short‐term wind power prediction model based on the improved Sparrow Search Algorithm (SSA) and Extreme Learning Machine(ELM) for anomalous information from farms. The objective is to enhance accuracy of prediction. employs extraction features utilizing raw history data farms, in conjunction with application Variable Importance Projection indices Partial Least Squares (PLS‐VIP). As ELM network susceptible influence randomly generated input weights thresholds at outset training, solution proposed whereby are optimized using SSA. optimal identified by SSA then applied model, thus forming SSA‐ELM model. To address limitations traditional SSA, namely its susceptibility local solutions poor global search ability, an algorithm proposed. introduces chaotic sequences exchange learning strategy original rationale behind incorporating quality initial solution, ensuring more uniform distribution sparrow positions and, consequently, diverse population. This, turn, enables achieve effective capability through utilization strategy. Subsequently, all fed into purposes. simulation results demonstrate that exhibits enhanced practical applicability © 2025 Institute Electrical Engineers Japan. Published Wiley Periodicals LLC.

Язык: Английский

---

Research on short-term photovoltaic power point-interval prediction method based on multi-scale similar day and EVO-TABiGRU

Measurement Science and Technology, Год журнала: 2025, Номер 36(4), С. 046011 - 046011

Опубликована: Апрель 4, 2025

Abstract Photovoltaic (PV) power generation, known for its environmental benefits and renewability, plays a critical role in advancing sustainable energy. However, the inherent randomness volatility of PV generation challenge stable operation systems with high penetration. Accurate prediction is essential ensuring safe grid integration reliable system operation. This study introduces an advanced short-term framework, combining multi-scale similar days (MSSD) selection trend-aware bidirectional gated recurrent unit (TABiGRU). First, MSSD employed to select historical data meteorological conditions predicted day as training samples, reducing impact on model. Then, enhance model’s ability capture trends dynamics, TABiGRU model proposed, which change rate features dynamic weight adjustment improve adaptability fluctuations. In addition, energy valley optimization algorithm used tune hyperparameters TABiGRU, preventing performance degradation due improper parameter settings. Furthermore, mitigate cumulative error issue point under uncertain conditions, adaptive bandwidth kernel density estimation generate high-quality intervals, providing more robust decision support scheduling. Finally, experimental results demonstrate that proposed method achieves accuracy stability various particularly showing significant advantages complex fluctuation scenarios, strong grid.

Язык: Английский

---

A Deep Learning Method for Photovoltaic Power Generation Forecasting Based on a Time-Series Dense Encoder

Energies, Год журнала: 2025, Номер 18(10), С. 2434 - 2434

Опубликована: Май 9, 2025

Deep learning has become a widely used approach in photovoltaic (PV) power generation forecasting due to its strong self-learning and parameter optimization capabilities. In this study, we apply deep algorithm, known as the time-series dense encoder (TiDE), which is an MLP-based encoder–decoder model, forecast PV generation. TiDE compresses historical time series covariates into latent representations via residual connections reconstructs future values through temporal decoder, capturing both long- short-term dependencies. We trained model using data from 2020 2022 Australia’s Desert Knowledge Australia Solar Centre (DKASC), with 2023 for testing. Forecast accuracy was evaluated R2 coefficient of determination, mean absolute error (MAE), root square (RMSE). 5 min ahead test, demonstrated high 0.952, MAE 0.150, RMSE 0.349, though performance declines longer horizons, such 1 h forecast, compared other algorithms. For one-day-ahead forecasts, it achieved 0.712, 0.507, 0.856, effectively medium-term weather trends but showing limited responsiveness sudden changes. Further analysis indicated improved cloudy rainy weather, seasonal reveals higher spring autumn, reduced summer winter extreme conditions. Additionally, explore model’s sensitivity input environmental variables, algorithmic versatility, implications errors on grid integration. These findings highlight TiDE’s superior robust adaptability across conditions, while also revealing limitations under abrupt

Язык: Английский

---

Time Series Forecasting for Energy Management: Neural Circuit Policies (NCPs) vs. Long Short-Term Memory (LSTM) Networks

Electronics, Год журнала: 2024, Номер 13(18), С. 3641 - 3641

Опубликована: Сен. 12, 2024

This paper investigates the effectiveness of Neural Circuit Policies (NCPs) compared to Long Short-Term Memory (LSTM) networks in forecasting time series data for energy production and consumption context predictive maintenance. Utilizing a dataset generated from Tuscan company specialized food refrigeration, we simulate scenario where employs 60 kWh storage system calculate battery charge discharge policies assess potential cost reductions increased self-consumption produced energy. Our findings demonstrate that NCPs outperform LSTM by leveraging underlying physical models, offering superior maintenance solutions production.

Язык: Английский

---

Improvement in the Forecasting of Low Visibility over Guizhou, China, Based on a Multi-Variable Deep Learning Model

Atmosphere, Год журнала: 2024, Номер 15(7), С. 752 - 752

Опубликована: Июнь 24, 2024

High-quality visibility forecasting benefits traffic transportation safety, public services, and tourism. For a more accurate forecast of the in Guizhou region China, we constructed several models via progressive refinements different compositions input observational variables adoption Unet architecture to perform hourly forecasts with lead times ranging from 0 72 h over Guizhou, China. Three Unet-based were according inputs meteorological variables. The model training multiple high-spatiotemporal-resolution numerical weather prediction (China Meteorological Administration, Guangdong, CMA-GD) produced higher threat score (TS), which led substantial improvements for thresholds compared CMA-GD. However, had larger bias (BS) than CMA-GD model. By introducing U2net architecture, there was further improvement TS by approximately factor two model, along significant reduction BS, enhanced stability forecast. In particular, U2net-based performed best terms below threshold 200 m, eightfold increase Furthermore, some TS, RMSE (root-mean-square error) LSTM_Attention spatial distribution showed that better at grid scale 3 km individual stations. summary, based on algorithm, variables, data best. key improving deep learning model’s capability, these could improve value support socioeconomic needs sectors reliant forecasting.

Язык: Английский

---

Cold Chain Logistics Center Layout Optimization Based on Improved Dung Beetle Algorithm

Symmetry, Год журнала: 2024, Номер 16(7), С. 805 - 805

Опубликована: Июнь 27, 2024

To reduce the impact of cold chain logistics center layout on economic benefits, operating efficiency and carbon emissions, a optimization method is proposed based improved dung beetle algorithm. Firstly, analysis relationship between non-logistics, multi-objective model established to minimize total cost, maximize adjacency correlation emissions; secondly, standard Dung Beetle Optimization (DBO) algorithm, in order further improve global exploration ability Chebyshev chaotic mapping an adaptive Gaussian–Cauchy hybrid mutation disturbance strategy are introduced DBO (IDBO) algorithm; finally, taking actual as example, algorithm applied optimize its layout, respectively. The results show that cost after IDBO reduced by 25.54% compared with original 29.93%, emission 6.75%, verifying effectiveness providing reference for design centers.

Язык: Английский

---

Environmental Impact Minimization Model for Storage Yard of In-Situ Produced PC Components: Comparison of Dung Beetle Algorithm and Improved Dung Beetle Algorithm

Buildings, Год журнала: 2024, Номер 14(12), С. 3753 - 3753

Опубликована: Ноя. 25, 2024

If PC components are produced on site under the same conditions, quality can be secured at least equal to that of factory production. In-situ production reduce environmental loads by 14.58% or more than production, and if number in-situ is increased, cost reduced up 39.4% compared Most existing studies focus optimizing layout logistics centers, relatively little attention paid parts for component yard planning effectively carbon dioxide emissions improve construction efficiency. Therefore, purpose this study develop an impact minimization model components. As a result applying developed model, optimization improved dung beetle algorithm was verified efficient improving neighboring correlation 22.79% reducing 18.33% algorithm. The proposed support construction, reconstruction, functional upgrade contributing low in industry.

Язык: Английский

---