Multi-Type Load Forecasting Model Based on Random Forest and Density Clustering with the Influence of Noise and Load Patterns DOI
Song Deng, Xia Dong, Tao Li

и другие.

Опубликована: Янв. 1, 2023

Load forecasting (LF) models lay a foundation for various smart-grid applications, whose accuracy is determined by the input load data. Prior LF studies mainly make restrictive assumptions on data, thus suffering from limited practicality in two folds: first, they model patterns over time, ignoring fact that real data are generated composition of multiple disparate electrical endeavours, leading to information loss perspective; and fail considering unexpected events which lead noises. To address these issues, we propose novel multi-type based random forest density clustering (MLF-RFDC), including three-fold ideas: 1) it each endeavour as an independent matrix; 2) detects corrects noisy entries matrix via low-rank structure; 3) harmonizes noise-free matrices all types ensemble perspective. Extensive experiments taken ten benchmark datasets three real-world datasets, results substantiate superiority our approach 11 state-of-the-art rival terms noise detection, restoration, accuracy.

Язык: Английский

Adaptive power flow analysis for power system operation based on graph deep learning DOI Creative Commons
Xiao Hu,

Jinduo Yang,

Yang Gao

и другие.

International Journal of Electrical Power & Energy Systems, Год журнала: 2024, Номер 161, С. 110166 - 110166

Опубликована: Авг. 14, 2024

Conventional model-driven methods are hard to handle large-scale power flow with multivariate uncertainty, variable topology, and massive real-time repetitive calculations. With the ability deal non-Euclidean graph-structured system data, graph deep learning shows great potential in modern calculation. However, general based calculation has limited adaptability because of its sole mapping node information black-box attributes. In this paper, an edge attention network (EGAT-PFC) model is proposed improved for analysis complex scenarios. First, dual-model structure constructed realize a complete covering all systems. Second, learnable coefficient mechanism fusing features ensure global can be completely considered. Third, mechanisms extended first-order neighborhood, dynamic normalization, regularization-based loss function designed improve training performance. Finally, visualized interpretability developed show valuable vulnerable nodes lines operation. The numerical simulation verifies that EGAT-PFC high accuracy, fast mapping, as well excellent topologies.

Язык: Английский

Процитировано

4
A ConvLSTM nearshore water level prediction model with integrated attention mechanism DOI Creative Commons
Jian Yang, Tianyu Zhang, Junping Zhang

и другие.

Frontiers in Marine Science, Год журнала: 2024, Номер 11

Опубликована: Окт. 4, 2024

Nearshore water-level prediction has a substantial impact on the daily lives of coastal residents, fishing operations, and disaster prevention mitigation. Compared to limitations high costs traditional empirical forecasts numerical models for nearshore prediction, data-driven artificial intelligence methods can more efficiently predict water levels. Attention mechanisms have recently shown great potential in natural language processing video prediction. Convolutional long short-term memory(ConvLSTM) combines advantages convolutional neural networks (CNN) Memory (LSTM), enabling effective data feature extraction. Therefore, this study proposes ConvLSTM level model that incorporates an attention mechanism. The extracts multiscale information from historical levels, mechanism enhances importance key features, thereby improving accuracy timeliness. structure was determined through experiments relevant previous studies using five years Zhuhai Tide Station surrounding wind speed rainfall training evaluation. results indicate outperforms four other baseline PCCs, RMSE, MAE, effectively predicting future levels at stations up 48 h advance. model, with showed average improvement approximately 10% test set, greater error reduction than long-term forecasts. During Typhoon Higos, reduced MAE best-performing by 3.2 2.4 cm 6- 24-hour forecasts, respectively, decreasing forecast errors 18% 11%, enhancing ability storm surges. This method provides new approach forecasting tides

Язык: Английский

Процитировано

4
Coupled Encoder-Decoder ConvLSTM Structure for the Parameter Identification Problems of Time-Dependent Partial Differential Equations DOI

琦 郭

Advances in Applied Mathematics, Год журнала: 2025, Номер 14(02), С. 341 - 355

Опубликована: Янв. 1, 2025

Язык: Английский

Процитировано

0
Dynamic Graph Attention Meets Multi-Scale Temporal Memory: A Hybrid Framework for Photovoltaic Power Forecasting Under High Renewable Penetration DOI Open Access
Xiaochao Dang,

Xiaoling Shu,

Fenfang Li

и другие.

Processes, Год журнала: 2025, Номер 13(3), С. 873 - 873

Опубликована: Март 16, 2025

In the context of accelerated global energy transition, power fluctuations caused by integration a high share renewable have emerged as critical challenge to security systems. The goal this research is improve accuracy and reliability short-term photovoltaic (PV) forecasting effectively modeling spatiotemporal coupling characteristics. To achieve this, we propose hybrid framework—GLSTM—combining graph attention (GAT) long memory (LSTM) networks. model utilizes dynamic adjacency matrix capture spatial correlations, along with multi-scale dilated convolution temporal dependencies, optimizes feature interactions through gated fusion unit. Experimental results demonstrate that GLSTM achieves RMSE values 2.3%, 3.5%, 3.9% for (1 h), medium-term (6 long-term (24 h) forecasting, respectively, mean absolute error (MAE) 3.8%, 6.2%, 7.0%, outperforming baseline models such LSTM, ST-GCN, Transformer reducing errors 10–25%. Ablation experiments validate effectiveness mechanism, 19% reduction in 1 h error. Robustness tests show remains stable under extreme weather conditions (RMSE 7.5%) data noise 8.2%). Explainability analysis reveals differentiated contributions features. proposed offers an efficient solution high-accuracy demonstrating its potential address key challenges integration.

Язык: Английский

Процитировано

0
Integrated multi-energy load prediction system with multi-scale temporal channel features fusion DOI
Dezhi Liu, Jiaming Zhu,

Mengyang Wen

и другие.

Measurement, Год журнала: 2025, Номер unknown, С. 117559 - 117559

Опубликована: Апрель 1, 2025

Язык: Английский

Процитировано

0
Short-Term Electricity-Load Forecasting by deep learning: A comprehensive survey DOI
Qi Dong, Rubing Huang, Chenhui Cui

и другие.

Engineering Applications of Artificial Intelligence, Год журнала: 2025, Номер 154, С. 110980 - 110980

Опубликована: Май 6, 2025

Язык: Английский

Процитировано

0
Sensitivity analysis and comparative assessment of novel hybridized boosting method for forecasting the power consumption DOI
Jing Zhou,

Qingdong Wang,

Hamed Khajavi

и другие.

Expert Systems with Applications, Год журнала: 2024, Номер 249, С. 123631 - 123631

Опубликована: Март 11, 2024

Язык: Английский

Процитировано

3
Comparative study on the performance of ConvLSTM and ConvGRU in classification problems—taking early warning of short-duration heavy rainfall as an example DOI Creative Commons
Meng Zhou, Jingya Wu, Mingxuan Chen

и другие.

Atmospheric and Oceanic Science Letters, Год журнала: 2024, Номер 17(4), С. 100494 - 100494

Опубликована: Март 30, 2024

Convolutional long short-term memory (ConvLSTM) and convolutional gated recurrent unit (ConvGRU) are two widely adopted deep learning models that combine mechanisms with operations for spatiotemporal sequences forecasting. To clarify the convergence speed classification ability of above models, using same model architecture to predict problem is necessary. This research treats district-level warning short-duration heavy rainfall in Beijing as a binary learning, composite radar reflectivity data Beijing–Tianjin–Hebei network from automatic weather stations used training performance evaluation. The results show ConvGRU approximately 25% faster than ConvLSTM. early-warning performances ConvLSTM have similar trends region, time, rain intensity, but most scores higher, few cases, has higher scores.

Язык: Английский

Процитировано

1
A Hybrid Power Load Forecasting Framework with Attention-Based Network and Multi-Scale Decomposition DOI
Jiaming Zhu, Dezhi Liu, Lili Niu

и другие.

Опубликована: Янв. 1, 2024

Язык: Английский

Процитировано

0
A Multi-Spatial Scale Ocean Sound Speed Prediction Method Based on Deep Learning DOI Creative Commons
Yü Liu, Benjun Ma, Zhiliang Qin

и другие.

Journal of Marine Science and Engineering, Год журнала: 2024, Номер 12(11), С. 1943 - 1943

Опубликована: Окт. 31, 2024

As sound speed is a fundamental parameter of ocean acoustic characteristics, its prediction central focus underwater acoustics research. Traditional numerical and statistical forecasting methods often exhibit suboptimal performance under complex conditions, whereas deep learning approaches demonstrate promising results. However, these methodologies fall short in adequately addressing multi-spatial coupling effects spatiotemporal weighting, particularly scenarios characterized by limited data availability. To investigate the interactions across multiple spatial scales to achieve accurate predictions, we propose STA-ConvLSTM framework that integrates attention mechanisms with convolutional long short-term memory neural networks (ConvLSTM). The core concept involves accounting for among various while extracting temporal information from assigning appropriate weights different entities. Furthermore, introduce an interpolation method temperature salinity based on KNN algorithm enhance dataset resolution. Experimental results indicate provides precise predictions speed. Specifically, relative measured data, it achieved root mean square error (RMSE) approximately 0.57 m/s absolute (MAE) about 0.29 m/s. Additionally, when compared single-dimensional analysis, incorporating scale considerations yielded superior predictive performance.

Язык: Английский

Процитировано

0