Multi-step prediction of dissolved oxygen in fish pond aquaculture using feature reconstruction-based deep neural network

Computers and Electronics in Agriculture, Journal Year: 2025, Volume and Issue: 232, P. 109997 - 109997

Published: Feb. 6, 2025

Language: Английский

---

BiGTA-Net: A Hybrid Deep Learning-Based Electrical Energy Forecasting Model for Building Energy Management Systems

Systems, Journal Year: 2023, Volume and Issue: 11(9), P. 456 - 456

Published: Sept. 2, 2023

The growth of urban areas and the management energy resources highlight need for precise short-term load forecasting (STLF) in systems to improve economic gains reduce peak usage. Traditional deep learning models STLF present challenges addressing these demands efficiently due their limitations modeling complex temporal dependencies processing large amounts data. This study presents a groundbreaking hybrid model, BiGTA-net, which integrates bi-directional gated recurrent unit (Bi-GRU), convolutional network (TCN), an attention mechanism. Designed explicitly day-ahead 24-point multistep-ahead building electricity consumption forecasting, BiGTA-net undergoes rigorous testing against diverse neural networks activation functions. Its performance is marked by lowest mean absolute percentage error (MAPE) 5.37 root squared (RMSE) 171.3 on educational dataset. Furthermore, it exhibits flexibility competitive accuracy Appliances Energy Prediction (AEP) Compared traditional models, reports remarkable average improvement approximately 36.9% MAPE. advancement emphasizes model’s significant contribution accentuating efficacy proposed approach power system optimizations smart city enhancements.

Language: Английский

---

Short-Term Power Load Forecasting Based on Secondary Cleaning and CNN-BILSTM-Attention

Energies, Journal Year: 2024, Volume and Issue: 17(16), P. 4142 - 4142

Published: Aug. 20, 2024

Accurate power load forecasting can provide crucial insights for system scheduling and energy planning. In this paper, to address the problem of low accuracy prediction, we propose a method that combines secondary data cleaning adaptive variational mode decomposition (VMD), convolutional neural networks (CNN), bi-directional long short-term memory (BILSTM), adding attention mechanism (AM). The Inner Mongolia electricity were first cleaned use K-means algorithm, then further refined with density-based spatial clustering applications noise (DBSCAN) algorithm. Subsequently, parameters VMD algorithm optimized using multi-strategy Cubic-T dung beetle optimization (CTDBO), after which was employed decompose twice-cleaned sequences into number intrinsic functions (IMFs) different frequencies. These IMFs used as inputs CNN-BILSTM-Attention model. model, CNN is feature extraction, BILSTM extracting information from sequence, AM assigning weights features optimize prediction results. It proved experimentally model proposed in paper achieves highest robustness compared other models exhibits high stability across time periods.

Language: Английский

---

Multi-Step Prediction of Wind Power Based on Hybrid Model with Improved Variational Mode Decomposition and Sequence-to-Sequence Network

Processes, Journal Year: 2024, Volume and Issue: 12(1), P. 191 - 191

Published: Jan. 15, 2024

Due to the complexity of wind power, traditional prediction models are incapable fully extracting hidden features multidimensional strong fluctuation data, which results in poor multi-step performance. To predict continuous power effectively future, an improved model combining variational mode decomposition (VMD) with sequence-to-sequence (Seq2Seq) is proposed. Firstly, sequence smoothed using VMD and parameters optimized by squirrel search algorithm (SSA) optimize effect. Then, subsequence obtained from decomposition, together original reconstructed into multivariate time series features. Finally, a Seq2Seq constructed, convolutional neural networks (CNNs) bidirectional gate recurrent units (BiGRUs) used learn coupling timing relationships input data encode them. The unit (GRU) decoded achieve prediction. Based on actual operating farm, case analysis conducted. Experimental show that SSA-VMD can effect, subsequences its highly accurate when applied predictions. has better than methods, as step size increases, advantages more obvious.

Language: Английский

---

PV Generation Prediction Using Multilayer Perceptron and Data Clustering for Energy Management Support

Energies, Journal Year: 2025, Volume and Issue: 18(6), P. 1378 - 1378

Published: March 11, 2025

Accurate PV power generation forecasting is critical to enable grid utilities manage energy effectively. This study presents an approach that combines machine learning with a clustering methodology improve the accuracy of predictions for management purposes. First, various models were compared, and multilayer perceptron (MLP) outperformed others by effectively capturing complex relationships between weather parameters output, obtaining following results: MSE: 3.069, RMSE: 1.752, MAE: 1.139. To performance MLP, characteristics are highly correlated outputs, such as irradiation sun elevation, grouped using K-means clustering. The elbow method identified four optimal clusters, individual MLP trained on each, reducing data complexity improving model focus. clustering-based significantly improved predictions, resulting in average metrics across all clusters following: 0.761, 0.756, 0.64. Despite these improvements, further research optimizing architecture required address inconsistencies achieve even better performance.

Language: Английский

---

Simulation and Reconstruction of Runoff in the High-Cold Mountains Area Based on Multiple Machine Learning Models

Water, Journal Year: 2023, Volume and Issue: 15(18), P. 3222 - 3222

Published: Sept. 10, 2023

Runoff from the high-cold mountains area (HCMA) is most important water resource in arid zone, and its accurate forecasting key to scientific management of resources downstream basin. Constrained by scarcity meteorological hydrological stations HCMA inconsistency observed time series, simulation reconstruction mountain runoff have always been a focus cold region research. Based on observations Yurungkash Kalakash Rivers, upstream tributaries Hotan River northern slope Kunlun Mountains at different periods, atmospheric circulation indices, we used feature analysis machine learning methods select input elements, train, simulate, preferences models runoffs two watersheds, reconstruct missing series River. The results show following. (1) Air temperature driver variability mountainous areas River, had strongest performance terms Pearson correlation coefficient (ρXY) random forest importance (FI) (ρXY = 0.63, FI 0.723), followed soil 0.043), precipitation, hours sunshine, wind speed, relative humidity, were weakly correlated. A total 12 elements selected as data. (2) Comparing simulated eight methods, found that gradient boosting performed best, AdaBoost Bagging with Nash–Sutcliffe efficiency coefficients (NSE) 0.84, 0.82, 0.78, while support vector regression (NSE 0.68), ridge 0.53), K-nearest neighbor 0.56), linear 0.51) poorly. (3) application four boosting, forest, AdaBoost, bagging, simulate for 1978–1998 was generally outstanding, NSE exceeding 0.75, reconstructing data period (1999–2019) could well reflect characteristics intra-annual inter-annual changes runoff.

Language: Английский

---

Short-Term Power Load Forecasting Using a VMD-Crossformer Model

Energies, Journal Year: 2024, Volume and Issue: 17(11), P. 2773 - 2773

Published: June 5, 2024

There are several complex and unpredictable aspects that affect the power grid. To make short-term load forecasting more accurate, a model utilizes VMD-Crossformer is suggested in this paper. First, ideal number of decomposition layers was ascertained using variational mode (VMD) parameter optimum approach based on Pearson correlation coefficient (PCC). Second, original data decomposed into multiple modal components VMD, then were reconstructed with components. Finally, input Crossformer network, which cross-dimensional dependence multivariate time series (MTS) prediction; is, dimension-segment-wise (DSW) embedding two-stage attention (TSA) layer designed to establish hierarchical encoder–decoder (HED), final prediction performed information from different scales. The experimental results show method could accurately predict electricity high accuracy reliability. MAE, MAPE, RMSE 61.532 MW, 1.841%, 84.486 respectively, for dataset I. 68.906 0.847%, 89.209 II. Compared other models, paper predicted better.

Language: Английский

---

Pneumonia Classification using deep learning: a comparative study

Published: April 21, 2024

Pneumonia is an infectious disease of the lungs, caused by viruses, bacteria or fungi. distinguished acute inflammation lung tissue, causing consolidation terminal bronchioles and alveoli. According to WHO (the World Health Organization), this causes about 4 million deaths. Among methods diagnosing pneumonia uses a chest X-ray. This widely used visualize pulmonary abnormalities. work aims at detection characterization development computer-assisted diagnosis systems (DAOC). We use deep learning algorithms X-ray images from database examine classification pneumonia. To improve accuracy, we compare several models. research contributes meeting growing demand for medical personnel addressing global incidence disorders.

Language: Английский

---

Short Term Load Forecasting for Smart Distribution System Planning Using Deep Neural Networks: A Hybrid Approach

International Journal of Electrical and Electronics Engineering, Journal Year: 2024, Volume and Issue: 11(5), P. 138 - 149

Published: May 31, 2024

Accurate load forecasting plays a crucial role in the management and control of electrical power distribution systems. Short-Term Load Forecasting (STLF) is particularly vital for planning, as it provides precise predictions immediate future. This paper introduces an innovative hybrid deep-learning model specifically designed STLF The proposed combines strengths Bidirectional Long Memory (Bi-LSTM) Gated Recurrent Unit (GRU) networks. study utilizes high-resolution real-world dataset, consisting historical consumption weather-related features, with 30-minute intervals from period January 1, 2006, to December 31, 2010. benchmarked against prominent standalone models such Bi-LSTM, GRU, LSTM, CNN, like CNN-LSTM ConvLSTM-GRU. model's performance evaluated using various validation metrics, including Rsquared error, Root Mean Squared Error (RMSE), (MSE), Absolute (MAE), Percentage (MAPE). results show that outperforms all conventional models, offering significant improvements forecast accuracy. Thus, highlights potential revolutionizing methodologies, paving way smart system.

Language: Английский

---

Fine-Grained Agricultural Facility Power Forecasting Based on Empirical Mode Decomposition

2022 International Joint Conference on Neural Networks (IJCNN), Journal Year: 2024, Volume and Issue: 47, P. 1 - 8

Published: June 30, 2024

Language: Английский

---