Comparative Analysis of Multiple Deep Learning Models for Forecasting Monthly Ambient PM2.5 Concentrations: A Case Study in Dezhou City, China

Atmosphere, Journal Year: 2024, Volume and Issue: 15(12), P. 1432 - 1432

Published: Nov. 28, 2024

Ambient air pollution affects human health, vegetative growth and sustainable socio-economic development. Therefore, data in Dezhou City China are collected from January 2014 to December 2023, multiple deep learning models used forecast PM2.5 concentrations. The ability of the is evaluated compared with observed using various statistical parameters. Although all eight can accomplish forecasting assignments, precision accuracy CNN-GRU-LSTM method 34.28% higher than that ANN method. result shows has best performance other seven models, achieving an R (correlation coefficient) 0.9686 RMSE (root mean square error) 4.6491 μg/m3. values CNN, GRU LSTM 57.00%, 35.98% 32.78% method, respectively. results reveal predictor remarkably improves performances benchmark overall forecasting. This research provides a new perspective for predictive ambient model provide scientific basis prevention control.

Language: Английский

---

Enhanced load forecasting for distributed multi-energy system: A stacking ensemble learning method with deep reinforcement learning and model fusion

Energy, Journal Year: 2025, Volume and Issue: unknown, P. 135031 - 135031

Published: Feb. 1, 2025

Language: Английский

---

A Hybrid Wavelet-Based Deep Learning Model for Accurate Prediction of Daily Surface PM2.5 Concentrations in Guangzhou City

Toxics, Journal Year: 2025, Volume and Issue: 13(4), P. 254 - 254

Published: March 28, 2025

Surface air pollution affects ecosystems and people’s health. However, traditional models have low prediction accuracy. Therefore, a hybrid model for accurately predicting daily surface PM2.5 concentrations was integrated with wavelet (W), convolutional neural network (CNN), bidirectional long short-term memory (BiLSTM), gated recurrent unit (BiGRU). The data meteorological factors pollutants in Guangzhou City from 2014 to 2020 were utilized as inputs the models. W-CNN-BiGRU-BiLSTM demonstrated strong performance during phase, achieving an R (correlation coefficient) of 0.9952, root mean square error (RMSE) 1.4935 μg/m3, absolute (MAE) 1.2091 percentage (MAPE) 7.3782%. Correspondingly, accurate is beneficial control urban planning.

Language: Английский

---

Predicting water demand for spraying operations in dry bulk ports: A hybrid approach based on data decomposition and deep learning

Advanced Engineering Informatics, Journal Year: 2025, Volume and Issue: 65, P. 103313 - 103313

Published: April 2, 2025

Language: Английский

---

A novel model combined with deep learning and Kalman filter augmentation for route-level bus arrival time prediction with error accumulation mitigation

Expert Systems with Applications, Journal Year: 2025, Volume and Issue: unknown, P. 127622 - 127622

Published: April 1, 2025

Language: Английский

---

Improved Prediction of Hourly PM2.5 Concentrations with a Long Short-Term Memory Optimized by Stacking Ensemble Learning and Ant Colony Optimization

Toxics, Journal Year: 2025, Volume and Issue: 13(5), P. 327 - 327

Published: April 23, 2025

To address the performance degradation in existing PM2.5 prediction models caused by excessive complexity, poor spatiotemporal efficiency, and suboptimal parameter optimization, we employ stacking ensemble learning for feature weighting analysis integrate ant colony optimization (ACO) algorithm model optimization. Combining meteorological collaborative pollutant data, a (namely stacking-ACO-LSTM model) with much shorter consuming time than that of only long short-term memory (LSTM) networks suitable concentration is established. It can effectively filter out variables higher weights, thereby reducing predictive power model. The hourly trained tested using real-time monitoring data Nanchang City from 2017 to 2019. results show established has high accuracy predicting concentration, compared same without considering space efficiency defective mean square error (MSE) decreases about 99.88%, coefficient determination (R2) increases 2.39%. This study provides new idea cities.

Language: Английский

---

A novel diagnosis methodology of gear oil for wind turbine combining stepwise multivariate regression and clustered federated learning framework

Research Square (Research Square), Journal Year: 2025, Volume and Issue: unknown

Published: April 23, 2025

Data-driven approaches demonstrate significant potential in accurately diagnosing faults wind turbines. To enhance diagnostic performance, we introduce a clustered federated learning framework (CFLF) to gear oil diagnosis. Initially, stepwise multivariate regression (SMR) model is introduced and optimized after data process, which integrates multiscale feature AIC diagnosis feature. Subsequently, tackle heterogeneity among different indicators, canonical correlation series of representations are extracted from the SMR models, combining CFLF method proposed assess performance oil. Actual analysis turbine showcase superior over single with higher prediction accuracy 35.73%. This study provides new technique for evaluating energy sector.

Language: Английский

---

Enhanced Load Forecasting for Distributed Multi-Energy System: A Stacking Ensemble Learning Method With Deep Reinforcement Learning And Model Fusion

Published: Jan. 1, 2025

Accurate multi-energy load forecasting for distributed systems is facing challenges due to the complexity of coupling and inherent stochasticity. In this regard, a novel stacking ensemble learning model based on reinforcement (RL) fusion proposed. First, feature selection performed using maximal information coefficient (MIC), data decomposed reconstructed through complete empirical mode decomposition with adaptive noise (CEEMDAN) sample entropy (SE). Subsequently, fused models strong predictive capabilities are selected as base learners, RL deep deterministic policy gradient (DDPG) excellent ability meta-learner. Next, hyperparameters learners optimized an improved arctic puffin optimization (APO) algorithm. Finally, constructed K-fold cross-validation. Tests real-world datasets demonstrate that proposed method achieves smaller prediction errors, enhanced robustness, greater reliability. Moreover, careful learner utilization meta-learner, up 1.53% improvement in determination (R²), 36.09% increase improves residual deviation (RPD), 102.96% reduction rooted mean square error (RMSE).

Language: Английский

---

Decomposing Spatio-Temporal Heterogeneity: Matrix-Informed Ensemble Learning for Interpretable Prediction

Knowledge-Based Systems, Journal Year: 2024, Volume and Issue: unknown, P. 112906 - 112906

Published: Dec. 1, 2024

Language: Английский

---