Load Forecasting for the Laser Metal Processing Industry Using VMD and Hybrid Deep Learning Models

Energies, Journal Year: 2023, Volume and Issue: 16(14), P. 5381 - 5381

Published: July 14, 2023

Electric load forecasting is crucial for the metallurgy industry because it enables effective resource allocation, production scheduling, and optimized energy management. To achieve an accurate forecasting, essential to develop efficient approach. In this study, we considered time factor of univariate time-series data implement various deep learning models predicting one hour ahead under different conditions (seasonal daily variations). The goal was identify most suitable model each specific condition. two hybrid were proposed. first combines variational mode decomposition (VMD) with a convolutional neural network (CNN) gated recurrent unit (GRU). second incorporates VMD CNN long short-term memory (LSTM). proposed outperformed baseline models. VMD–CNN–LSTM performed well seasonal conditions, average RMSE 12.215 kW, MAE 9.543 MAPE 0.095%. Meanwhile, VMD–CNN–GRU variations, value 11.595 9.092 0.079%. findings support practical application electrical in diverse scenarios, especially concerning variations.

Language: Английский

---

A multiscale adaptive framework based on convolutional neural network: Application to fluid catalytic cracking product yield prediction

Petroleum Science, Journal Year: 2024, Volume and Issue: 21(4), P. 2849 - 2869

Published: Jan. 22, 2024

Since chemical processes are highly non-linear and multiscale, it is vital to deeply mine the multiscale coupling relationships embedded in massive process data for prediction anomaly tracing of crucial parameters production indicators. While integrated method adaptive signal decomposition combined with time series models could effectively predict variables, does have limitations capturing high-frequency detail operation state when applied complex processes. In light this, a novel Multiscale Multi-radius Multi-step Convolutional Neural Network (MsrtNet) proposed mining spatiotemporal information. First, industrial from Fluid Catalytic Cracking (FCC) using Complete Ensemble Empirical Mode Decomposition Adaptive Noise (CEEMDAN) extract multi-energy scale information feature subset. Then, convolution kernels varying stride padding structures established decouple long-period encapsulated within data. Finally, reconciliation network trained reconstruct results obtain final output. MsrtNet initially assessed its capability untangle among variables Tennessee Eastman (TEP). Subsequently, performance evaluated predicting product yield 2.80 × 106 t/a FCC unit, taking diesel gasoline as examples. conclusion, can achieve maximum reduction 11% error compared other time-series models. Furthermore, robustness transferability underscore promising potential broader applications.

Language: Английский

---

Improving short-term forecasting of solar power generation by using an EEMD-BiGRU model: A comparative study based on seven standalone models and six hybrid models

International Journal of Green Energy, Journal Year: 2024, Volume and Issue: 21(14), P. 3135 - 3158

Published: May 27, 2024

Accurate and timely forecasting is critical for grid-connected solar power safety stability, achieved through machine learning (ML) both common real-time applications. To mitigate the impact of nonstationarity volatility in generation, we employed empirical mode decomposition (EMD), ensemble (EEMD), variational (VMD), complete EEMD with adaptive noise (CEEMDAN) to decompose time series into frequency components, reducing fluctuations noise. A combination four methods (EMD, EEMD, VMD, CEEMDAN) two ML models, bidirectional gated recurrent unit (BiGRU) long short-term memory (BiLSTM) were utilized construct six hybrid models (EMD-BiLSTM, EMD-BiGRU, EEMD-BiLSTM, EEMD-BiGRU, VMD-BiGRU, CEEMDAN-BiGRU), which validated on a dataset from 20 MW station Hebei, compared seven standalone backpropagation neural networks (BPNN), support vector machines (SVM), (LSTM), (GRU), convolutional (CNN), BiLSTM, BiGRU, these demonstrated enhanced forecast accuracy. Of these, CEEMDAN-BiGRU significantly reduced prediction errors, percentage reductions root mean square error (RMSE), absolute (MAE), (MAPE) ranging 44.18 ~ 49.43%, 43.67 48.59%, 44.64% ~53.53%, respectively. The EEMD-BiGRU model outperformed all achieving an RMSE 0.7662, MAE 0.3990, MAPE 7.982%, R2 0.9865. findings this study can provide insights applying based generation.

Language: Английский

---

Prediction of Short-Term Solar Irradiance Using the ProbSparse Attention Mechanism for a Sustainable Energy Development Strategy

Sustainability, Journal Year: 2025, Volume and Issue: 17(3), P. 1075 - 1075

Published: Jan. 28, 2025

Sustainability refers to a development approach that meets the needs of present generation without compromising ability future generations meet their own needs. Solar energy is an inexhaustible and renewable resource. From perspective resource utilization, solar power has high degree sustainability. Therefore, one most important ways transform structure promote sustainable economy society, it great significance for promoting construction resource-conserving environmentally friendly society. However, resources also exhibit strong unpredictability; therefore, this paper proposes novel artificial intelligence (AI) model short-term irradiance prediction in photovoltaic generation. Leveraging ProbSparse attention mechanism within encoder-decoder architecture, AI efficiently captures both short- long-term dependencies input sequence. The dingo algorithm innovatively redesigned optimize hyperparameters proposed model, enhancing convergence. Data preprocessing involves feature selection based on mutual information, multiple imputations data cleaning, median filtering. Evaluation metrics include mean absolute error (MAE), root square (RMSE), coefficient determination (R2). demonstrates improved efficiency robust performance prediction, contributing advancements management electrical systems.

Language: Английский

---

Power generation efficiency and resources saving of the hydropower industry using the extended data based convolutional neural network

Renewable Energy, Journal Year: 2025, Volume and Issue: unknown, P. 122541 - 122541

Published: Jan. 1, 2025

Language: Английский

---