Load Day-Ahead Automatic Generation Control Reserve Capacity Demand Prediction Based on the Attention-BiLSTM Network Model Optimized by Improved Whale Algorithm

Energies, Год журнала: 2024, Номер 17(2), С. 415 - 415

Опубликована: Янв. 15, 2024

Load forecasting is a research hotspot in academia; the context of new power systems, prediction and determination load reserve capacity also important. In order to adapt forms day-ahead automatic generation control (AGC) demand method based on Fourier transform attention mechanism combined with bidirectional long short-term memory neural network model (Attention-BiLSTM) optimized by an improved whale optimization algorithm (IWOA) proposed. Firstly, response time, used refine distinction between various types demand, AGC band calculated using Parseval’s theorem obtain sequence. The maximum mutual information coefficient explore relevant influencing factors sequence concerning data characteristics Then, historical daily sequences features are input into Attention-BiLSTM model, automatically find optimal hyperparameters better results. Finally, arithmetic simulation results show that proposed this paper has best performance upper (0.8810) lower (0.6651) bounds (R2) higher than other models, it smallest mean absolute percentage error (MAPE) root square (RMSE).

Язык: Английский

---

Land-atmosphere and ocean–atmosphere couplings dominate the dynamics of agricultural drought predictability in the Loess Plateau, China

Journal of Hydrology, Год журнала: 2024, Номер unknown, С. 132225 - 132225

Опубликована: Окт. 1, 2024

Язык: Английский

---

Predicting Regional Heavy Precipitation Occurrences in the Southwest Iran Using Machine Learning Models and Atmospheric Variables

Опубликована: Янв. 1, 2025

This study assessed the effectiveness of various machine learning models and logistic regression for predicting regional heavy precipitation events in southwest Iran. We used a time-delay scenario, analyzing atmospheric data from one to five days preceding events. Feature selection methods averaging techniques were compared optimize model performance. Random Forest (RF) achieved highest overall accuracy (0.848) using 1-4 prior with "Both" Chi-Square feature selection. While RF outperformed decision trees, remained competitive (accuracy 0.804) specific methods. Statistical tests showed no significant differences between models. Zonal wind humidity emerged as crucial predictor variables, particularly model. Analyzing Outgoing Longwave Radiation (OLR) vapor flux anomalies revealed consistent sequence leading precipitation. Negative OLR indicated strong initial convection, followed by intensification eastward movement Mediterranean Sea cyclone. enhanced surrounding water bodies, culminating These findings offer valuable insights improving weather forecasting early warning systems, especially regions vulnerable extreme weather.

Язык: Английский

---

Moving beyond post hoc explainable artificial intelligence: a perspective paper on lessons learned from dynamical climate modeling

Geoscientific model development, Год журнала: 2025, Номер 18(3), С. 787 - 802

Опубликована: Фев. 11, 2025

Abstract. AI models are criticized as being black boxes, potentially subjecting climate science to greater uncertainty. Explainable artificial intelligence (XAI) has been proposed probe and increase trust. In this review perspective paper, we suggest that, in addition using XAI methods, researchers can learn from past successes the development of physics-based dynamical models. Dynamical complex but have gained trust because their failures sometimes be attributed specific components or sub-models, such when model bias is explained by pointing a particular parameterization. We propose three types understanding basis evaluate alike: (1) instrumental understanding, which obtained passed functional test; (2) statistical make sense modeling results techniques identify input–output relationships; (3) component-level refers modelers' ability point parts architecture culprit for erratic behaviors crucial reason why functions well. demonstrate how sought achieved via intercomparison projects over several decades. Such routinely leads improvements may also serve template thinking about AI-driven science. Currently, methods help explain focusing on mapping between input output, thereby increasing Yet, further our models, will build that interpretable amenable understanding. give recent examples literature highlight some recent, albeit limited, achieving explaining behavior. The merit they stronger and, extension, downstream uses data.

Язык: Английский

---

Optimizing rainfall prediction in coastal and inland areas: a comparative analysis of forecasting models in eThekwini district, South Africa

International Journal of Business Ecosystem and Strategy (2687-2293), Год журнала: 2025, Номер 7(1), С. 180 - 197

Опубликована: Март 7, 2025

While floods and droughts are natural occurrences in the earth’s hydrological cycle, their escalating frequency intensity have become a major concern for governments throughout globe. Developing nations, such as South Africa, weary of these extreme weather events because they understand lack necessary resources infrastructure to deal with them. The eThekwini Municipality serves prime example how vulnerable developing nations' regions devastating effects droughts, multiple devastated area, resulting fatalities, damaging public infrastructure, demolishing houses. scale damage from reveals that significant gaps exist disaster preparedness Region. Rainfall forecasting is vital tool has been underutilised can be used preemptively manage or mitigate flooding enhance water resource management region. Machine learning models particular very useful rainfall forecasting; hence, goal this study was evaluate most efficient precipitation northern central regions, which coastal inland areas, respectively. data spanning 32 years obtained meteorological stations both SARIMA, ARIMA, ETS machine were evaluated based on ability capture seasonal patterns, handle non-stationarity, provide accurate predictions. Model performance analysed, comparisons made using root mean squared error (RMSE), absolute (MAE), scaled (MASE) evaluation metrics. study's findings indicate effective SARIMA (0,0,0) (2,0,0) [12] (1,0,0) [12]. These valuable insights meteorologists, hydrologists, policymakers involved regional climate modelling management.

Язык: Английский

---

A Machine Learning Model Superior to Dynamic Subseasonal Temperature Forecasting

Hans Journal of Data Mining, Год журнала: 2025, Номер 15(02), С. 176 - 183

Опубликована: Янв. 1, 2025

Язык: Английский

---

A self-organizing interval type-2 fuzzy neural network for multi-step time series prediction

Applied Soft Computing, Год журнала: 2025, Номер unknown, С. 113221 - 113221

Опубликована: Май 1, 2025

Язык: Английский

---

Optimization of graph wavenet model for dissolved oxygen prediction using self-distillation and whale optimization algorithm

Journal of Water Process Engineering, Год журнала: 2025, Номер 75, С. 108013 - 108013

Опубликована: Май 29, 2025

Язык: Английский

---

Energy-driven TBM health status estimation with a hybrid deep learning approach

Expert Systems with Applications, Год журнала: 2024, Номер 249, С. 123701 - 123701

Опубликована: Март 19, 2024

Язык: Английский

---

Moving beyond post-hoc XAI: Lessons learned from dynamical climate modeling

Опубликована: Янв. 30, 2024

Abstract. AI models are criticized as being black boxes, potentially subjecting climate science to greater uncertainty. Explainable artificial intelligence (XAI) has been proposed probe and increase trust. In this Perspective, we suggest that, in addition using XAI methods, researchers can learn from past successes the development of physics-based dynamical models. Dynamical complex but have gained trust because their failures be attributed specific components or sub-models, such when model bias is explained by pointing a particular parameterization. We propose three types understanding basis evaluate alike: (1) instrumental understanding, which obtained passed functional test; (2) statistical make sense modelling results techniques identify input-output relationships; (3) Component-level refers modelers’ ability point parts architecture culprit for erratic behaviors crucial reason why functions well. demonstrate how component-level sought achieved via intercomparison projects over several decades. Such routinely leads improvements may also serve template thinking about AI-driven science. Currently, methods help explain focusing on mapping between input output, thereby increasing Yet, further our models, will build that interpretable amenable understanding. give recent examples literature highlight some recent, albeit limited, achieving explaining behaviour. The merit they stronger modeling and, extension, downstream uses data.

Язык: Английский

---