The rapid deployment of photovoltaic generation increases uncertainty in power-system operation and strengthens the need for ultra-short-term forecasts with reliable uncertainty estimates. Point-forecasting approaches alone are often insufficient for dispatch and reserve decisions because they do not quantify risk. This study investigates probabilistic forecasting of short-horizon solar generation using quantile regression on a public dataset of solar output and meteorological variables. This study proposes a hybrid attention–convolution model that combines an attention-based encoder to capture long-range temporal dependencies with a causal temporal convolution module that extracts fast local fluctuations using only past information, preventing information leakage. The two representations are fused and decoded jointly across multiple future horizons to produce consistent quantile trajectories. Experiments against representative machine-learning and deep-learning baselines show improved probabilistic accuracy and competitive central forecasts, while illustrating an important sharpness–calibration trade-off relevant to risk-aware grid operation. Key novelties include a multi-horizon quantile formulation at 15 min resolution for one-hour-ahead PV increments, a HybridCNN–STTransformer that fuses causal temporal convolutions with Transformer attention, and a horizon-token decoder that models inter-horizon dependencies to produce consistent multi-step quantile trajectories; reliability/sharpness diagnostics and post hoc calibration are discussed for operational risk-aware use.

We propose the Hybrid Spatio-Temporal Transformer with Multi-Branch CNN/RNN and Extreme-Event Head (HST–MB–CREH), a hybrid spatio-temporal deep learning architecture for joint short-term photovoltaic (PV) power forecasting and the detection of rare extreme events, to support the reliable operation of renewable-rich power systems. The model combines a spatio-temporal transformer encoder with three convolutional neural network (CNN)/recurrent neural network (RNN) branches (CNN → long short-term memory (LSTM), LSTM → gated recurrent unit (GRU), CNN → GRU) and a dense pathway for tabular meteorological and calendar features. A multitask output head simultaneously performs the regression of PV power and binary classification of extremes defined above the 95th percentile. We evaluate HST–MB–CREH on the publicly available Renewable Power Generation and Weather Conditions dataset with hourly resolutions from 2017 to 2022, using a 5-fold TimeSeriesSplit protocol to avoid temporal leakage and to cover multiple seasons. Compared with tree ensembles (RandomForest, XGBoost), recurrent baselines (Stacked GRU, LSTM), and advanced hybrid/transformer models (Hybrid Multi-Branch CNN–LSTM/GRU with Dense Path and Extreme-Event Head (HMB–CLED) and Spatio-Temporal Multitask Transformer with Extreme-Event Head (STM–EEH)), the proposed architecture achieves the best overall trade-off between accuracy and rare-event sensitivity, with normalized performance of RMSE_z = 0.2159 ± 0.0167, MAE_z = 0.1100 ± 0.0085, mean absolute percentage error (MAPE) = 9.17 ± 0.45%, R² = 0.9534 ± 0.0072, and AUC_ext = 0.9851 ± 0.0051 across folds. Knowledge extraction is supported via attention-based analysis and permutation feature importance, which highlight the dominant role of global horizontal irradiance, diurnal harmonics, and solar geometry features. The results indicate that hybrid spatio-temporal multitask architectures can substantially improve both the forecast accuracy and robustness to extremes, making HST–MB–CREH a promising building block for intelligent decision-support tools in smart grids with a high share of PV generation.

Multi-parametric industrial time-series forecasting is important for predictive monitoring, soft sensing, and decision support in complex process industries. This study proposes a process-informed hybrid ARX-residual deep sequence architecture for short-horizon multi-output forecasting of fluid catalytic cracking unit (FCCU) product yields. The model combines a frozen linear ARX branch with causal temporal convolution, a three-layer GRU, causal self-attention, additive attention pooling, residual gating, and a consistency-regularized multi-output objective. Experiments were performed on an open simulated FCCU benchmark containing 20,160 one-minute observations from seven normal and disturbed operating scenarios. The leakage-controlled pipeline uses scenario-wise chronological splitting, train-only median imputation and standardization, exclusion of fault-timing metadata and direct algebraic target components, and 30 min causal windows for five-minute-ahead forecasting. Across ten prespecified random seeds, the proposed model achieved a historical-test RMSE of 0.022809 $\pm$ 0.000116, compared with 0.022751 for ARX-like Ridge. Dependence-aware statistical analyses did not establish a consistent statistical advantage for either model across the evaluated runs. Rolling-origin evaluation likewise showed similar point accuracy, while leave-one-scenario-out evaluation revealed severe extrapolation failure for the unseen pressure-drop scenario. The results support the reproducibility of the proposed architecture and the importance of the ARX-based dynamic structure, but they do not establish the statistical superiority of the full deep residual architecture. Independent industrial data or newly prespecified simulation trajectories are required for confirmatory external validation.

The paper presents a proposed clinical intelligence system, which is called NLP-Bayesian Optimization Clinical Model (NBO-CM), to study large-scale unstructured electronic health record narratives in the MIMIC-IV discharge and radiology note datasets with the help of a structured pipeline of clinical text preprocessing, feature extraction and probabilistic modeling to solve the linguistic variability, missing information and uncertainty in clinical decisions. Text preprocessing methods are first applied in the standardization of clinical narratives, such as text tokenization, text normalization, text lemmatization, text segmentation, and text negation detection. Next, the unstructured text is converted into structured clinical variables with the help of feature extraction techniques, including named entity recognition, medical concept normalization using UMLS/SNOMED ontologies, generating contextual embeddings, and vectorizing text using the Term Frequency–Inverse Document Frequency (TF-IDF) technique. Bayesian Networks are then used to model these features, with dependency structure learning based on the Poor-and-Rich Optimization algorithm (PRO), having the ability to explore probabilistic relationships efficiently, and parameter estimation based on expectation–maximization, providing robust learning with incomplete and uncertain conditions of data. Lastly, probabilistic reasoning and inference are used to predict diseases, prioritize risks and make clinical inferences with clear measures of uncertainty and interpretability. Experimental analysis of real-world MIMIC-IV clinical notes indicates that the framework is more effective in terms of diagnostic accuracy, predictive strength, and clinical explainability than traditional machine learning methods, resulting in a scaled and explainable framework of intelligent clinical decision support systems in complex care settings.

Modern power transmission grids are increasingly operated under volatile load, variable generation, and near-limit line-flow conditions. In such environments, deterministic line-flow forecasting is insufficient for operational decision support because operators require calibrated risk probabilities, uncertainty intervals, and reliable early warning signals. This article proposes B-TGPRF, a Bayesian Temporal Graph Probabilistic Risk Forecaster for calibrated overload-risk forecasting in power transmission grids. The proposed framework is positioned as a hybrid probabilistic graph-temporal forecasting system rather than as a new end-to-end graph neural network; its novelty lies in the leakage-controlled sequential combination of temporal forecasting, electrical graph descriptors, calibrated Bayesian risk estimation, residual correction, and compact interval uncertainty assessment. The model integrates temporal line-flow, load, and generation features with graph-topological descriptors, operating-regime indicators, residual correction, conformal interval estimation, probability calibration, and a Bayesian risk layer. A leakage-controlled data preparation pipeline was built using an open large-scale benchmark for machine learning applications in transmission grids. The final modeling dataset contains more than 3.48 million observations, 100 selected critical lines, train-only risk thresholds, and chronological train, validation, test, and external-like scenario splits. B-TGPRF was compared with persistence baselines, linear models, Bayesian baselines, tree ensembles, boosting models, neural temporal models, and compact state-of-the-art-style temporal and graph-temporal architectures. On the strict external-like test, B-TGPRF achieved MAE = 0.3650, RMSE = 0.5478, R² = 0.9962, Brier Score = 0.0147, and ECE = 0.0064. The results show that the proposed model provides a strong overall balance between line-flow forecasting accuracy, calibrated risk estimation, compact interval prediction, and low false-positive risk-signaling, while boosting models remain highly competitive for pure risk-class detection.

This study evaluates a causal, six-target, 30 min forecasting component for diesel hydrotreating using campaign-separated synthetic data, not a deployed digital twin. An empirical, engineering-constrained simulator generated 30 original campaigns; the recovered code reproduced their datasets byte-for-byte but the simulator did not implement mechanistic balances. The archived AURA-SSM hybrid combines a frozen Ridge anchor, a released measurement observer, multiscale context, bounded corrections, and conditional empirical risk weights; despite its identifier, it has no learned state-space layer. On two previously inspected evaluation domains, frozen NRMSE was 0.16996 and 0.18085, versus ten-degradation means of 0.19170 and 0.20490. These retrospective improvements over leading frozen controls were below the prespecified 3% practical threshold; adaptive Ridge was more accurate on average. Additional matched controls, CRISP/TimeFormer adaptations, and a prespecified 16-campaign synthetic evaluation clarified limitations. On new common/modified simulator structures, AURA-SSM achieved ten-degradation NRMSE of 0.18804/0.20332, outperforming NLinear but not a fixed convex blend; context LightGBM was more accurate with clean measurements. A separately corrected observer passed causal tests and improved prolonged-outage prediction without replacing archived results. Interval undercoverage, lack of mechanistic validation, and absent independent industrial data constrain applicability.