Temporal Features Research Articles

DAB-LSTMNN: A novel respiratory disease diagnosis deep CNN framework based on attention scheme and BLSTM neural networks

Respiratory diseases have a significant impact on modern society, posing considerable public health risks. With the emergence of the COVID-19 pandemic, addressing respiratory issues has become increasingly urgent and important. Recently, artificial intelligence-based methods utilizing the acoustic recordings of suspected patients have shown promise in the diagnosis of various respiratory diseases, enabling localized treatment and containment of their spread. As the existing methods cannot efficiently extract subtle differences between the sound samples, thereby limiting their generalization ability. To improve the diagnosis accuracy, this paper proposes a novel multi-channel, multi-modal deep learning architecture based on the attention mechanism. The proposed framework combines a deep convolutional neural network (DCNN) with a bidirectional long short-term memory (BLSTM) network, and also utilizes the attention scheme to extract temporal and spectral features from different modalities of speech data (e.g. coughs, counting sounds and sustained vowel articulation). The proposed method effectively classifies COVID-19 patients, asthma patients and healthy individuals, with a test accuracy of 89.27 ± 0.1%, and an F1 score of 85.42 ± 0.2%. The experimental results validate the feasibility of our method, and also indicate that it is competitive with the existing deep networks.

Predicting goal probabilities with improved xG models using event sequences in association football.

In association football, predicting the likelihood and outcome of a shot at a goal is useful but challenging. Expected goal (xG) models can be used in a variety of ways including evaluating performance and designing offensive strategies. This study proposed a novel framework that uses the events preceding a shot, to improve the accuracy of the expected goals (xG) metric. A combination of previously explored and unexplored temporal features is utilized in the proposed framework. The new features include; "advancement factor", and "player position column". A random forest model was used, which performed better than published single-event-based models in the literature. Results further demonstrated a significant improvement in model performance with the inclusion of preceding event information. The proposed framework and model enable the discovery of event sequences that improve xG, which include; opportunities built up from the sides of the 18-yard box, shots attempted from in front of the goal within the opposition's 18-yard box, and shots from successful passes to the far post.

Few-shot Learning for Sign Language Recognition with Embedding Propagation

Sign language is a primary channel for the deaf and hard-hearing to communicate. Sign language consists of many signs with different variations in hand shapes, motion patterns, and positioning of hands, faces, and body parts. This makes sign language recognition (SLR) a challenging field in computer vision research. This paper tackles the problem of few-shot SLR, where models trained on known sign classes are utilized to recognize instances of unseen signs with only a few examples. In this approach, a transformer encoder is employed to learn the spatial and temporal features of sign gestures, and an embedding propagation technique is used to project these features into the embedding space. Subsequently, a label propagation method is applied to smooth the resulting embeddings. The obtained results demonstrate that combining embedding propagation with label propagation enhances the performance of the SLR system and achieved an accuracy of 76.6%, which surpasses the traditional few-shot prototypical network's accuracy of 72.4%.

A Deep Learning PM2.5 Hybrid Prediction Model Based on Clustering–Secondary Decomposition Strategy

Accurate prediction of PM2.5 concentration is important for pollution control, public health, and ecological protection. However, due to the nonlinear nature of PM2.5 data, the accuracy of existing methods suffers and performs poorly in both short-term and long-term predictions. In this study, a deep learning hybrid prediction model based on clustering and quadratic decomposition is proposed. The model utilizes the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) to decompose the PM2.5 sequences into multiple intrinsic modal function components (IMFs), and clusters and re-fuses the subsequences with similar complexity by permutation entropy (PE) and K-means clustering. For the fused high-frequency sequences, a secondary decomposition is performed using the whale optimization algorithm (WOA) optimized variational modal decomposition (VMD). Finally, the nonlinear and temporal features are captured for prediction using the long- and short-term memory neural network (LSTM). Experiments show that this proposed model exhibits good stability and generalization ability. It does not only make accurate predictions in the short term, but also captures the trends in the long-term prediction. There is a significant performance improvement over the baseline models. Further comparisons with existing models outperform the current state-of-the-art models.

Spatial–Temporal Characteristics and Driving Factors of Visible and Invisible Non-Grain Production of Cultivated Land in Hebei Province Based on GlobeLand 30 and MODIS-EVI

The growing problem of non-grain production of cultivated land (NGPOCL) has increased food security risk, garnering attention from China and other nations worldwide. Current research predominantly focuses on the internal planting structure of cultivated land. To more comprehensively measure the level of NGPOCL, we categorized NGPOCL into two types: visible non-grain production of cultivated land (VNGPOCL) and invisible non-grain production of cultivated land (INGPOCL). VNGPOCL and INGPOCL scopes were extracted utilizing land use and vegetation index data, exploring their spatial–temporal characteristics and driving factors through spatial feature analysis and multiple linear regression methods. The findings are as follows: (1) The degree of VNGPOCL shifted from mild to moderate, with its rate increasing from 5.16% in 2000–2010 to 10.82% in 2010–2020. Furthermore, the spatial variation in VNGPOCL indicated a growing east–west disparity while showing a reduction in north–south differences, reflecting significant spatial agglomeration effects. (2) There was a dramatic increase in areas classified as having moderate to severe INGPOCL, with the rate rising from 14.24% in 2000 to 41.47% by 2020. The east–west and north–south disparities concerning INGPOCL diminished rapidly, also indicating strong spatial agglomeration effects. (3) The driving factors for VNGPOCL and INGPOCL differed significantly depending on developmental stages. The results contribute valuable insights into accurately characterizing the spatial–temporal features associated with NGPOCL in Hebei Province while enhancing risk management strategies related to NGPOCL.

AI-driven optimization of agricultural water management for enhanced sustainability

Optimizing agricultural water resource management is crucial for food production, as effective water management can significantly improve irrigation efficiency and crop yields. Currently, precise agricultural water demand forecasting and management have become key research focuses; however, existing methods often fail to capture complex spatial and temporal dependencies. To address this, we propose a novel deep learning framework that combines remote sensing technology with the UNet-ConvLSTM (UCL) model to effectively integrate spatial and temporal features from MODIS and GLDAS datasets. Our model leverages the high-resolution spatial data from UNet and the temporal dependencies captured by ConvLSTM to significantly improve prediction accuracy. Experimental results demonstrate that our UCL model achieves the best R2\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$R^2$$\\end{document} compared to existing methods, reaching 0.927 on the MODIS dataset and 0.935 on the GLDAS dataset. This approach highlights the potential of AI and remote sensing technologies in addressing critical challenges in agricultural water management, contributing to more sustainable and efficient food production systems.

Prediction of QT interval in ECGs through baseline wander removal and deep learning technique

Abstract Background In electrocardiograms (ECGs), the QT interval, calculated from the onset of the Q-wave to the end of the T-wave, reflects the repolarization status of the heart. Currently, the QT interval automatically measured in 12-lead electrocardiograms is inaccurately derived as the median value, rather than utilizing the more precise mean value of QT intervals from all leads. This approach fails to provide information on heterogeneous QT prolongation crucial for predicting arrhythmias, as it does not accurately represent the values of QT intervals in each lead. Purpose Traditional analysis of ECGs based on QT intervals necessitates direct interpretation by cardiologists, leading to potential errors in diagnosis. In response to this challenge, we present a model that integrates low-pass filtering and continuous wavelet transformation to eliminate baseline wander noise occurring during ECG measurements. Moreover, to account for temporal features, we incorporate Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (Bi-LSTM) to estimate QT intervals. Methods The dataset used in this study was the QT Database, comprising ECG data for 105 patients recorded at a sampling rate of 250Hz over a duration of 15 minutes. Each ECG was associated with annotations for two leads and information regarding the P, QRS, and T waves. We structured the dataset into a ratio of approximately 8:1:1, resulting in 51 patients for training (15,490 samples), 6 patients for validation (1,564 samples), and 7 patients for testing (1,682 samples). Results For the evaluation of deep learning, we selected Accuracy and F1-Score as metrics. We experimented with two configurations: the proposed model with two Bi-LSTM layers and an alternative with a single Bi-LSTM layer of 100 units. We also tested a similar recurrent neural network, GRU, with the same unit configuration. The results indicated that the method with two Bi-LSTM layers outperformed others in both Accuracy and F1-Score. Conclusion By incorporating CNN and Bi-LSTM, we conducted QT interval estimation, leveraging temporal features. This methodology facilitated the elimination of diverse noise types during ECG measurements, resulting in enhanced accuracy for QT interval estimation through the consideration of temporal ECG characteristics. In our future research endeavors, we plan to deploy this model for predicting arrhythmias, specifically targeting conditions like long QT syndrome and drug-induced QT prolongation.Methods and result

Algorithm Improvement for Mobile Event Detection with Intelligent Tunnel Robots

Mobile inspections conducted by intelligent tunnel robots are instrumental in broadening the inspection reach, economizing on inspection expenditures, and augmenting the operational efficiency of inspections. Despite differences from fixed surveillance, mobile-captured traffic videos have complex backgrounds and device conditions that interfere with accurate traffic event identification, warranting more research. This paper proposes an improved algorithm based on YOLOv9 and DeepSORT for intelligent event detection in an edge computing mobile device using an intelligent tunnel robot. The enhancements comprise the integration of the Temporal Shift Module to boost temporal feature recognition and the establishment of logical rules for identifying diverse traffic incidents in mobile video imagery. Experimental results show that our fused algorithm achieves a 93.25% accuracy rate, an improvement of 1.75% over the baseline. The algorithm is also applicable to inspection vehicles, drones, and autonomous vehicles, effectively enhancing the detection of traffic events and improving traffic safety.

Temporal dynamics of nucleus accumbens neurons in male mice during reward seeking

The nucleus accumbens (NAc) regulates reward-motivated behavior, but the temporal dynamics of NAc neurons that enable “free-willed” animals to obtain rewards remain elusive. Here, we recorded Ca2+ activity from individual NAc neurons when mice performed self-paced lever-presses for sucrose. NAc neurons exhibited three temporally-sequenced clusters, defined by times at which they exhibited increased Ca2+ activity: approximately 0, −2.5 or −5 sec relative to the lever-pressing. Dopamine D1 receptor (D1)-expressing neurons and D2-neurons formed the majority of the −5-sec versus −2.5-sec clusters, respectively, while both neuronal subtypes were represented in the 0-sec cluster. We found that pre-press activity patterns of D1- or D2-neurons could predict subsequent lever-presses. Inhibiting D1-neurons at −5 sec or D2-neurons at −2.5 sec, but not at other timepoints, reduced sucrose-motivated lever-pressing. We propose that the time-specific activity of D1- and D2-neurons mediate key temporal features of the NAc through which reward motivation initiates reward-seeking behavior.

A digital twin-assisted intelligent fault diagnosis method for hydraulic systems

As the complexity of modern engineering systems increases, traditional fault detection models face growing challenges in achieving accuracy and reliability. This paper presents a novel Digital Twin-assisted fault diagnosis framework specifically designed for hydraulic systems. The framework utilizes a virtual model, constructed using Modelica, which is integrated with real-time system data through a first-of-its-kind bidirectional data consistency evaluation mechanism. The integrated data is further refined using a two-dimensional signal warping algorithm to enhance its reliability. This optimized twin data is then employed to train a multi-channel one-dimensional convolutional neural network-gated recurrent unit model, effectively capturing both spatial and temporal features to improve fault detection. The subsea blowout preventer in lab is used to study the performance of the method. The results show that the accuracy is 95.62 %. Compared to current methods, this is a significant improvement. By integrating DT technology, data consistency optimization, and advanced deep learning techniques, this framework provides a scalable and reliable solution for predictive maintenance in complex engineering systems.

Based on Deep Learning Model and Flink Streaming Computing Short Term Photovoltaic Power Generation Prediction for Suburban Distribution Network

With the advancement of global energy internet construction, accurate prediction of new energy generation power such as photovoltaic is an important foundation for ensuring the safety and economic working of new power systems. A short-term photovoltaic power generation prediction method for suburban distribution networks based on deep learning model fusion and Flink flow calculation is proposed to address the challenges of complex power grids, diversified disturbance factors, and isolated monitoring points. This method uses Bi directional Long Short Term Memory(BiLSTM) to extract cross sequential nonlinear characteristic of photovoltaic power generation time series data. Compared with standard LSTM, BiLSTM can consider both historical and future information simultaneously, thus extracting richer extracted features from power generation time series data. This method also integrates attention mechanism to capture the importance distribution of historical temporal features for power generation prediction, effectively solving the problem of long-term temporal dependence in standard LSTM models. The Flink streaming computing framework embeds a trained BiLSTM-Attention photovoltaic power generation prediction model, enabling real-time prediction and monitoring analysis of photovoltaic power generation at various monitoring points in the suburban distribution network. This article uses a dataset of a suburban photovoltaic power station for validation, and trains the model with historical power generation data, meteorological factors, weather types, seasons, and other information as inputs. The BiLSTM-Attention fusion model studys the temporal characteristics of power generation, and has high accuracy in predicting short-term photovoltaic power generation in different scenarios. The Flink streaming computing platform can not only process high throughput predicted power data, but also has low time delay.

Graph Neural Network-Based Speech Emotion Recognition: A Fusion of Skip Graph Convolutional Networks and Graph Attention Networks

In speech emotion recognition (SER), our research addresses the critical challenges of capturing and evaluating node information and their complex interrelationships within speech data. We introduce Skip Graph Convolutional and Graph Attention Network (SkipGCNGAT), an innovative model that combines the strengths of skip graph convolutional networks (SkipGCNs) and graph attention networks (GATs) to address these challenges. SkipGCN incorporates skip connections, enhancing the flow of information across the network and mitigating issues such as vanishing gradients, while also facilitating deeper representation learning. Meanwhile, the GAT in the model assigns dynamic attention weights to neighboring nodes, allowing SkipGCNGAT to focus on both the most relevant local and global interactions within the speech data. This enables the model to capture subtle and complex dependencies between speech segments, thus facilitating a more accurate interpretation of emotional content. It overcomes the limitations of previous single-layer graph models, which were unable to effectively represent these intricate relationships across time and in different speech contexts. Additionally, by introducing a pre-pooling SkipGCN combination technique, we further enhance the ability of the model to integrate multi-layer information before pooling, improving its capacity to capture both spatial and temporal features in speech. Furthermore, we rigorously evaluated SkipGCNGAT on the IEMOCAP and MSP-IMPROV datasets, two benchmark datasets in SER. The results demonstrated that SkipGCNGAT consistently achieved state-of-the-art performance. These findings highlight the effectiveness of the proposed model in accurately recognizing emotions in speech, offering valuable insights and a solid foundation for future research on capturing complex relationships within speech signals for emotion recognition.

A Novel Dictionary Learning Algorithm Based on Prior Knowledge for fMRI Data Analysis

ABSTRACTTask‐based functional magnetic resonance imaging (fMRI) has been widely utilized for brain activation detection and functional network analysis. In recent years, the K‐singular value decomposition (K‐SVD) algorithm has gained increasing attention in the research of fMRI data analysis methods. In this study, we propose a novel temporal feature region‐growing constrained K‐SVD algorithm that incorporates task‐based fMRI temporal prior knowledge and utilizes a region‐growing algorithm to infer potential activation locations. The algorithm incorporates temporal and spatial constraints to enhance the detection of brain activation. Specifically, this paper improves the three stages of the traditional K‐SVD algorithm. First, in the dictionary initialization stage, the automatic target generation process with an independent component analysis algorithm is utilized in conjunction with prior knowledge to enhance the accuracy of initialization. Second, in the sparse coding stage, the region‐growing algorithm is employed to infer potential activation locations based on temporal prior knowledge, thereby imposing spatial constraints to limit the extent of activation regions. Finally, in the dictionary learning stage, soft constraints and low correlation constraints are applied to reinforce the consistency with prior knowledge and enhance the robustness of learning for task‐related atoms. The proposed method was validated on simulated and real fMRI data, showing superior performance in detecting brain activation compared with traditional methods. The results indicate that the algorithm accurately identifies activated brain regions, providing an effective approach for studying brain function in clinical applications.

Feasibility Analysis for Predicting Indian Ocean Bigeye Tuna (Thunnus obesus) Fishing Grounds Based on Temporal Characteristics of FY-3 Microwave Radiation Imager Data

Efficient and accurate fishery forecasting is of great significance in ensuring the efficiency of fishery operations. This paper proposes a fishery forecasting method using a brightness temperature (TB) time series spatial feature extraction and fusion model. Using Indian Ocean bigeye tuna fishery data from 2009 to 2021 as a reference, this paper discusses the feasibility of fishery forecasting using FY-3 Microwave Radiation Imager (MWRI) Level 1 TB data. For this paper, we designed a deep learning network model for radiometer TB time series feature extraction (TimeTB-FishNet) based on the Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU), and Attention mechanism. After expanding the dimensions of TB features, the model uses them together with spatiotemporal feature factors (year, month, longitude, and latitude) as features. By adding the GRU and Attention to the CNN, the CNN-GRU-Attention model architecture is established and can extract deep time series spatial features from the data to achieve the best results. In the model validation experiments, the TimeTB-FishNet model performed optimally, with a coefficient of determination (R2) of 0.6643. In the generalization experiments, the R2 also reached 0.6261, with a root mean square error (RMSE) of 46.6031 kg/1000 hook. When the sea surface height (SSH) was introduced, the R2 further reached 0.6463, with a lower RMSE of 45.1318 kg/1000 hook. The experimental results show that the proposed method and model are feasible and effective. The proposed model can directly use enhanced radiometer TB data without relying on lagging ocean environmental product data, performing deep temporal and spatial feature extraction for fishery forecasting. This method can provide a reference for the fishing of bigeye tuna in the Indian Ocean.

Enhancing Daily Crash Count Prediction using Deep Learning with Window Size Selection and Seasonality Predictor Integration

The escalating need for proactive safety measures, coupled with advancements in data collection and analytical techniques, has significantly refined the accuracy of crash count predictions, shifting from annual scales to finer daily or hourly estimates. This research places emphasis on recurrent neural networks (RNNs), specifically the long short-term memory (LSTM) model, acknowledged for effectively managing sequential data in time series predictions. Paramount considerations encompass the treatment of input data, including the decision to incorporate temporal features alongside endogenous historical target values, and the establishment of an optimal window size for data input. Despite the critical nature of these facets, exhaustive studies concurrently investigating both under controlled conditions are scarce. This research addresses this gap, assessing diverse scenarios featuring distinct temporal treatments and window sizes, and employing an LSTM model with uniform fine-tuned parameters to ensure a fair comparison. Findings indicate a significant variation in performance among models employing different window sizes and month predictor integration, under identical RNN structures and LSTM configurations. The best model outperformed the least effective by roughly 30% concerning root mean square error (RMSE) and ranking correlation between predicted and actual target crash counts in the test datasets. Unlike seasonality predictor treatment, diverse window size selections did not lead to statistically significant differences in model performance. Generally, a window size of 3 performed worst under most conditions, followed by a size of 14. Sizes of 7 and 28 performed best in nearly an equal numbers of cases.

Detecting command injection attacks in web applications based on novel deep learning methods

Web command injection attacks pose significant security threats to web applications, leading to potential server information leakage or severe server disruption. Traditional detection methods struggle with the increasing complexity and obfuscation of these attacks, resulting in poor identification of malicious code, complicated feature extraction processes, and low detection efficiency. To address these challenges, a novel detection model, the Convolutional Channel-BiLSTM Attention (CCBA) model, is proposed, leveraging deep learning techniques to enhance the identification of web command injection attacks. The model utilizes dual CNN convolutional channels for comprehensive feature extraction and employs a BiLSTM network for bidirectional recognition of temporal features. An attention mechanism is also incorporated to assign weights to critical features, improving the model’s detection performance. Experimental results demonstrate that the CCBA model achieves 99.3% accuracy and 98.2% recall on a real-world dataset. To validate the robustness and generalization of the model, tests were conducted on two widely recognized public cybersecurity datasets, consistently achieving over 98% accuracy. Compared to existing methods, the proposed model offers a more effective solution for identifying web command injection attacks.

Visual field prediction based on temporal-spatial feature learning

Glaucoma stands as the leading irreversible cause of blindness worldwide. Regular visual field examinations play a crucial role in both diagnosing and treating glaucoma. Predicting future visual field changes can assist clinicians in making timely interventions to manage the progression of this disease. To integrate temporal and spatial features from past visual field examination results and enhance visual field prediction, a convolutional long short-term memory (ConvLSTM) network was employed to construct a predictive model. The predictive performance of the ConvLSTM model was validated and compared with other methods using a dataset of perimetry tests from the Humphrey field analyzer at the University of Washington (UWHVF). Compared to traditional methods, the ConvLSTM model demonstrated higher prediction accuracy. Additionally, the relationship between visual field series length and prediction performance was investigated. In predicting the visual field using the previous three visual field results of past 1.5~6.0 years, it was found that the ConvLSTM model performed better, achieving a mean absolute error of 2.255 dB, a root mean squared error of 3.457 dB, and a coefficient of determination of 0.960. The experimental results show that the proposed method effectively utilizes existing visual field examination results to achieve more accurate visual field prediction for the next 0.5~2.0 years. This approach holds promise in assisting clinicians in diagnosing and treating visual field progression in glaucoma patients.

Fusion Attention for Action Recognition: Integrating Sparse-Dense and Global Attention for Video Action Recognition.

Conventional approaches to video action recognition perform global attention over the entire video patches, which may be ineffective due to the temporal redundancy of video frames. Recent works on masked video modeling adopt a high-ratio tube masking and reconstruction strategy as a pre-training method to mitigate the problem of focusing on spatial features well but not on temporal features. Inspired by this pre-training method, we propose Fusion Attention for Action Recognition (FAR), which fuses the sparse-dense attention patterns specialized for temporal features with global attention during fine-tuning. FAR has three main components: head-split sparse-dense attention (HSDA), token-group interaction, and group-averaged classifier. First, HSDA splits the head of multi-head self-attention to fuse global and sparse-dense attention. The sparse-dense attention is divided into groups of tube-shaped patches to focus on temporal features. Second, token-group interaction is used to improve information exchange between divided patch groups. Finally, the group-averaged classifier uses spatio-temporal features from different patch groups to improve performance. The proposed method uses the weight parameters that are pre-trained with VideoMAE and MVD, and achieves higher performance (+0.1-0.4%) with less computation than models fine-tuned with global attention on Something-Something V2 and Kinetics-400. Moreover, qualitative comparisons show that FAR captures temporal features quite well in highly redundant video frames. The FAR approach demonstrates improved action recognition with efficient computation, and exploring its adaptability across different pre-training methods presents an interesting direction for future research.

Unsupervised selection of acoustic indices: An experimental comparison for characterizing unlabeled audio recordings from sub-Andean forest soundscapes

We present an experimental comparison of applying six unsupervised (i.e. not relying on class labels) and almost parameterless feature selection methods for ranking acoustic indices. The study is aimed at guiding the practitioner in choosing appropriate acoustic indices when, in absence of class labels, a small but meaningful number of features to characterize soundscapes is desired. Forty acoustic indices were considered, which correspond to seventeen temporal, spectral and soundscape features, along with their basic statistics. Three publicly available soundscape datasets, registered in a sub-Andean forest, were used for the experiments; moreover, several subsets were considered according to the different times of the day. Results reveal that the Acoustic Evenness Index is the most important one in terms of representational power according to the six considered selection criteria, followed by the Acoustic Complexity Index when conditions are relaxed to examine features ranked among the top-five. Besides, the Bioacoustic Index, the Acoustic Diversity Index and the Root Mean Square Energy stand out as important features when characterizing days in their separated parts.

An attention fused sequence -to-sequence convolutional neural network for accurate solar irradiance forecasting and prediction using sky images

The challenge of accurately forecasting ultra-short-term solar irradiance for photovoltaic systems is complicated by rapidly changing weather, and while ground-based sky images offer potential improvements, effectively extracting spatiotemporal data from these images remains a significant hurdle for current computer vision models. A new hybrid model, "An Attention Fused Sequence-to-Sequence Convolutional Neural Network," is being developed to address this challenge. The model predicts intra-hour GHI, DNI, and DHI with a 10-min lead time by combining a Convolutional Neural Network (for spatial feature extraction from sky images), an attention mechanism (to focus on relevant regions), and a sequence-to-sequence model (for temporal feature extraction from time-series data). The proposed Model is trained using the NREL Solar Radiation Research Laboratory Dataset while evaluating the model with the Mean Bias Error, Mean Absolute Error, Root Mean Squared Error, R Squared, and forecasting skill score. The 10-min and sequence length 2 interval is considered to be the best performing across most of the evaluation metrics with an MBE value is 2.321 W/m2, MAE value of 39.490 W/m2, RMSE value of 62.086 W/m2, R2 value is 0.909 W/m2, FSS value of 23.589 W/m2 and an MBE of 4.876 W/m2, MAE of 56.887 W/m2, RMSE of 85.346 W/m2, R2 of 0.834 and FSS of 24.881 W/m2 respectively. Furthermore, the sensitivity analysis reveals that the proposed model's performance is influenced by both the sequence length and the lead time. The proposed framework outperforms other techniques for ultra-short-term PV generation forecasting, demonstrating its potential for practical deployment in PV systems to improve grid reliability and energy management.