Faculty Publications
Permanent URI for this communityhttps://idr.nitk.ac.in/handle/123456789/18736
Publications by NITK Faculty
Browse
5 results
Search Results
Item Enhancing High-Frequency PV Power Forecast Using Optimal Hyperparameter Setting in LSTM(Springer Science and Business Media Deutschland GmbH, 2025) Kumar, A.; Kashyap, Y.; Nasar, R.Solar energy plays a significant role in the world’s shift to renewable and sustainable energy. So, accurate forecasting techniques are essential for effective grid management and smooth integration into current energy infrastructures. Traditional solar forecasting approaches often encounter limitations in capturing the complex and nonlinear relationships inherent in solar power generation patterns. In response to these challenges, the present paper demonstrates the forecast analysis of high-frequency (HF) PV power components, which is obtained with the decomposition of actual PV power data. The focus of this paper is on the analysis of high-frequency PV power components as they exhibit high fluctuation. To capture this high fluctuation feature present in PV power, a moving average filter is applied to smooth the input data and potentially enhance the 60 min ahead forecasting performance using the long short-term memory (LSTM) model. The best-performing LSTM model has secured MAE= 1.114 % and RMSE = 2.608 % for 60 min ahead forecast. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.Item LBF-NoC: Learning-Based Framework to Predict Performance, Power and Area for Network-On-Chip Architectures(World Scientific, 2022) Kumar, A.; Talawar, B.Extensive large-scale data and applications have increasing requests for high-performance computations which is fulfilled by Chip Multiprocessors (CMP) and System-on-Chips (SoCs). Network-on-Chips (NoCs) emerged as the reliable on-chip communication framework for CMPs and SoCs. NoC architectures are evaluated based on design parameters such as latency, area, and power. Cycle-accurate simulators are used to perform the design space exploration of NoC architectures. Cycle-accurate simulators become slow for interactive usage as the NoC topology size increases. To overcome these limitations, we employ a Machine Learning (ML) approach to predict the NoC simulation results within a short span of time. LBF-NoC: Learning-based framework is proposed to predict performance, power and area for Direct and Indirect NoC architectures. This provides chip designers with an efficient way to analyze various NoC features. LBF-NoC is modeled using distinct ML regression algorithms to predict overall performance of NoCs considering different synthetic traffic patterns. The performance metrics of five different (Mesh, Torus, Cmesh, Fat-Tree and Flattened Butterfly) NoC architectures can be analyzed using the proposed LBF-NoC framework. BookSim simulator is employed to validate the results. Various architecture sizes from 2×2 to 45×45 are used in the experiments considering various virtual channels, traffic patterns, and injection rates. The prediction error of LBF-NoC is 6% to 8%, and the overall speedup is 5000× to 5500× with respect to BookSim simulator. © 2022 World Scientific Publishing Company.Item Utilizing support vector regression modeling to predict pyro product yields from microwave-assisted catalytic co-pyrolysis of biomass and waste plastics(Elsevier Ltd, 2023) Ramesh, P.; Sankar Rao, C.S.; Surya, D.V.; Kumar, A.; Basak, T.The rise in plastic waste production has led to the development of co-pyrolysis of waste plastics and biomass as a potential solution. This process converts waste into valuable resources, including chemicals and pollutant-absorbing materials. Accurately predicting product yields is crucial and involves considering feedstock characteristics and pyrolysis conditions. No previous work on machine learning (ML) predicts pyro-products considering catalyst and blend as input features. This study used a support vector machine (SVM) to predict pyro-product yields from microwave-assisted co-pyrolysis of biomass and plastics. SVM models were trained, validated, and then applied to new data. The results showed high predictive accuracy, with R2 values of 0.96, 0.93, and 0.91 for bio-oil, biochar, and biogas, respectively. The SVM model demonstrated strong predictive capabilities, indicating effective generalization ability based on statistical parameters. Additionally, SVM models incorporating all features performed better than those based on 'elementary analysis (EA)' and 'proximate analysis (PA)' alone. The pearson correlation coefficient (PCC) approach assessed the correlation between input features to remove highly correlated variables. The partial dependence analysis reveals the individual effects of influential factors and their interactions in the co-pyrolysis process, highlighting significant features like carbon, hydrogen, ash, volatile matter, and nitrogen content that influence oil, char, and gas yields, thereby providing valuable insights for optimization strategies in co-pyrolysis. © 2023 Elsevier LtdItem MSSEAG-UNet: A Novel Deep Learning Architecture for Cloud Segmentation in Fisheye Sky Images and Solar Energy Forecast(Institute of Electrical and Electronics Engineers Inc., 2025) Kumar, A.; Kashyap, Y.; Sharma, K.; Vittal, K.; Shubhanga, K.N.This study analyzes sky images captured using a ground-based fisheye camera, aiming to address the challenge of accurately segmenting clouds, which is difficult due to their fuzzy and indistinct boundaries and uneven lighting conditions. Accurate segmentation of clouds in ground-based sky images is crucial for accurate solar energy forecasting. Motivated by these challenges, this article has proposed a novel deep learning architecture called multispatial squeeze-and-excite attention gated U-Net (MSSEAG-UNet) for cloud segmentation in ground-based fisheye sky images. The proposed architecture integrates a multispatial convolutional (MS-CNN) block and squeeze-and-excitation (SE) blocks in the encoder path to improve multiscale feature extraction (MFF) and recalibrate feature maps, while an attention block is incorporated in the decoder path to emphasize key cloud features. The segmentation performance of the MSSEAG-UNet is compared with five benchmark models, and results show that the proposed model outperforms than all benchmarks models. Furthermore, the segmented cloud images produced by the MSSEAG-UNet are used to calculate the cloud percentage, which is then integrated with the original sky images using a multicolumn convolutional model for global horizontal irradiance (GHI) forecast. GHI forecast is conducted for 15-, 30-, and 60-min ahead timesteps, with the best results achieved for the 60-min forecast, yielding mean absolute error (MAE), mean square error (mse), and RMSE values of 6.245%, 0.683%, and 8.265%, respectively. These results highlight the effectiveness of the proposed approach in improving both cloud segmentation accuracy and short-term solar irradiance forecasting. © 1980-2012 IEEE.Item An uncertainty-aware decision support system: Integrating text narratives and conformal prediction for trustworthy accident code classification(Institution of Chemical Engineers, 2025) Kumar, A.; Senapati, A.; Upadhyay, R.; Chatterjee, S.; Bhattacherjee, A.; Samanta, B.It is imperative to assign accident classification codes to the Mine Safety and Health Administration (MSHA) accident data for effective data analysis and risk assessment. Although trained personnel are capable of performing this task, the manual process is both time-consuming and resource-intensive. Automating the classification process with machine learning (ML) algorithms promises to expedite code assignment. However, ML predictions typically lack uncertainty metrics. This study proposes an uncertainty-aware hierarchical classification framework that assists human experts in efficiently and accurately assigning accident codes. Several text representation techniques combined with different ML algorithms were employed within a hierarchical architecture to assign classification codes. Low-frequency codes were consolidated into a single category, with a primary classifier distinguishing between these and a secondary classifier further classifying the grouped categories. Regularized Adaptive Prediction Sets (RAPS) was integrated to quantify uncertainty. Highly confident predictions yielding single-class sets were automatically classified, whereas multi-class sets were flagged for manual review. Primary Classifier with XGBoost with word2vec text representation achieved the best performance, with 95.12 % coverage, 37.02 % single-class prediction sets at 96.11 % accuracy, and an average prediction set size of 2.39. Whereas the secondary classifier, a logistic regression model with TF-IDF representation, yielded 96.19 % coverage, an average set size of 1.80, and 53.66 % single-class prediction sets with 98.90 % accuracy. Additionally, sensitivity analysis determined that a 95 % coverage guarantee offers the best trade-off between prediction set size and coverage. The framework effectively integrates conformal prediction to quantify uncertainty and aid human experts in improving the decision-making process in safety management. Although the framework is broadly applicable across different sectors, it needs to be retrained on domain-specific data for effective use. © 2025 The Institution of Chemical Engineers