Saroj Biswas

Prof. Saroj Biswas

orcid

Research Focus & Metrics

Academic Career Length

16Years (2011-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
25 November 2021 in IEEE Transactions on Industry Applications
Verified
Journal Article
Closed access
SZSaeedeh Ziyabari
Saeedeh Ziyabari
Liang DuLiang Du
Liang Du
SBSaroj Biswas
Saroj Biswas
This work introduces a novel multi-branch attentive gated recurrent residual network (ResAttGRU) consisting of multiple branches of residual networks (ResNet), gated recurrent units (GRUs), and the attention mechanism. The proposed multi-branch ResAttGRU is capable of modeling data at various resolutions, extracting hierarchical features, and capturing short- and long-term dependencies. Moreover, this network also presents a strong multi-time-scale representative within the proposed architecture, while GRUs can exploit temporal information at less computational cost than the popular long-short-term memory (LSTM). The novelty of the proposed architecture is to employ multiple convolutional-based branches with different filter lengths to learn multi-time-scale features jointly, accelerate the learning process, and reduce overfitting by leveraging shared representations as the auxiliary information. This study also compares the multi-branch ResAttGRU networks with state-of-the-art deep learning methods using 18 years of NSRDB data at 12 solar sites. Finally, the proposed multi-branch ResAttGRU requires 7.1% fewer parameters than multi-branch residual LSTM (ResLSTM) while achieving similar average RMSE, MAE, and R-squared values.

16 August 2023 in IEEE Transactions on Industry Applications
Verified
Journal Article
Closed access
SZSaeedeh Ziyabari
Saeedeh Ziyabari
ZZZhenyu Zhao
Zhenyu Zhao
Liang DuLiang Du
Liang Du
SBSaroj Biswas
Saroj Biswas
The increasing penetration of solar generation into power grids has promoted the need for accurate and reliable short-term solar irradiance forecasting. Existing methods utilizing advanced deep learning architectures have shown advanced performance compared to conventional time-series analytical techniques but in general encountered shortcomings in modeling spatial correlations among neighboring solar generation sites, exploring the similarity of long-term, time-varying patterns, and alleviating overfitting issues in convolutional and recurrent neural networks, such as the popular Long Short-term Memory (LSTM). To effectively but yet reliably tackle these challenges in the existing literature, this article proposes a spatio-temporal framework consisting of a multi-branch hybrid Residual network and the Transformer architecture (ResTrans). The proposed framework has been tested on two groups' real-world data containing 17 years-long data from different solar sites in Philadelphia, USA, including 12 and 18 locations, respectively. Compared to other hybrid benchmark architectures, including single-branch ResTrans and multi-branch ResNet-LSTM (ResLSTM), single-branch ResLSTM, and CNN-LSTM, the proposed multi-branch ResTrans achieves the highest forecasting accuracy with an average RMSE of 0.049 $(W/m^{2})$ , an average MAE of 0.031 $(W/m^{2})$ , and an $R^{2}$ coefficient of 97%.

6 February 2025 in Energies
Verified
Journal Article
Open access
PDF file
SBSaroj Biswas
Saroj Biswas
JCJim Shih-Jiun Chen
Jim Shih-Jiun Chen
The Betz constant is the well-known aerodynamic limit of the maximum power which can be extracted from wind using wind turbine technologies, under the assumption that the wind speed is uniform across a blade disk. However, this condition may not hold for large wind turbines, since the wind speed may not be constant along their height; rather, it may vary with the location due to surface friction from tall buildings and trees, the topography of the Earth’s surface, and radiative heating and cooling in a 24 h cycle. This paper derives a new power coefficient for large wind turbines based on the power law exponent model of the wind gradient and height. The proposed power coefficient is a function of the size of the rotor disk and the Hellmann exponent, which describes the wind gradient based on wind stability at various locations, and it approaches the same value as the Betz limit for wind turbines with small rotor disks. It is shown that for large offshore wind turbines, the power coefficient was about 1.27% smaller than that predicted by the Betz limit, whereas for onshore turbines in human-inhabited areas with stable air, the power coefficient was about 8.7% larger. Our results are significant in two ways. First, we achieve generalization of the well-known Betz limit through elimination of the assumption of a constant wind speed across the blade disk, which does not hold for large wind turbines. Second, since the power coefficient depends on the location and air stability, this study offers guidelines for wind power companies regarding site selection for the installation of new wind turbines, potentially achieving greater energy efficiency than that predicted by the Betz limit.

13 July 2025 in Journal of Imaging
Verified
Journal Article
Open access
PDF file
VKVahid Khalkhali
Vahid Khalkhali
HLHayan Lee
Hayan Lee
JNJoseph Nguyen
Joseph Nguyen
Sergio Zamora-ErazoSergio Zamora-Erazo
Sergio Zamora-Erazo
CRCamille Ragin
Camille Ragin
SBSaroj Biswas
Saroj Biswas
The absence of skin color information in skin cancer datasets poses a significant challenge for accurate diagnosis using artificial intelligence models, particularly for non-white populations. In this paper, based on the Monk Skin Tone (MST) scale, which is less biased than the Fitzpatrick scale, we propose MST-AI, a novel method for detecting skin color in images of large datasets, such as the International Skin Imaging Collaboration (ISIC) archive. The approach includes automatic frame, lesion removal, and lesion segmentation using convolutional neural networks, and modeling normal skin tones with a Variational Bayesian Gaussian Mixture Model (VB-GMM). The distribution of skin color predictions was compared with MST scale probability distribution functions (PDFs) using the Kullback-Leibler Divergence (KLD) metric. Validation against manual annotations and comparison with K-means clustering of image and skin mean RGBs demonstrated the superior performance of the MST-AI, with Kendall’s Tau, Spearman’s Rho, and Normalized Discounted Cumulative Gain (NDGC) of 0.68, 0.69, and 1.00, respectively. This research lays the groundwork for developing unbiased AI models for early skin cancer diagnosis by addressing skin color imbalances in large datasets.

22 June 2016 in arXiv
Verified
Preprint
Open access
PDF file
ASAnibal Sanjab
Anibal Sanjab
Walid SaadWalid Saad
Walid Saad
IGIsmail Guvenc
Ismail Guvenc
Arif I. SarwatArif I. Sarwat
Arif I. Sarwat
SBSaroj Biswas
Saroj Biswas
The cyber-physical nature of the smart grid has rendered it vulnerable to a multitude of attacks that can occur at its communication, networking, and physical entry points. Such cyber-physical attacks can have detrimental effects on the operation of the grid as exemplified by the recent attack which caused a blackout of the Ukranian power grid. Thus, to properly secure the smart grid, it is of utmost importance to: a) understand its underlying vulnerabilities and associated threats, b) quantify their effects, and c) devise appropriate security solutions. In this paper, the key threats targeting the smart grid are first exposed while assessing their effects on the operation and stability of the grid. Then, the challenges involved in understanding these attacks and devising defense strategies against them are identified. Potential solution approaches that can help mitigate these threats are then discussed. Last, a number of mathematical tools that can help in analyzing and implementing security solutions are introduced. As such, this paper will provide the first comprehensive overview on smart grid security.

1 March 2011 in Naval Engineers Journal
Verified
Journal Article
Closed access
FFFrank Ferrese
Frank Ferrese
QDQing Dong
Qing Dong
NNNat Nataraj
Nat Nataraj
SBSaroj Biswas
Saroj Biswas
The US Navy has a continuing interest and investment in basic and applied research in the area of automation and control. The potential naval applications for this research are numerous and wide ranging. The need for advances in control and automation systems exists from missile defense, to shipboard auxiliary systems, to naval aircraft, and virtually everywhere in between. This research is performed in industry, academia, and in naval laboratories across the nation. This paper will detail particular research in control theory being performed in the area of automation and controls in the naval laboratories. A particle swarm optimization algorithm is used to manipulate the state and control weighting matrices of a linear quadratic regulator to achieve an optimal control for a desired eigenstructure. The algorithm is demonstrated on a nonlinear power system model, and is found to be highly effective in the stabilization of the system output performance, showing both rapid convergence and a closed loop eigenstructure very close to the specified eigenstructure.