Takele Ferede Agajie
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Takele Ferede Agajie has a total of 182 co-authors. Here are the top co-authors by number of shared publications:
- Abdelaziz Rabehi
Telecommunication and Smart Systems Laboratory, Faculty of Sciences and Technology, Ziane Achour University, Djelfa 17000, Algeria
8 shared publications - Isaac Amoussou
Department of Electrical and Electronic Engineering, Faculty of Engineering and Technology, University of Buéa, Buéa, Cameroon
8 shared publications - Ahmed Farag Ali
Department of Electrical and Electronic Engineering Technology, University of Johannesburg, Johannesburg 2006, South Africa
7 shared publications - Baseem Khan
Department of Electrical and Computer Engineering, Hawassa University, Hawassa, Ethiopia
7 shared publications - Emmanuel Tanyi
Department of Electrical and Electronic Engineering, Faculty of Engineering and Technology, University of Buea, Buea P.O. Box 63, Cameroon
7 shared publications - Armand Fopah-Lele
Department of Mechanical and Industrial Engineering, Faculty of Engineering and Technology, University of Buea, Buea P.O. Box 63, Cameroon
5 shared publications
Latest Publications
- Open Access
- 1 November 2026
- Published by Mustansiriyah University College of Engineering in Journal of Engineering and Sustainable Development
- Vol. 30 (6) , 835-852
Night Jasmine Blooming Optimization (NJBO): A Novel Nature-Inspired Algorithm for Electrical Engineering Applications
Optimization algorithms have become vital to the solution of complex optimization problems that are difficult to solve using conventional methods. In this paper, we introduce a novel nature-inspired metaheuristic algorithm called the Night Jasmine Blooming Optimization (NJBO) algorithm, which is inspired by the blooming of the night jasmine plant (Cestrum nocturnum). The suggested algorithm combines adaptive hormonal regulation, Gibberellin-based disturbance, scent-guided exploitation, elite preservation, and Nelder–Mead local refinement, which allows us to have a balanced exploration–exploitation process. It was tested with thirteen classical benchmark functions and the CEC2017 benchmark suites and benchmarked with several well-known and recent metaheuristic methods. As a practical application, NJBO was used for identification of an IIR system and optimal design of a five-stage Cockcroft–Walton voltage multiplier circuit. Additionally, the scalability analysis, runtime evaluation, component contribution analysis, parameter sensitivity analysis, and statistical validation with the Friedman ranking test and Wilcoxon signed-rank test were performed to assess the system completely. The results show that NJBO is a very effective general-purpose optimization algorithm with competitive convergence performance, good optimization accuracy, and stability for the benchmark and engineering optimization problems that were investigated.
- 1 November 2026
- Published by Mustansiriyah University College of Engineering in Journal of Engineering and Sustainable Development
- Vol. 30 (6) , 835-852
Open AccessNight Jasmine Blooming Optimization (NJBO): A Novel Nature-Inspired Algorithm for Electrical Engineering Applications
Optimization algorithms have become vital to the solution of complex optimization problems that are difficult to solve using conventional methods. In this paper, we introduce a novel nature-inspired metaheuristic algorithm called the Night Jasmine Blooming Optimization (NJBO) algorithm, which is inspired by the blooming of the night jasmine plant (Cestrum nocturnum). The suggested algorithm combines adaptive hormonal regulation, Gibberellin-based disturbance, scent-guided exploitation, elite preservation, and Nelder–Mead local refinement, which allows us to have a balanced exploration–exploitation process. It was tested with thirteen classical benchmark functions and the CEC2017 benchmark suites and benchmarked with several well-known and recent metaheuristic methods. As a practical application, NJBO was used for identification of an IIR system and optimal design of a five-stage Cockcroft–Walton voltage multiplier circuit. Additionally, the scalability analysis, runtime evaluation, component contribution analysis, parameter sensitivity analysis, and statistical validation with the Friedman ranking test and Wilcoxon signed-rank test were performed to assess the system completely. The results show that NJBO is a very effective general-purpose optimization algorithm with competitive convergence performance, good optimization accuracy, and stability for the benchmark and engineering optimization problems that were investigated.
- Open Access
- 10 September 2026
- Published by SAGE Publications in Energy Exploration & Exploitation
Enhancing photovoltaic system performance using an advanced hybrid MPPT strategy: a performance and efficiency assessment under variable environmental conditions
Photovoltaic (PV) power generation is highly sensitive to variations in solar irradiance and temperature, making reliable maximum power point tracking (MPPT) essential for maximizing energy extraction under dynamic operating conditions. In this study, we developed and systematically evaluated a hybrid artificial neural network–genetic algorithm (ANN–GA) MPPT strategy that combines nonlinear MPP prediction with global optimization. The proposed controller was evaluated using an MSX-60 PV module model under irradiance levels ranging from 300 to 1000 W/m 2 and temperatures ranging from 291–323 K. The proposed method was benchmarked under identical operating conditions against seven other methods using operating ranges, root mean square error (RMSE), mean absolute error (MAE), R 2 , tracking efficiency, and control-loop execution time as performance metrics. ANN–GA achieved the highest maximum-power range (18.64–54.22 W) and current range (1.39–3.55 A), while yielding RMSE values of 5.39 W, 1.21 V, and 0.28 A for P mp , V mp , and I mp , respectively. Compared with the standalone ANN, these errors were reduced by approximately 11.9%, 10.4%, and 12.5%, respectively, confirming the added value of GA-based optimization. The corresponding R 2 values reached 0.894, 0.893, and 0.893, while the tracking efficiencies were 87.4%, 88.6%, and 94.3%, respectively. The repeated-run analysis further showed that ANN–GA maintained lower RMSE values and narrower error distributions than GA–P&O (perturb and observe), indicating improved repeatability and robustness across independent runs and different PV module configurations. In addition, ANN–GA required approximately 88% less control-loop execution time than GA–P&O, the second-best hybrid method, with reductions of 88.19% for the Solarex MSX-60 module and 88.17% for the Kyocera KC200GT module. Overall, the results show that the proposed hybrid strategy provides a favorable balance among tracking accuracy, repeatability, and computational efficiency under the investigated simulation conditions.
- 10 September 2026
- Published by SAGE Publications in Energy Exploration & Exploitation
Open AccessEnhancing photovoltaic system performance using an advanced hybrid MPPT strategy: a performance and efficiency assessment under variable environmental conditions
Photovoltaic (PV) power generation is highly sensitive to variations in solar irradiance and temperature, making reliable maximum power point tracking (MPPT) essential for maximizing energy extraction under dynamic operating conditions. In this study, we developed and systematically evaluated a hybrid artificial neural network–genetic algorithm (ANN–GA) MPPT strategy that combines nonlinear MPP prediction with global optimization. The proposed controller was evaluated using an MSX-60 PV module model under irradiance levels ranging from 300 to 1000 W/m 2 and temperatures ranging from 291–323 K. The proposed method was benchmarked under identical operating conditions against seven other methods using operating ranges, root mean square error (RMSE), mean absolute error (MAE), R 2 , tracking efficiency, and control-loop execution time as performance metrics. ANN–GA achieved the highest maximum-power range (18.64–54.22 W) and current range (1.39–3.55 A), while yielding RMSE values of 5.39 W, 1.21 V, and 0.28 A for P mp , V mp , and I mp , respectively. Compared with the standalone ANN, these errors were reduced by approximately 11.9%, 10.4%, and 12.5%, respectively, confirming the added value of GA-based optimization. The corresponding R 2 values reached 0.894, 0.893, and 0.893, while the tracking efficiencies were 87.4%, 88.6%, and 94.3%, respectively. The repeated-run analysis further showed that ANN–GA maintained lower RMSE values and narrower error distributions than GA–P&O (perturb and observe), indicating improved repeatability and robustness across independent runs and different PV module configurations. In addition, ANN–GA required approximately 88% less control-loop execution time than GA–P&O, the second-best hybrid method, with reductions of 88.19% for the Solarex MSX-60 module and 88.17% for the Kyocera KC200GT module. Overall, the results show that the proposed hybrid strategy provides a favorable balance among tracking accuracy, repeatability, and computational efficiency under the investigated simulation conditions.
- Open Access
- 30 August 2026
- Published by Springer Nature in Discover Computing
- Vol. 29 (1) , 584
Microcontroller-based internet of things control system for autonomous hidden camera detection using adaptive decision-making and hybrid MCDM techniques
- 30 August 2026
- Published by Springer Nature in Discover Computing
- Vol. 29 (1) , 584
Open AccessMicrocontroller-based internet of things control system for autonomous hidden camera detection using adaptive decision-making and hybrid MCDM techniques
- Open Access
- 20 August 2026
- Published by SAGE Publications in Energy Exploration & Exploitation
Intelligent fault diagnosis and simulation-based modelling of grid-connected photovoltaic systems under desert dust conditions
Photovoltaic (PV) power plants operating in desert environments experience continuous efficiency losses because dust accumulation gradually reduces solar radiation reaching the module surface, leading to lower energy production even under favourable weather conditions. Accurate prediction of normal operating behaviour therefore provides a reference for distinguishing genuine faults from natural fluctuations in plant performance. This study proposes a data-driven framework for PV power prediction and residual-based fault diagnosis at the Aoulef PV power plant in southern Algeria. Artificial neural networks (ANN) and adaptive neuro-fuzzy inference systems (ANFIS) were developed from measured solar irradiance and ambient temperature to estimate the healthy-state PV power output. Model performance was assessed through regression analysis and statistical error indices, whereas fault detection relied on residuals computed from the difference between measured and predicted power. Both models achieved excellent prediction accuracy, with coefficients of determination approaching 0.99. The ANN produced lower prediction errors than the ANFIS, with a root mean square error of 0.009 and an mean absolute error of 0.004, which improved the sensitivity of the residual-based diagnosis. Dust accumulation, identified as fault F13, generated clear residual deviations that enabled automatic fault detection without interrupting plant operation. The findings indicate that the ANN framework combines high predictive accuracy with low computational demand, offering a practical and reliable solution for intelligent monitoring and maintenance of PV systems operating under harsh desert conditions.
- 20 August 2026
- Published by SAGE Publications in Energy Exploration & Exploitation
Open AccessIntelligent fault diagnosis and simulation-based modelling of grid-connected photovoltaic systems under desert dust conditions
Photovoltaic (PV) power plants operating in desert environments experience continuous efficiency losses because dust accumulation gradually reduces solar radiation reaching the module surface, leading to lower energy production even under favourable weather conditions. Accurate prediction of normal operating behaviour therefore provides a reference for distinguishing genuine faults from natural fluctuations in plant performance. This study proposes a data-driven framework for PV power prediction and residual-based fault diagnosis at the Aoulef PV power plant in southern Algeria. Artificial neural networks (ANN) and adaptive neuro-fuzzy inference systems (ANFIS) were developed from measured solar irradiance and ambient temperature to estimate the healthy-state PV power output. Model performance was assessed through regression analysis and statistical error indices, whereas fault detection relied on residuals computed from the difference between measured and predicted power. Both models achieved excellent prediction accuracy, with coefficients of determination approaching 0.99. The ANN produced lower prediction errors than the ANFIS, with a root mean square error of 0.009 and an mean absolute error of 0.004, which improved the sensitivity of the residual-based diagnosis. Dust accumulation, identified as fault F13, generated clear residual deviations that enabled automatic fault detection without interrupting plant operation. The findings indicate that the ANN framework combines high predictive accuracy with low computational demand, offering a practical and reliable solution for intelligent monitoring and maintenance of PV systems operating under harsh desert conditions.
- Open Access
- 1 August 2026
- Published by AIP Publishing in AIP Advances
- Vol. 16 (8)
A multiple sub-objectives experimental based approach for improved model design of photovoltaic systems
One of the key steps to accurately model photovoltaic (PV) systems is the use of metaheuristic algorithms. The technique generally utilizes experimental and estimated current-vs-voltage I(V) curves obtained under standard test conditions (STC) to perform model optimization based on a single objective function. This method has one main drawback: the model derived through this approach usually presents output results that greatly deviate from the actual experimental data when estimating the cells’ performance under climatic conditions different from those of STC. The model using this technique, therefore, becomes unreliable under varying conditions despite its efficiency at STC. This problem has been frequently observed after testing many PV technologies. Thus, in this work, a novel modeling approach based on a weighted-sum multiple sub-objectives optimization method is introduced to address the challenge. The idea is to make the objective function more general, beyond STC, so as to be able to optimize the PV model using many experimental I(V) curves with a wide range of solar irradiance at the same time. This proposed idea has not been reported in the literature to the best of our knowledge. To test the effectiveness of the novel approach, commercial BLD 200-72M monocrystalline and SW175 polycrystalline PV models are employed as a test sample. A comparison is performed between the proposed and literature modeling concepts. The final results proved that the proposed approach has a competitive performance compared to the existing literature method. The proposed approach can achieve an accuracy improvement of up to 61.980%.
- 1 August 2026
- Published by AIP Publishing in AIP Advances
- Vol. 16 (8)
Open AccessA multiple sub-objectives experimental based approach for improved model design of photovoltaic systems
One of the key steps to accurately model photovoltaic (PV) systems is the use of metaheuristic algorithms. The technique generally utilizes experimental and estimated current-vs-voltage I(V) curves obtained under standard test conditions (STC) to perform model optimization based on a single objective function. This method has one main drawback: the model derived through this approach usually presents output results that greatly deviate from the actual experimental data when estimating the cells’ performance under climatic conditions different from those of STC. The model using this technique, therefore, becomes unreliable under varying conditions despite its efficiency at STC. This problem has been frequently observed after testing many PV technologies. Thus, in this work, a novel modeling approach based on a weighted-sum multiple sub-objectives optimization method is introduced to address the challenge. The idea is to make the objective function more general, beyond STC, so as to be able to optimize the PV model using many experimental I(V) curves with a wide range of solar irradiance at the same time. This proposed idea has not been reported in the literature to the best of our knowledge. To test the effectiveness of the novel approach, commercial BLD 200-72M monocrystalline and SW175 polycrystalline PV models are employed as a test sample. A comparison is performed between the proposed and literature modeling concepts. The final results proved that the proposed approach has a competitive performance compared to the existing literature method. The proposed approach can achieve an accuracy improvement of up to 61.980%.
- Open Access
- 1 July 2026
- Published by Mustansiriyah University College of Engineering in Journal of Engineering and Sustainable Development
- Vol. 30 (4) , 469-478
Next-Level 5G Channel Prediction Using Residual Convolutional Neural Network (ResNet)
Accurate channel estimation is critical for achieving the high reliability and low latency required by 5G New Radio (5G NR) systems. Conventional estimators, including Least Squares (LS) and Practical Channel Estimation (PCE), rely on linear models that fail to capture nonlinear channel characteristics. Linear Minimum Mean Square Error (LMMSE) improves accuracy but requires prior channel statistics and a costly matrix inversion, limiting its real-time applicability. This paper proposes a lightweight ResNet-based channel estimator that treats the complex time-frequency channel matrix as a two-dimensional image. Convolutional layers exploit spatial correlations across subcarriers and Orthogonal Frequency Division Multiplexing (OFDM) symbol dimensions, while residual skip connections stabilize training and enable correction over a coarse LS estimate. The model is trained end-to-end using Demodulation Reference Signal (DMRS) pilot patterns in a fully 3GPP-compliant 5G NR environment based on TDL-A and TDL-C channel models (TR 38.901). Simulations over 0–30 dB SNR show consistent superiority over all benchmarks. At SNR = 10 dB, the proposed method achieves a Mean Squared Error (MSE) of 0.0177 — a 45% gain over PCE, ~90% over LS, and 29% over ChannelNet — with only ≈0.29 million parameters, single-stage training, and an inference time of 2–5 ms per OFDM frame, suitable for real-time 5G deployments.
- 1 July 2026
- Published by Mustansiriyah University College of Engineering in Journal of Engineering and Sustainable Development
- Vol. 30 (4) , 469-478
Open AccessNext-Level 5G Channel Prediction Using Residual Convolutional Neural Network (ResNet)
Accurate channel estimation is critical for achieving the high reliability and low latency required by 5G New Radio (5G NR) systems. Conventional estimators, including Least Squares (LS) and Practical Channel Estimation (PCE), rely on linear models that fail to capture nonlinear channel characteristics. Linear Minimum Mean Square Error (LMMSE) improves accuracy but requires prior channel statistics and a costly matrix inversion, limiting its real-time applicability. This paper proposes a lightweight ResNet-based channel estimator that treats the complex time-frequency channel matrix as a two-dimensional image. Convolutional layers exploit spatial correlations across subcarriers and Orthogonal Frequency Division Multiplexing (OFDM) symbol dimensions, while residual skip connections stabilize training and enable correction over a coarse LS estimate. The model is trained end-to-end using Demodulation Reference Signal (DMRS) pilot patterns in a fully 3GPP-compliant 5G NR environment based on TDL-A and TDL-C channel models (TR 38.901). Simulations over 0–30 dB SNR show consistent superiority over all benchmarks. At SNR = 10 dB, the proposed method achieves a Mean Squared Error (MSE) of 0.0177 — a 45% gain over PCE, ~90% over LS, and 29% over ChannelNet — with only ≈0.29 million parameters, single-stage training, and an inference time of 2–5 ms per OFDM frame, suitable for real-time 5G deployments.
- Open Access
- 1 May 2026
- Published by Mustansiriyah University College of Engineering in Journal of Engineering and Sustainable Development
- Vol. 30 (3) , 434-445
Deep Learning-Based Monitoring System to Enhance IoT Network Performance
The rapid growth and increasing complexity of Internet of Things (IoT) networks require efficient real-time monitoring and anomaly detection mechanisms. Traditional machine learning approaches often struggle to handle the dynamic and high-dimensional traffic generated by IoT environments. This study investigates the effectiveness of deep learning models, including Feedforward Neural Networks (FFNN), Convolutional Neural Networks (CNN), and Multilayer Perceptron (MLP), for enhancing IoT network monitoring. The models were trained using both synthetic and real-world IoT traffic datasets in MATLAB with Adam and Stochastic Gradient Descent with Momentum (SGDM) optimizers to improve convergence and training stability. Experimental results demonstrate that deep learning models outperform traditional machine learning techniques in detecting complex traffic patterns and anomalies. Among the evaluated models, CNN achieved the highest accuracy of 94%, compared with Decision Trees (78.5%) and Support Vector Machines (85.7%). CNNs effectively capture spatiotemporal traffic characteristics, while MLPs efficiently model nonlinear relationships in network data. The proposed framework provides a scalable, reliable approach to real-time IoT network monitoring.
- 1 May 2026
- Published by Mustansiriyah University College of Engineering in Journal of Engineering and Sustainable Development
- Vol. 30 (3) , 434-445
Open AccessDeep Learning-Based Monitoring System to Enhance IoT Network Performance
The rapid growth and increasing complexity of Internet of Things (IoT) networks require efficient real-time monitoring and anomaly detection mechanisms. Traditional machine learning approaches often struggle to handle the dynamic and high-dimensional traffic generated by IoT environments. This study investigates the effectiveness of deep learning models, including Feedforward Neural Networks (FFNN), Convolutional Neural Networks (CNN), and Multilayer Perceptron (MLP), for enhancing IoT network monitoring. The models were trained using both synthetic and real-world IoT traffic datasets in MATLAB with Adam and Stochastic Gradient Descent with Momentum (SGDM) optimizers to improve convergence and training stability. Experimental results demonstrate that deep learning models outperform traditional machine learning techniques in detecting complex traffic patterns and anomalies. Among the evaluated models, CNN achieved the highest accuracy of 94%, compared with Decision Trees (78.5%) and Support Vector Machines (85.7%). CNNs effectively capture spatiotemporal traffic characteristics, while MLPs efficiently model nonlinear relationships in network data. The proposed framework provides a scalable, reliable approach to real-time IoT network monitoring.
- RetractionOpen Access
- 3 March 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Retraction Note: Optimization of off-grid hybrid renewable energy systems for cost-effective and reliable power supply in Gaita Selassie Ethiopia
- 3 March 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
RetractionOpen AccessRetraction Note: Optimization of off-grid hybrid renewable energy systems for cost-effective and reliable power supply in Gaita Selassie Ethiopia
- Open Access
- 22 February 2026
- Published by Springer Nature in Journal on Advances in Signal Processing
- Vol. 2026 (1)
IoT-enabled ECG signal preprocessing and CNN-based classification: a novel approach for enhanced healthcare monitoring
- 22 February 2026
- Published by Springer Nature in Journal on Advances in Signal Processing
- Vol. 2026 (1)
Open AccessIoT-enabled ECG signal preprocessing and CNN-based classification: a novel approach for enhanced healthcare monitoring
- Open Access
- 13 January 2026
- Published by Springer Nature in Journal on Wireless Communications and Networking
- Vol. 2026 (1)
Uplink RIS-assisted NOMA/OMA networks: a theoretical study
In this paper, we have investigated the outage probabilities (OPs) and the ergodic rates (ERs) of uplink reconfigurable intelligent surface (RIS)-assisted non-orthogonal and orthogonal multiple-access (NOMA and OMA) networks. Particularly, we have tackled a challenging scenario in which both user equipments (UEs) need assistance from the same RIS unit to communicate with the base station (BS). We have modeled all the UEs $$\rightarrow$$ RIS and RIS $$\rightarrow$$ BS links as Nakagami- m variates. To assess the performance of these networks, we have derived the necessary statistics for the effective channel gain of the reflected (UE $$\rightarrow$$ RIS $$\rightarrow$$ BS) link for the randomly phase-shifted UE, while adopting its already-existing counterpart statistics for the coherently phase-shifted one. On the basis of that, we have derived closed-form expressions for the OPs and the ERs for both RIS-assisted NOMA and OMA schemes. Finally, to get further insights on these results, we have elaborated on the high signal-to-noise ratio (SNR) regime.- 13 January 2026
- Published by Springer Nature in Journal on Wireless Communications and Networking
- Vol. 2026 (1)
Open AccessUplink RIS-assisted NOMA/OMA networks: a theoretical study
In this paper, we have investigated the outage probabilities (OPs) and the ergodic rates (ERs) of uplink reconfigurable intelligent surface (RIS)-assisted non-orthogonal and orthogonal multiple-access (NOMA and OMA) networks. Particularly, we have tackled a challenging scenario in which both user equipments (UEs) need assistance from the same RIS unit to communicate with the base station (BS). We have modeled all the UEs $$\rightarrow$$ RIS and RIS $$\rightarrow$$ BS links as Nakagami- m variates. To assess the performance of these networks, we have derived the necessary statistics for the effective channel gain of the reflected (UE $$\rightarrow$$ RIS $$\rightarrow$$ BS) link for the randomly phase-shifted UE, while adopting its already-existing counterpart statistics for the coherently phase-shifted one. On the basis of that, we have derived closed-form expressions for the OPs and the ERs for both RIS-assisted NOMA and OMA schemes. Finally, to get further insights on these results, we have elaborated on the high signal-to-noise ratio (SNR) regime.