Vision Transformers in Medical Imaging: A Review
Emerald U. Henry*, Onyeka Emebo, Conrad Asotie Omonhinmin
arXiv preprint, 2022. DOI: 10.48550/arXiv.2211.10043. (*co-first author — ~100 citations)

A comparative systematic review of Vision Transformer (ViT) architectures in medical imaging across 200+ literature sources, spanning MRI, CT, Ultrasound, and Whole Slide Imaging modalities. We describe the transformer model, compare it with a diversity of convolutional neural networks (CNNs), and detail transformer-based approaches for medical image classification, segmentation, registration, and reconstruction with a focus on the imaging modality. The analysis quantitatively determines that hybrid CNN+ViT methods outperform pure CNN approaches by approximately 20% on standard medical datasets.

Conditional Monitoring and Fault Detection of Wind Turbines based on Kolmogorov-Smirnov's Non-Parametric Test and Machine Learning
Olayinka S. Ohunakin*, Emerald U. Henry*, Olaniran J. Matthew, Victor U. Ezekiel, Damola Adelekan, Ayodele Oyeniran
Energy Reports, Vol. 11 (2024), 2577–2591. Elsevier, SCIE/Scopus indexed. DOI: 10.1016/j.egyr.2024.01.081. (*co-first authors — 28 citations)

This research presents a new method for conditional monitoring based on the wind turbine power curve. The Kolmogorov-Smirnov distribution test from the field of statistics is employed in the assessment of turbine data and the detection of abnormality (faults) in wind turbines. The process begins with anomaly detection and filtration of faulty SCADA data by a quantile based filtration approach. Useful data comprising wind speed, density, ambient temperature and pitch angle are utilized in the development of wind turbine power curve models that represents actualities within wind farms. The radial basis function (RBF), multi-layer Perceptron (MLP) and gradient boosting (GBR) methods utilized for model development are compared for predictive accuracy using Mariano-Preve test, the null hypothesis assumes equal predictive ability (EPA). If rejected, an algorithm compares the coefficients of correlation of the models and selects the closest to one (unity). The most accurate model is utilized for the creation of a bin-wise distribution from past data and bin-wise confidence levels from the plot of wind speed and output power. Cochran’s method validates the minimum sample size that will possess a sampling distribution similar to that of the population, a fault is detected if there is a reasonable difference between the sample distribution and population distribution. The Kolmogorov-Smirnov test, having a null hypothesis of equivalent distributions, signals a fault if the null hypothesis is rejected. Two wind turbine SCADA data sets associated with two fault events are used for the assessment of our method. Our results indicate that our method effectively highlights abnormalities in power output relating to increased bearing temperature and reduced generator rpm, aiding in the detection of faults long before they occur.

Neural Network-Based Wind Turbine Power Curve Model Using Several Wind Farms' Influencing Parameters and Topography
Olayinka S. Ohunakin*, Emerald U. Henry*, Victor U. Ezekiel, Olaniran J. Matthew, Damola Adelekan, Mutali Napfumbada
In: Localized Energy Transition in the 4th Industrial Revolution, ed. Laseinde & Eboka, pp. 162–190. Taylor & Francis, 2024. DOI: 10.1201/9781032651958-10. (*co-first authors)

Wind turbine power curve (WTPC) modelling is of great importance for energy assessment and forecasting. In previous works, WTPC models were developed based on wind speed only. However, in this research, we developed modelling methods that represent actual WTPC by extensively considering wind farms’ topography, and several field conditions (other than wind speed only) that are found to influence the power output of wind turbines such as climate variability, the effect of neighbouring wind turbines, turbulence intensity, wake effect, ambient temperature, atmospheric pressure, wind direction, and terrain conditions. We analyze the radial basis function (RBF) and multi-layer perceptron (MLP) architectures for sensitivity and modelling accuracy. A filtered dataset is passed into the models and fitting accuracies are computed alongside sensitivity analysis. The best-performing models are compared with numerous parametric and non-parametric WTPC modeling schemes. It is found that the quantile filtering (QF-NN) models outperforms all other models in terms of fitting accuracy, and outperforms all selected hybrid models in terms of computation time.

Techno-economic assessment of offshore wind energy potential at selected sites in the Gulf of Guinea
Olayinka S. Ohunakin, Olaniran J. Matthew, Windmanagda Sawadogo Emerald U. Henry Victor U. Ezekiel Adelekan S. Damola

Offshore wind power has been found to stand out among the most dynamic renewable energy technologies. With its long coastal line, Nigeria has an overwhelming advantage in developing marine energy resources to relieve the power crisis effectively. This work analyzed and characterized observation data of sea-surface wind speed and direction at 30-minute intervals between 1979 and 2015 at five synoptic offshore stations in the Gulf of Guinea. The seasonal variations in hourly surface wind speed and directions as well as the Weibull distribution of wind speed and wind power at 100 m hub height were examined. The wind shears, capacity factors, and accumulated energy outputs for seven offshore wind turbine types were determined for the selected locations. In addition, the economic analysis of the selected offshore turbines using levelized cost of energy was carried out, while sensitivity analysis of the total levelized cost of energy to key input parameters was further determined. The results revealed large spatial and temporal variations in wind speed and wind power in the Gulf of Guinea. The most viable offshore site for wind energy exploitation was Agbami (the deepest offshore site), while Bonny (the shallow coastal site) had the least. The findings established very good fits (having mean bias (between − 0.08 ms− 1 and − 2.44 ms− 1 ), percentage bias (between − 0.47% and − 13.98%), correlation coefficients (be- tween 0.97 and 0.98), Chi-square (between 0.2 and 1.2), and root mean square error (between 1.2 ms− 1 and 3.1 ms− 1 )) between Weibull distribution and the actual wind data. The wind turbines with the highest and the lowest wind power densities, capacity factors, and power outputs across the seasons and sites were V236-15.0 MW and Siemens SWT113, respectively. The levelized cost of energy was considered for the deep waters due to the moderately high-capacity factors. The highest values ranged between 101.48 and 137.12 USD/MWh at Sea Eagle with V236-15 MW and V117-4.2 MW, respectively, while the lowest ranged from 52.29 to 69.66 USD/MWh at Agbami with V236-15 MW and Siemens (SWT113), respectively. The exploitation of Nigeria’s offshore wind resources could also be dedicated to producing renewable hydrogen and can serve to meet the country’s ambitious targets set for carbon neutrality by 2060.

Design of the electrical system of a mini-racecar: A report
Emerald Henry

Formula SAE is a global competition, which challenges students to design and build a formula-style racecar, which they then compete with in a series of dynamic and static events. Hebron Motorsports, the formula student team in Covenant University designed and is currently implementing their designs against the 2023 Formula Student competition in Silverstone UK. The electrical team considered three main systems to ensure durability and simplicity, being the first design ever attempted. They include; critical vehicle systems comprising engine control unit, sensors and actuators, the FSAE required safety system comprising safety shutdown buttons, switches, relay networks, break system plausibility device and circuit protection, electro-pneumatic shifting and additional designs comprising digitally controlled shifting, gear count display, neutral detector lights. Additionally, we considered compact wiring, purchasing cables of required length and compactness. We also ensured to employ standard terminations of cables to ensure system integrity. For the Safety Circuit Board, we ensured to incorporate appropriate redundancy in design that led us to print them as opposed to using perf boards whereas the shifting system circuit was implemented on a perf board. Schematics and detailed documentations were drafted on paper in order to help convey the information to people of other disciplines. Each system was prototyped on a bench or by utilizing bred boards before they were implemented on the vehicle hence, separate wirings had to be provided.