49 research outputs found
MangoLeafBD Dataset
No full textType of data: 240x320 mango leaf images.Data format: JPG.Number of images: 4000 images. Of these, around 1800 are of distinct leaves, and the rest are prepared by zooming and rotating where deemed necessary.Diseases considered: Seven diseases, namely Anthracnose, Bacterial Canker, Cutting Weevil, Die Back, Gall Midge, Powdery Mildew, and Sooty Mould.Number of classes: Eight (including the healthy category).Distribution of instances:Each of the eight categories contains 500 images.How data are acquired: Captured from mango trees through the mobile phone camera.Data source locations: Four mango orchards of Bangladesh, namely Sher-e-Bangla Agricultural University orchard, Jahangir Nagar University orchard, Udaypur village mango orchard, and Itakhola village mango orchard.Where applicable: Suitable for distinguishing healthy and diseases leaves (two-class prediction) as well as for differentiating among various diseases (multi-class prediction)
Matrix converter based voltage regulator for managing smart microgrid with high renewable energy penetrations
No full textThis thesis examines the use of matrix converter (MC) in voltage regulators intended for use in the low voltage (LV) distribution network. Voltage balance and voltage regulation can be controlled by adding a series compensation voltage with a transformer. The MC supplies the
injection transformer with an appropriate voltage which may be significantly unbalanced.
Prior-art MCs have been predominantly applied in motor drives which have relatively balanced
voltages and currents. This thesis shows the existing methods are not ideal for very unbalanced situations. It is shown that a traditional MC supplying an unbalanced load develops current harmonics at its input terminal. This thesis provides an improved MC modulation method that
can eliminate input harmonics. This result is confirmed by simulation and experimental work.
The thesis extends the traditional 3× 3 matrix into four wire compatible MC topologies including 3×4, 4×3 and 4×4 converters. For these cases, and generalised cases, solutions are presented for switch commutation. These are implemented with a field programmable gate array (FPGA). An experimental study is made of switch commutation for silicon and silicon carbide metal oxide semiconductor field-effect transistors (MOSFETs)
Matrix converter based voltage regulator for managing smart microgrid with high renewable energy penetrations
No full textThis thesis examines the use of matrix converter (MC) in voltage regulators intended for use in the low voltage (LV) distribution network. Voltage balance and voltage regulation can be controlled by adding a series compensation voltage with a transformer. The MC supplies the
injection transformer with an appropriate voltage which may be significantly unbalanced.
Prior-art MCs have been predominantly applied in motor drives which have relatively balanced
voltages and currents. This thesis shows the existing methods are not ideal for very unbalanced situations. It is shown that a traditional MC supplying an unbalanced load develops current harmonics at its input terminal. This thesis provides an improved MC modulation method that
can eliminate input harmonics. This result is confirmed by simulation and experimental work.
The thesis extends the traditional 3× 3 matrix into four wire compatible MC topologies including 3×4, 4×3 and 4×4 converters. For these cases, and generalised cases, solutions are presented for switch commutation. These are implemented with a field programmable gate array (FPGA). An experimental study is made of switch commutation for silicon and silicon carbide metal oxide semiconductor field-effect transistors (MOSFETs).</p
Dynamic Stability Analysis in Distribution Networks with Dynamic Loads and Renewable Energy Sources
No full textThe main focus of this thesis is the modelling, control and stability analysis of
distribution networks with distributed generators and dynamic loads in which the
dynamic loading effects in standard distribution networks are taken into consideration. Different case studies are considered, such as fault effects, a worst-case
scenario and nodal voltage analyses of different network configurations. The dynamic modelling includes line resistance, which has been neglected in the existing
literature, but which the case studies show is a critical parameter affecting system
stability. An LQR controller is proposed for minimising the effect of resistance in
the distribution network. A novel linear zero dynamic controller (LZDC) is design
to maintain the voltage and angle stability for distributed generators. Some elementary notions for the LZDC are introduced such as relative degree, Lie derivative
and exact linearization. This thesis also presents a new concept of a multiple input
multiple output (MIMO) LZDC for a three-phase grid-connected photovoltaic (PV)
system to enhance its stability and robustness under different weather conditions.
Grid-connected PV systems are highly nonlinear systems in which most of the non-
linearities occur due to the intermittency of sunlight and the switching functions of
their converters and inverters. The proposed controller overcomes the limitations of
other controllers, such as the PI, hysteresis, predictive and sliding-mode controllers,
and it is proven that this system operates at unity power factor. The effectiveness of
the proposed control strategies are demonstrated through time-domain simulation
studies conducted using the standard industry-based software environment
An improved modulation method for matrix converters with unbalanced output loading
No full textIn this paper Instantaneous Reactive Power (IRP) theory is applied to develop an improved modulation method for Matrix Converters (MCs) operating with unbalanced loads. IRP theory shows that, for unbalanced loading conditions, the instantaneous power contains a constant component and a component that fluctuates at twice the fundamental frequency. In conventional modulation schemes the power fluctuation results in input current harmonics. The new method removes the harmonic distortion by synthesizing an input current vector that contains positive and negative sequence components that satisfy the power balance requirements but have only fundamental frequency components. The paper shows that the conventional modulation method may be converted to produce the required input current vector by phase modulation of the input current vector
A matrix converter based voltage regulator for MV rural feeders
No full textIn this paper a Matrix Converter (MC) has been used with a series voltage injection transformer to regulate the voltages on a Medium-Voltage (MV) rural distribution feeder. This approach is well suited to pole top regulator applications as DC bus capacitors are avoided. The MC is controlled using a space vector modulation (SVM) method that is capable of regulating the positive-sequence voltage on a feeder while simultaneously cancelling negative sequence voltages produced by high levels of load or distributed generation unbalance or by the use of non-transposed lines. The study of the MC modulation processes is undertaken with an exact switching model for the higher frequency behaviours of the converter while a small signal Instantaneous Reactive Power (IRP) based average signal model is used to study the multi-cycle dynamic regulatory performances
A matrix converter based voltage regulator for MV rural feeders
No full textIn this paper a Matrix Converter (MC) has been used with a series voltage injection transformer to regulate the voltages on a Medium-Voltage (MV) rural distribution feeder. This approach is well suited to pole top regulator applications as DC bus capacitors are avoided. The MC is controlled using a space vector modulation (SVM) method that is capable of regulating the positive-sequence voltage on a feeder while simultaneously cancelling negative sequence voltages produced by high levels of load or distributed generation unbalance or by the use of non-transposed lines. The study of the MC modulation processes is undertaken with an exact switching model for the higher frequency behaviours of the converter while a small signal Instantaneous Reactive Power (IRP) based average signal model is used to study the multi-cycle dynamic regulatory performances
An improved modulation method for matrix converters with unbalanced output loading
No full textIn this paper Instantaneous Reactive Power (IRP) theory is applied to develop an improved modulation method for Matrix Converters (MCs) operating with unbalanced loads. IRP theory shows that, for unbalanced loading conditions, the instantaneous power contains a constant component and a component that fluctuates at twice the fundamental frequency. In conventional modulation schemes the power fluctuation results in input current harmonics. The new method removes the harmonic distortion by synthesizing an input current vector that contains positive and negative sequence components that satisfy the power balance requirements but have only fundamental frequency components. The paper shows that the conventional modulation method may be converted to produce the required input current vector by phase modulation of the input current vector
Selecting Update Blocks of Convolutional Neural Networks using Genetic Algorithm in Transfer Learning
No full textThe performance of convolutional neural networks (CNN) for computer vision problems depends heavily on their architectures. Transfer learning performance of a CNN strongly relies on selection of its trainable layers. Selecting the most effective update layers for a certain target dataset often requires expert knowledge on CNN architecture which many practitioners do not possess. General users prefer to use an available architecture (e.g. GoogleNet, ResNet, EfficientNet etc.) that is developed by domain experts. With the ever-growing number of layers, it is increasingly becoming difficult and cumbersome to handpick the update layers. Therefore, in this paper we explore the application of a genetic algorithm to mitigate this problem. The convolutional layers of popular pre-trained networks are often grouped into modules that constitute their building blocks. We devise a genetic algorithm to select blocks of layers for updating the parameters. By experimenting with EfficientNetB0 pre-trained on ImageNet and using three popular image datasets - namely Food-101, CIFAR-100 and MangoLeafBD - as target datasets, we show that our algorithm yields similar or better results than the baseline in terms of accuracy, and requires lower training and evaluation time due to learning a smaller number of parameters. We also devise a measure called block importance to measure each block’s efficacy as an update block and analyze the importance of the blocks selected by our algorithm
