246 research outputs found
The detection of abrupt changes using recursive identification for power system fault analysis
No full textCopyright © 2008 Elsevier B.V.This paper describes the application of the recursive parameter estimation technique used to detect the abrupt changes in the signals recorded during disturbances in the power network of South Africa. The recursive identification technique uses M parallel Kalman filters. Main focus has been to estimate the time-instants of the changes in the signal model parameters during the pre-fault condition and following the events like initiation of fault, circuit-breaker opening, auto-reclosure of the circuit-breakers and the like. After segmenting the fault signal precisely into these event-specific sections, further signal processing and analysis can be performed on these segments, leading to automated fault recognition and analysis. In the scope of this paper, we focus on the first task, that is, segmenting the fault signal into event-specific sections using the recursive identification technique.Abhisek Ukil and Rastko Živanovićhttp://www.elsevier.com/wps/find/journaldescription.cws_home/504085/description#descriptio
Automated analysis of power systems disturbance records: Smart Grid big data perspective
No full textAnalysis of faults and disturbances play crucial roles in secure and reliable electrical power supply. Digital fault recorders (DFR) enable digital recording of the power systems transient events with high quality and huge quantity. However, transformation of data to information, expectedly in an automated way, is a big challenge for the power utilities worldwide. This is a key focus for realizing the `Smart Grid'. In this paper, the architecture and specifications for the primary and the secondary information for the automated systems are described. This provides qualitative and quantitative guidelines about the information to derive out of the disturbance data. A quantified estimate of big data for the substations, has been estimated in the paper. Possible ways of reducing the big data by utilizing intelligent segmentation techniques are described, substantiated by real example. Utilization of centralized protection and remote disturbance analysis for reducing big disturbance data are also discussed.Abhisek Ukil, Rastko Zivanovi
Adjusted Haar wavelet for application in the power systems disturbance analysis
No full textAbrupt change detection based on the wavelet transform and threshold method is very effective in detecting the abrupt changes and hence segmenting the signals recorded during disturbances in the electrical power network. The wavelet method estimates the time-instants of the changes in the signal model parameters during the pre-fault condition, after initiation of fault, after circuit-breaker opening and auto-reclosure. Certain kinds of disturbance signals do not show distinct abrupt changes in the signal parameters. In those cases, the standard mother wavelets fail to achieve correct event-specific segmentations. A new adjustment technique to the standard Haar wavelet is proposed in this paper, by introducing 2n adjusting zeros in the Haar wavelet scaling filter, n being a positive integer. This technique is quite effective in segmenting those fault signals into pre- and post-fault segments, and it is an improvement over the standard mother wavelets for this application. This paper presents many practical examples where recorded signals from the power network in South Africa have been used. © 2007 Elsevier Inc. All rights reserved.Abhisek Ukil and Rastko Živanovićhttp://www.elsevier.com/wps/find/journaldescription.cws_home/622818/description#descriptio
Abrupt change detection in power system fault analysis using wavelet transform
No full textThis paper describes the application of the wavelets used to detect the abrupt changes in the signals recorded during disturbances in the electrical power network in South Africa. Main focus has been to estimate exactly the timeinstants of the changes in the signal model parameters during the pre-fault condition and following events like initiation of fault, circuit-breaker opening, auto-reclosure of the circuit-breakers using the wavelet transform, particularly the dyadicorthonormal wavelet transform. The key idea is to decompose the fault signals into effective detailed and smoothed version using the multiresolution signal decomposition technique based on discrete wavelet transform. Then we apply the threshold method on the decomposed signals to estimate the change time-instants, segmenting the fault signals. After segmenting the fault signal precisely into the event-specific sections, further signal processing and analysis can be performed on these segments, leading to automated fault recognition and analysis. In the scope of this paper, we focus on the first task i.e., segmentation of the fault signal into event-specific sections using the wavelet transform and threshold method. This paper presents application on recorded signals in the transmission network of South Africa.Abhisek Ukil, Rastko Živanovi
Abrupt change detection in power system fault analysis using adaptive whitening filter and wavelet transform
No full textCopyright © 2005 Elsevier B.V. All rights reserved.This paper describes the application of the adaptive whitening filter and the wavelet transform used to detect the abrupt changes in the signals recorded during disturbances in the electrical power network in South Africa. Main focus has been to estimate exactly the time-instants of the changes in the signal model parameters during the pre-fault condition and following events like initiation of fault, circuit-breaker opening, auto-reclosure of the circuit-breakers. The key idea is to decompose the fault signals, de-noised using the adaptive whitening filter, into effective detailed and smoothed version using the multiresolution signal decomposition technique based on discrete wavelet transform. Then we apply the threshold method on the decomposed signals to estimate the change time-instants, segmenting the fault signals into the event-specific sections for further signal processing and analysis. This paper presents application on the recorded signals in the power transmission network of South Africa. © 2005 Elsevier B.V. All rights reserved.Abhisek Ukil and Rastko Živanovićhttp://www.elsevier.com/wps/find/journaldescription.cws_home/504085/description#descriptio
Application of abrupt change detection in power systems disturbance analysis and relay performance monitoring
No full text"This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright. In most cases, these works may not be reposted without the explicit permission of the copyright holder."This paper describes the application of the abrupt change detection technologies to detect the abrupt changes in the signals recorded during disturbances in the electrical power network of South Africa for disturbance analysis and relay performance monitoring. The aim is to estimate the time instants of the changes in the signal model parameters during the prefault condition, after initiation of fault, after the circuit-breaker opening and autoreclosure, etc. After these event-specific segmentations, the synchronization of the different digital fault recorder recordings are done based on the fault inception timings. The synchronized signals are segmented again. This synchronized segmentation is the first step toward automatic disturbance recognition, facilitating further complex feature vector analysis and pattern recognition. Besides, the synchronized, segmented recordings can be directly used to analyze certain kinds of disturbances and monitor the relay performance. This paper presents many practical examples from the power network in South Africa.Abhisek Ukil and Rastko Zivanovi
Simulation and study of hybrid energy storage system for microgrid
No full textHybrid Energy Storage System (HESS), comprising of batteries and ultracapacitors, is
becoming increasingly important for micro grids. While batteries are characterized by low
discharge current but higher operation time ranging from 1hour to 25 hours, ultracapacitors are
used for heavy discharge current in short period of time usually in milliseconds to few seconds.
Purpose of this project is to study the effect of highly fluctuating load demands and associated transients on the stability of HESS operation.
During this project, the author models batteries, ultracapacitors along with converters to
simulate the HESS in MATLAB. Due to time constraint, this report is strictly limited to evaluation
of the stability of HESS between energy sources. The author will also design the control algorithm
for dealing with highly fluctuating load demands. Stability of HESS can be evaluated based on
optimum power allocation between different energy sources without adding stress on any of them.Bachelor of Engineerin
Aprendizaje computacional para análisis de señales de sensores
No full textObjetivo: el objetivo general es construir modelos precisos de aprendizaje automático para resolver desafíos prácticos como la escasez de datos de entrenamiento, la construcción de modelos compactos y la preservación de la privacidad de los datos para un conjunto diverso de tareas de análisis de señales de sensores. Con la proliferación de Internet de las cosas (IoT), los avances de las tecnologías de detección, las increíbles mejoras hacia el poder de cómputo junto con el progreso sobresaliente de los algoritmos y herramientas de inteligencia artificial, los investigadores están encontrando nuevas vías para crear diferentes aplicaciones útiles y direcciones de investigación novedosas. El trabajo de investigación se centra en la construcción de modelos para el aprendizaje computacional de tareas de análisis que involucran diferentes tipos de señales de sensores de sensores como electrocardiograma, fonocardiograma, acelerómetro, medidor de energía, etc. Muchos sensores pueden considerarse como la micro-representación de la fisiología humana y la actividad humana y tales sensores contienen información sensible. Por lo tanto, nuestra tarea principal es la habilitación de técnicas de preservación de la privacidad como parte de los modelos de detección computacional que analizan las señales de los sensores e infieren decisiones críticas.
Metodología : se entiende que la atención médica remota es una de las aplicaciones críticas de IoT y resolvemos el problema de la protección de la privacidad de los datos al proponer la eliminación del riesgo de la gestión de datos confidenciales mediante la privacidad diferencial, donde la protección de privacidad controlada habilitada por el usuario en datos de atención médica confidenciales puede ser empleado. El método de protección de la privacidad propuesto ofusca los datos confidenciales para garantizar que se realice una protección adecuada mientras que la utilidad no se ve gravemente comprometida, y el control de la habilitación de la privacidad está dirigido por el usuario. La limitación de este trabajo es que el algoritmo de aprendizaje automático que realiza la tarea de análisis requiere una ingeniería de funciones artesanal, que no solo restringe la escalabilidad del aprendizaje computacional, sino que también depende del costoso proceso de generación de funciones asistida por expertos o conocimiento del dominio. y selección. Desarrollamos detección integrada de inteligencia que realiza tareas de clasificación supervisadas utilizando un método novedoso de aprendizaje profundo (DL) de red neuronal convolucional ajustada por hiperparámetros sin esfuerzos de ingeniería de requisitos. Ampliamos nuestra investigación para abordar el problema integral de la escasez de datos de entrenamiento en la generación de modelos DL. Se sabe que los modelos DL exigen ejemplos de entrenamiento sustanciales para la construcción confiable del modelo computacional. Las tareas prácticas de análisis de señales de sensores a menudo se proporcionan con un número limitado de ejemplos de capacitación, principalmente debido a los costos asociados con la anotación de expertos. Proponemos un método novedoso de aprendizaje efectivo bajo la limitación de datos de entrenamiento utilizando el descubrimiento atribuido por Shapley de un subconjunto de entradas que influyen positivamente para construir un modelo DL efectivo basado en redes residuales.
Resultados : nuestro novedoso método de preservación de la privacidad propone el principio de incertidumbre de los datos del sensor, de modo que se emplea la incertidumbre estadística controlada para la información confidencial usando como definición de protección de la privacidad que las probabilidades a priori y a posteriori de encontrar información privada no cambian más allá de un umbral predefinido y la ganancia del adversario en el acceso a datos confidenciales se vuelva insignificante. La estimación de hiperparámetros propuesta a partir de las características de la señal de entrada facilita la construcción del modelo CNN compacto y demostramos que el modelo propuesto supera constantemente los algoritmos de última generación relevantes para la tarea de aprendizaje computacional dada de detección de condiciones de fibrilación auricular a partir de registros de ECG de una sola derivación. Con la novedosa arquitectura push-pull DL propuesta, donde la selección del subconjunto de entrada a través de la atribución del valor de Shapley empuja el modelo a una dimensión más baja mientras que el entrenamiento adversario aumenta la capacidad de aprendizaje del modelo sobre datos no vistos, demostramos un rendimiento superior a algoritmos actuales de última generación para tareas de clasificación sobre diversos conjuntos de señales de sensores de series temporales.
Conclusión : hemos propuesto un marco holístico para resolver los desafíos prácticos y de investigación del análisis computacional de las señales de los sensores, incluida la preservación de la privacidad de los datos, el algoritmo de aprendizaje profundo para la generación de modelos compactos, el modelo computacional efectivo bajo el problema de la escasez de datos de entrenamiento. En resumen, el trabajo de investigación proporciona un enfoque unificado para desarrollar un análisis computacional práctico para diversos conjuntos de datos de sensores.Objective- The general objective is to build accurate machine learning models to solve practical challenges like training data scarcity, compact model construction and data privacy preservation for diverse set of sensor signal analysis tasks. With the proliferation of Internet of Things (IoT), advancements of sensing technologies, incredible enhancements towards computing power along with the outstanding progress of Artificial Intelligence algorithms and tools, researchers are finding new avenues to build different useful applications and novel research directions. The research work focuses on the construction of models for computational learning of analysis tasks involving different types of sensor signals from sensors like Electrocardiogram, Phonocardiogram, accelerometer, energy meter etc. In general, we can consider sensors as the micro-representation of our ambient world. Given that sensors capture near-human information, they usually contain sensitive data. Hence, our foremost task is the enablement of privacy preserving techniques as part of the computational sensing models that analyze the sensor signals and infer critical decision.
Methodology- It is understood that remote healthcare is one of the critical applications of IoT and we solve the problem of data privacy protection by proposing de-risking of sensitive data management using differential privacy, where user-enabled controlled privacy protection on sensitive healthcare data can be employed. We propose a novel data privacy preservation method that obfuscates the sensitive component of the sensor data while utility is not severely compromised, while user controls the quantum of privacy. The proposed machine learning algorithm requires subtly hand-crafted feature engineering, which not only restricts the scalability of the computational learning, but also depends on the expensive process of expert or domain-knowledge aided feature generation and selection. We develop intelligence-embedded sensing that does supervised classification tasks using novel deep learning (DL) method of hyperparameter-adjusted convolutional neural network without feature engineering efforts. We extend research to address the integral problem of training data scarcity in DL model generation. It is known that DL models demand substantial training examples for reliable construction of the computational model. Practical sensor signal analysis tasks are often provided with limited number of training examples mainly due to the costs associated with expert annotation. We propose a novel method of effective learning under training data limitation using Shapley-attributed discovery of subset of positively influencing inputs to construct an effective Residual network-based DL model.
Results- Our novel privacy preserving method proposes sensor data uncertainty principle, such that controlled statistical uncertainty is employed to the sensitive information with the definition of privacy protection that the prior and posterior probabilities of finding private information does not change beyond a pre-defined threshold and the adversary's gain of sensitivity data access becomes insignificant. The proposed hyperparameter estimation from the input signal characteristics facilitates compact CNN model construction. We demonstrate that our model consistently performs superior over the relevant state-of-the-art algorithms for the given computational learning task of Atrial Fibrillation condition detection from single-lead ECG recordings. We propose an unique push-pull DL architecture, where, firstly Shapley value attributed input subset selection pushes the model parameters towards lower dimension and subsequently, we augment the learnability of the model through adversarial training. We demonstrate the efficacy of proposed model that empirically outperforms the current state-of-the-art algorithms in diverse set of time series sensor signal classification tasks.
Conclusion- We have proposed a holistic framework to solve the practical and research challenges of computational analysis of sensor signals including the data privacy preservation, deep learning algorithm for compact model generation, effective computational model under training data scarcity issue. In summary, the research work provides a unified approach to develop practical computational analysis for diverse set of sensor data
Design of deadbeat control method for Hybrid Energy Storage System in microgrids
No full textTraditional power generation methods use large amounts of fossil fuels, resulting in
carbon emissions and worsening the greenhouse effect, so green energy generation
has been developed, and with it the need for energy storage for microgrids. The
Hybrid Energy Storage System (HESS) includes batteries and supercapacitors.
Supercapacitors extend battery life and can manage high frequency charging and
discharging. As green energy generation is easily affected by external factors, the
HESS can charge and discharge the microgrid to ensure smooth output power and
enhance the stability of microgrid operation. In this paper, a more efficient controller
is investigated to control the charging and discharging of the HESS to the microgrid
through feedback from the current and voltage loops to output more stable power and
improve the output power quality. Simulink simulation results were used to compare
the simulation results of PI control and Deadbeat control, to observe the advantages
and disadvantages of Deadbeat control and the improvements to the HESS. The first
step of the study was to build a HESS using Simulink (supported by Matlab) and the
second step was to build PI control and Deadbeat control respectively in the HESS.
The third step was to vary the sun irradiance and load resistance values to observe the
simulation results of the two controls. According to the simulation results it can be
seen that Deadbeat control has a faster settling time and smaller overshoot values than
traditional PI control
- …
