Search

In this paper, a hybrid feature extraction and classification technique is proposed for fault detection in a power distribution network. The scheme uses the wavelet packet transform (WPT) technique for signal analysis and the support vector machine (SVM) technique is used for fault classification. The WPT is used to extract fault current statistical feature from 1/4 of the post fault current signal. The extracted features are subsequently utilized to train and test the SVM scheme for fault classification. Prior to the classification of the fault, the genetic algorithm (GA) technique is used to determine the optimal parameters of the SVM classifier. The results show that the fault classification on distribution line can be determined with high accuracy even for high impedance faults. The classification results are improved from 93.8 % to 98.8 %, when using the WPT technique. The proposed method is tested on a machine learning platform called ORANGE.

In this day and age, there is a proliferate concern from all governments across the globe to barricade the environment from the greenhouse gases, which absorb infrared radiation. As a result, solar photovoltaic (PV) energy has been an expeditiously growing renewable energy source and will eventually undertake a prominent role in the global energy generation. The selection and purchasing of energy efficient transformers which meet the operational requirements of the solar photovoltaic energy generation plants then becomes a part of the Independent Power Producers (IPP’s) investment plan of action. Taking these into account, this work proposes a novel procedure that put into effect the intricate financial analysis necessitated to precisely evaluate the transformer service lifetime no-load and load loss factors. This procedure correctly set forth the transformer service lifetime loss factors as a result of solar PV plant’s sporadic generation profile and related levelized costs of electricity into the computation of the transformer’s total ownership cost. The results are then critically compared with the conventional transformer total ownership cost (TCO), and demonstrate the significance of the sporadic energy generation nature of the solar PV plant on the total ownership cost. The findings indicate that the latter play a crucial role for developers and Independent Power Producers (IPP’s) in making the purchase decision during a tender bid where competing offers from different transformer manufactures are evaluated. Additionally, the susceptibility analysis of different factors engrossed in the transformer service lifetime cost is carried out, factors including the levelized cost of electricity, solar PV plant’s generation modes and the loading profile are examined.

This paper presents a high impedance fault (HIF) detection scheme based on the Hubbard-Stanovich (HS) transform and decision tree (DT). The HS-transform is used to extract features from the fault signal. The extracted features are used as input to train the DT classifier. Different cases are studied including HIF, load switching, capacitor switching and normal current. The proposed method is tested using a 22 kV practical network and implemented on the MATLAB/Simulink platform. The classification analysis was performed using a data science platform called WEKA. The presented results show that the proposed scheme is able to distinguish HIF from other power system operational cases. The proposed method showed good accuracy results.

The global energy sector has been faced with energy scarcity challenges over the past few years. This coupled with environmental concerns due to electricity generation using fossil fuels. This has led to the growth of eco-friendly renewable energy distributed generation (REDG) for electricity generation. In the planning of integrating REDG into the distribution power grid, inappropriate sizing and location of the REDG will result in power system voltage instability, increase in power losses and operational cost increase. It is therefore essential to optimize the location and sizing of REDG for effective performance improvement. In this paper, we propose a genetic algorithm (GA) technique to optimize the size and location of the REDG. The proposed GA technique is validated using the 33 IEEE bus system. From the results obtained it is observed that the active power losses are reduced and the voltage stability of the system is improved

The modern electricity society has seen an increase in the integration of renewable energy sources into the power grid. This has significantly improved the electricity supply and sustainability. However, there are technical challenges related to the power quality (PQ) analysis for grid integrated systems. In this paper, we propose a PQ detection technique using the discrete wavelet transform (DWT) and support vector machine (SVM). The DWT analysis technique is used to extract statistical features, which are used as input to train the SVM classifier. The parameters of the SVM are optimised using the harris hawks optimisation (HHO) algorithm. Various scenarios of cases which may affect the quality of network performance are investigated. These events include the voltage sag, voltage swell, notch, transient fault conditions and sudden load increment. The proposed method is validated using the modified practical Eskom network. The presented results show that the proposed scheme correctly classified different cases.

Fault determination and isolation is an important aspect for maintaining the health index of a power grid. In this paper, a protection fault scheme is proposed. The protection scheme uses a discrete wavelet transform (DWT), genetic algorithm (GA) and neural network (NN). The DWT technique is used to analyze fault current signals at different levels. From the analyzed signals, statistical features are extracted to minimize the size of the original signal to improve the computational efficiency. Subsequently, the features are used to train and test the NN fault classifier scheme. To improve the performance of the classier, the GA technique is used to determine the optimal parameters of the NN scheme. The scheme is tested on a practical network and a high accuracy is obtained for fault determination.

Integration of renewable energy distribution generators (REDGs) into the existing power grid has become more popular due to the technical benefits. The optimal location and sizing of the REDGs, particularly the photovoltaic (PV) and wind turbines (WT), remains a challenge due to the stochastic nature of these energy sources. In this work, we propose a method based on the Harris Hawks Optimization (HHO) algorithm to solve the optimal planning of REDGs problem. The HHO technique is used to minimize the active power losses and maximize the voltage stability of a network integrated with the REDGs. The proposed HHO technique is validated on the standard 33 and 69 IEEE bus systems. The results show that the proposed HHO technique improves the voltage stability and minimizes the active power losses in the system.

The protection of electrical distribution network against lightning with high reliability is a critical aspect for power utilities. Lightning strokes influences the performance of a distribution systems. In this paper, we investigate the option of using streamer to minimize the effect of lightning in the Doulas MMS Coal 22kV plant. The network is modelled using MATLAB. Furthermore, we compared the surge arrestors and streamers and determined a suitable option for the application.

Water may be treated differently in different communities depending on the quality of the water that enters the treatment plant. Typically, surface water requires more treatment and filtration than groundwater because lakes, rivers, and streams contain more sediment and pollutants and are more likely to be contaminated than groundwater. Some water supplies may also contain disinfection by-products, inorganic chemicals, organic chemicals, and radionuclides. Specialized methods for controlling formation or removing them can also be part of water treatment. The study presents an approach to prioritize Water treatment plant maintenance activities based on system reliability and cost-effectiveness. The object of this approach is to minimize the weighted average system reliability index (WASRI) by ranking maintenance tasks based on their marginal benefit to cost ratios, where the benefit is defined as improvement in WASRI. Estimations of WASRI improvement are obtained by utilizing the linear relationship between the benefit obtained from a maintenance task and the change of failure rate of the maintained component.

The integration of renewable energy distributed generation (REDG) into the existing distribution networks (DNs) has over the years become popular. This practice has brought significant and numerous advantages in the technical, economic, and environmental sectors. However, locating REDGs in a random location may result in poor voltage profile and an increase in the active power losses. It is therefore imperative that the location of REDGs be optimized to obtain satisfactory results. In this research work, we propose an analysis-based method for REDGs location using a sensitive index algorithm. The method is carried out in order to improve the voltage profile of the power grid system and minimize the active power losses in the system. The validation of the proposed method is tested using an IEEE-24 bus system node arrangement. The results presented show that optimizing the locality of the REDGs improves the voltage and minimizes the power losses significantly.

The challenge with most developing countries is to maintain a reliable and sustainable electricity supply. This has a depleting effect on the economic development of most states. In order to reduce the impact of energy shortages, there has been an extensive attempt to use renewable energy sources to generate electricity. There are however technical challenges of integrating the existing power system grid with the renewable external sources. These challenges include adequate power system protection, energy security, and reliability of external sources. In this paper, we investigate the fault pattern recognition and detection in a power distribution grid integrated with the wind energy source. A reduced Eskom 22kV and wind power energy source integrated is modeled using MATLAB/Simulink. From the integrated model, various types of power systems faults are generated. We further investigated the use of local polynomial approximation (LPA) for signal decomposition and support vector machine (SVM) for fault classification and detection. We also tested the performance of the naive Bayes classifier. In this paper, a hybrid technique based on LPA and SVM is proposed for fault pattern recognition and detection in a power distribution integrated system with the wind energy source. The proposed method was further tested using machine learning platforms WEKA and Orange. The results of the classifiers gave the accuracy of between 98 and 99 %.

Power quality (PQ) detection in power systems is a critical aspect of condition monitoring. This helps in ensuring that technical operations are performed within permissible voltage and frequency limits. In this paper, we propose a PQ detection technique for disturbances associated with integrating wind power into the power distribution grid. An Eskom power system with wind power integration is modeled in Digsilent Power Factory software.. Discrete wavelet transform (DWT) is used to decompose voltage signals collected at the source terminals to detect PQ disturbances. The DWT technique is further used to calculate the energy content of the different distorted signals to classify PQ events. This is carried out to analyze the effect of disturbances on voltage and frequency of the power system. The technique aims to detect disturbances such as voltage swell, sag and frequency variations which may arise from integrating wind power into the power grid. From the results presented, it may be concluded operations such as synchronization of the wind power into the grid, power system faults with the integrated wind power and the wind power outage has a negative impact on the PQ of the system. The PQ analysis are analysed used MATLAB.

Integrating renewable energy distributed generation (REDG) into the existing power distribution grid has become a significantly exercise to carry out. This is because of several technical, economic and environmental benefits accruing from it. However, optimal location and sizing of REDG especially photovoltaic (PV) and wind turbine (WT), is still a difficult task due to the natural dependency of these renewable source on meteorological conditions. In this paper, we proposed a technique based on particle swarm optimization (PSO) to solve the location and sizing of REDG problem. The proposed PSO algorithm is used to minimize the power losses and maximize the voltage stability of the power grid distribution system integrated with REDG. The paper also presents a comparison of the proposed method and other related techniques for REDG sizing and location. The proposed method is validated using the IEEE 33 bus systems.

This paper presents aerosol studies carried out in Gaborone, the capital city of Botswana. The Gaborone aerosol is varied consisting of elements from Si to Au. Traffic contribution to the aerosol of Botswana is clearly visible as illustrated by strong positive bromine and lead correlation. The use of unleaded petrol could be the cause of the decrease of ambient lead (Pb) and bromine (Br) concentrations when the present measurements are compared to previous measurements. The elements present in the aerosol of Gaborone range from silicon to lead.

The generation of electricity from renewable energy sources (RES) is becoming more popular globally. This is because primary sources of electricity such as coal have a negative environmental impact. The introduction of RES into the existing power distribution grid has brought technical challenges. Fault detection with high reliability in power distribution network integrated with RES is one of the major challenges. In this paper, we propose a technique for fault detection in an integrated network. A reduced 22kV integrated power system is modelled in Digsilent Power Factory. Various fault current signals are generated from the model. Discrete wavelet transform (DWT) is used to extract statistical features from the fault current obtained through the study of the model. Subsequently, the extracted features are fed into the support vector machine scheme for fault detection and classification. In this paper, we also tested the performance of neural network (NN) and decision tree (DT) classifiers. A combined technique comprising of DWT and SVM is proposed. The proposed method is tested using a machine learning platform WEKA. The proposed method showed impressive classification results

Recently the growing demand for electricity has led to the increase of integrating renewable energy sources (RES) into the grid. This has proven to be an advantage for both power systems and environmental impacts. The introduction of RES has proven to be effective in minimizing power outages due to electricity shortage. Subsequently, this technology has been effective in preserving the atmosphere. However, RES integration has challenges which they bring into the power grid. In this paper, we analyzed the effect of connecting the wind turbine into the grid under fault conditions. From the results obtained, connecting RES increases the fault current which in-turn has the potential to increase the damages within the power grid.

Power distribution system plays a vital role in the development of a society and should be treated with vigorous care. Power system reliability empowers the utility to measure the performance of the network and to take necessary steps to improve it where needed. The economic growth of the utility in-terms of revenue is largely depended on the performance of the system. In this paper, the performance of a 22kV distribution line is analyzed. The analysis includes the root cause analysis, distribution indices, voltage regulation and proposes mitigating factors to improve the performance of the system.

Fault detection in power systems with high reliability has been a subject of concern for protection engineers over the years. In this paper, we model a 132kV distribution system in Power World software package. Various types of fault cases are obtained through the study of the model. Stationery wavelet transform (SWT) is used to decompose the signal into its coefficients and extract statistical features. Subsequent to the extraction of features, support vector machine (SVM) and artificial neural network (ANN) schemes are used for fault detection and classification. Support vector regression (SVR) scheme is used for fault location. A hybrid technique for fault diagnostic comprising of SWT-(SVM, ANN) and SVR is proposed. The method showed good accuracy of classification and minimum error for fault estimation. The proposed method is tested on a machine learning platform WEKA.

The water infrastructure utilizes a significant large amount of electricity. This is a concern in South Africa, where large-scale load shedding is required to reduce electricity demand. Since the important part of operational costs is energy related there is a comprehensible motivation for reducing energy consumption. The objective of this paper is to review, clarify and quantify the factors driving energy consumption based on the existing literature review. In addition, the review aimed where energy is consumed the most, provide information regarding what utilities can do to minimize energy consumption and achieve the optimized operation of the plant. Given that almost all of energy consumption in water utilities comes from pumping, there are also other vast, unexplored areas. Further research is still needed to improve energy efficiency in water supply systems

Fault detection, classification and estimation in power systems is one of the most critical aspects of the engineering society. This goes beyond engineering factors to economic implications. Thus, proper applications of protection schemes are required to minimize the equipment damage resulting from a fault. This paper presents a method which tries to proactively detect, classify and estimate the position of the fault. A simplified two bus 132 kV system is modelled to study the effect of the proposed fault diagnostic method. The proposed method has a fault feature extraction technique done by stationary wavelet transform (SWT) on the fault signal. Relevance Vector Machine (RVM) and Support vector machine (SVM) schemes are applied for fault classification and detection. Fault location along the distribution line is achieved by using Support Vector Regression (SVR). The proposed method comprises of SWT-RVM and SVR schemes and tested using MATLAB.

Power systems have undergone major transformations in recent times which have a brought dynamic change in the manner which transmission system is operated. This transformation can be attributed to factors such as population growth and industrialization, amongst others. What has been classified as the major issue of this transformation is power flow control in the interconnected systems. It is for this and other reasons that developed countries are installing and using power flow control devices in their power systems at an alarming rate. In this paper, the effect of using a phase shift transformer (PST) is demonstrated for power flow control.

High-impedance faults (HIFs) detection with high reliability has been a prominent challenge for protection engineers over the years. This is mainly because of the nature and characteristics this type of fault has. Although HIFs do not directly pose danger to the power system equipment, they pose a serious threat to the public and agricultural environment. In this paper, a technique which comprises of a signal decomposition technique, feature extraction, feature selection and fault classification is proposed. A practical experiment was conducted to validate the proposed method. The scheme is implemented in MATLAB and tested on the machine intelligence platform WEKA. The scheme was tested on different classifiers and showed impressive results for both simulations and practical cases.

In this paper, a method for detection and classification of faults in an electrical power distribution system is presented. Digsilent Power Factory software was used to model a section of a 66 kV power system. Fault incidents were instantiated on the model. The signal obtained from fault incidences were subsequently fed as input to discrete wavlet transform in order to obtained fault features and subsequently the features were then used as inputs for a support vector machine (SVM) and artificial neural network (ANN) for fault classification and detection. In addition, a Gaussian Process Regression (GPR) technique was employed for estimation of fault locations along the distribution line. Fault detection, classification and location estimation scheme were developed in MATLAB. The method showed that most faults on electric power distribution network can be classified with a good accuracy and minimum fault estimation error. The method is further validated on a real world power system. A hybrid method is thus proposed for detection, classification and estimation of fault location in a distribution network.

The global energy sectors have been faced with a number of challenges of the past years. These challenges include; the electrical stability, the environmental concerns over using fossil fuel to generate electricity and the ever-growing cost of electricity. To increase the capacity of energy accessibility and clean energy production, renewable energy distributed generations (REDGs) have been used as an alternative sources of electricity generation. The integration of REDGs into the power grid has been a common practice and has contributed effectively to energy availability and improvement of the environmental coverage. However, there are technical and financial problems that are accompanied by integrating REDGs into the existing power grid. To minimize the impact of these problems, mathematical optimization tools must be used. In this paper, we present a review of numerous techniques proposed to minimize the technical effects of integrating REDGs into the power grid. The paper also presents possible recommendations which may be focused on by further future research.

This paper presents aerosol studies carried out in the capital city of Botswana. The use of backward air mass trajectories has shown that elevated concentrations of fine particles correspond to situations when the air mass originated in the Pretoria and Johannesburg region in South Africa. This illustrated transboundary pollution and also served as an example of long distance transport. The results also show a significant contribution of a local generated pollution. The data has been subjected to principal component analysis (PCA) in order to make source apportionment.

The challenge with high impedance faults has been a subject of concern for many decades and has proven to a complex problem. This problem has proven to be complex. This complexity brings even more challenges to the utility companies as a high impedance fault poses dangers to the community. It is for such reasons that optimal detection of high impedance faults is achieved with higher reliability. In this paper, a scheme for HIF detection on power distribution systems is proposed. The scheme includes a decomposition section using the wavelet packet transform. The wavelet packet transform is used to decompose a signal into its coefficients; after the determination of coefficients, statistical features (energy and entropy) are used. The signal is decomposed at level 4. Furthermore, particle swarm optimization is used to determine the parameters for a support vector machine. The support vector machine scheme is used to classify high impedance faults from other power system operations. The effectiveness of the J48 and fuzzy logic reasoning classifier has been further investigated. Based on simulation results support vector machine has proved to be more effective with 99.6% accuracy when using particle swarm optimization and 92% accuracy without particle swarm optimization. The J48 and fuzzy logic reasoning produced 88.3% and 87.5% classification accuracy respectively. A practical experiment has been conducted to validate the proposed method and support vector machine has showed impressive results.

This paper presents a technique for high impedance fault diagnosis in a power distribution network. A segment of a 22 kV Eskom real power system is modelled in Power World Software. Various cases including High Impedance Faults (HIFs), normal condition, load switching and capacitor switching are subsequently investigated through the study of the model. Discrete Wavelet Transform (DWT) is used as a feature extraction technique. The extracted features are subsequently fed into an Artificial Neural Network (ANN) and Gaussian Process Regression (GPR) Schemes in order to effectively diagnose HIFs. ANN is used for fault classification and detection from other power system conditions. Furthermore, the GPR scheme is used to estimate the location of the fault. Thus a hybrid technique which comprises of DWT-ANN-GPR is proposed for HIF diagnosis. The results obtained shows that the scheme is fairly accurate and has minimum estimation error for fault location.

Power systems reliability assessment allows power utility companies to make informative decision with regards to the balance between quality of supply and economical investment. Power systems indices enable utilities to measure quality and performance of a distribution network. As it can be expected, the level of performance of power system network that must be maintained is closely linked to revenue. A network with poor performance, can only expect low revenue. In this paper we investigate the performance of Moehoek 22 kV overhead feeder line. Moehoek power line is situated in South Africa in the area of the Mpumalanga Province; the distribution indices of the line are calculated. The study includes the determination of causes of failure, analysis of the outages and a process that should be put in place in order to improve the performance of a feeder.

The detection of high impedance faults (HIFs) on a power distribution system has been a subject of concern for many decades. This poses a very unique challenge to the protection engineers, as it seems to be invisible to be detected by conventional protection schemes. The major concern about HIFs is that they pose a safety risk, as these faults are associated with arcing which may be dangerous for the surroundings. In this work, we propose a technique, which uses feature extraction, classification and a locating algorithm. Discrete wavelet transform (DWT) is used to extract meaningful information, support vector machine (SVM) is used as a classifier and a support vector regression (SVR) scheme is used as a fault location estimator. The technique is tested on a network of a power utility.

A High Impedance Fault (HIF) is a long standing complex type of a fault, because of its distinctive nature. The major concern with HIF is the potential risk it poses to human lives, because of its association with arcing. From the point of safety and reliability HIF is still a challenge for protection engineers. In this paper a HIF model is adopted and the combinations of wavelet transform and support vector machine is presented to detect a HIF. Discrete Wavelet Transform (DWT) is used as a feature extractor to extract useful information from the distorted HIF current signal. For classification purposes Support Vector Machine (SVM) is used to distinguish HIF from other events such as normal load, capacitor switching, and load switching. An Eskom network is studied and modelled in MATLAB/SIMULINK. The waveform results are fed into a DWT tool for feature extraction and the results from DWT are used to train the SVM for classification and ultimately detecting HIF.

Absorption of sunlight by sub-micron particles is an important factor in calculations of the radiation balance of the earth and thus in climate modelling. Carbon-containing particles are generally considered as the most important in this respect. Major sources of these particles are generally considered to be bio-mass burning and vehicle exhaust. In order to characterise size fractionated particulate matter in a rural village in Botswana with respect to light absorption and elemental content experiments were performed, in which simultaneous sampling was made with a dichotomous impactor and a laboratory-made sampler, made compatible with black carbon analysis by reflectometry. The dichotomous impactor was equipped with Teflon filters and the other sampler with glass fibre filters. Energy dispersive X-ray fluorescence was used for elemental analysis of both kinds of filters. It appeared that Teflon filters were the most suitable for the combination of mass-, elemental- and black carbon measurements. The black carbon content in coarse (2.5–10 μm) and fine (

The mobility of different elements in the environment is an important factor to consider when assessing the risk of pollution. Airborne particles of micrometre or smaller sizes are known to be transported over vast distances, and may therefore affect regions far away from their sources. As a complement to the more traditional element analysis of fine-particle aerosols, the possibility of obtaining information on the element contents in particle fractions of different environmental mobilities and/or character was explored. A four-stage sequential leaching procedure was tested and applied to establish the distribution of metals in four different fine-particle fractions, namely (1) the fraction considered to be mobile in the environment, (2) the fraction bound to carbonates and oxides, (3) the fraction bound to organic matter, and (4) the fraction bound to silicates. Subtraction of energy-dispersive x-ray fluorescence (XRF) spectra was used to evaluate the element concentrations in the filters. For validation, total-reflection XRF was used in a few cases for analysis of leachates. The methods were applied to airborne particles collected in two locations in Botswana. In both locations the K, Ca, Mn, Zn, Br and Pb compounds were found to occur extensively in the environmentally mobile fraction, whereas Ti and Fe were found to dominate in the silicate fraction. Copyright © 2002 John Wiley & Sons, Ltd.

Airborne particles have been collected using a dichotomous virtual impactor at three different locations in the eastern part of Botswana: Serowe, Selibe-Phikwe, and Francistown. The particles were separated into two fractions (fine and coarse). Sampling at the three locations was done consecutively during the months of July and August, which are usually dry and stable. The sampling time for each sample was 12 hours during the day. For elemental composition, energy-dispersive x-ray fluorescence technique was used. Correlations and principal component analysis with varimax rotation were used to identify major sources of aerosol particles. In all the three places, soil was found to be the main source of aerosol particles. A copper-nickel mine and smelter at Selibe-Phikwe was found to be not only a source of copper and nickel particles in Selibe-Phikwe but also a source of these particles in far places like Serowe. In Selibe-Phikwe and Francistown, car exhaust was found to be the major source of fine particles of lead and bromine.

This paper presents a hybrid technique using a signal processing method and an intelligent scheme for power system fault diagnostics. The protection of a power line is critical for maintaining a sustainable power supply. Moreover, electrical protection schemes are required to distinguishes faults accordingly. In the present work, the discrete wavelet transform method is utilized to break down the fault current signal into sub-signal bands. Furthermore, the neural network (NN) algorithm scheme is applied to diagnose different faults which may occur in power grid network. The performance of the protection scheme relies mostly on the ability of the scheme to accurately classify fault. In the present work, the particle swarm optimization (PSO) method is implemented to evaluate the input parameters of the NN classifier. The presented results show that the ANN classified the faults with an accuracy margin of 99%.