Zuhaina Zakaria

Application of Fuzzy Clustering to Determine Electricity Consumers' Load Profiles

In a regulated environment, load profiles have been employed to provide information for forecasting, system planning and demand side planning. However, in the deregulated environment, consumers can purchase electricity from any provider regardless of size and location. As a result, load profiles have become more significant. The determination of customer load profile may facilitate utility companies with better marketing strategies and improved efficiency in operating the current facilities. This paper examined the capability of fuzzy clustering to determine consumers load profiles on the basis of their electricity behaviour. Two techniques in fuzzy clustering namely, fuzzy relation and fuzzy c-means (FCM) were employed in this work. The load data used in this work are from actual measurements from different feeders derived from a distribution network. Cluster validity indices will be used to determine the optimum clusters. The performance of each algorithm will be evaluated by employing adequacy indices i.e. mean index adequacy (MIA) and clustering dispersion indicator (GDI).

Quantum-Inspired Evolutionary Programming-Artificial Neural Network for prediction of undervoltage load shedding

This paper presents new intelligent-based technique namely Quantum-Inspired Evolutionary Programming-Artificial Neural Network (QIEP-ANN) to predict the amount of load to be shed in a distribution systems during undervoltage load shedding. The proposed technique is applied to two hidden layers feedforward neural network with back propagation. The inputs to the ANN are the load buses and the minimum voltage while the outputs are the amount of load shedding. ANN is trained to perform a particular function by adjusting the values of the connections (weights) between elements, so that a particular input leads to a specific target output. The network is trained based on a comparison of the output and the target, until the network output matches the target. The parameters of ANN are optimally selected using Quantum-Inspired Evolutionary Programming (QIEP) optimization technique for accurate prediction. The QIEP-ANN is developed to search for the optimal training parameters such as number of neurons in hidden layers, the learning rate and the momentum rate. This method has been tested on IEEE 69-bus distribution test systems. The results show better prediction performance in terms of mean square error (MSE) and coefficients of determination (R2) as compared to classical ANN.

Determination of fuzziness parameter in load profiling via Fuzzy C-Means

Load profiling has become an important issue in power industry and has gain more attention from utility company worldwide due to deregulation and liberalization. A lot of work had been done to obtain a method to determine typical load profiles (TLPs) of electricity consumers. Load profiles represents consumers electricity consumption pattern and provide useful data to both consumer and electricity provider. This paper presents the TLPs determination through clustering technique by using Fuzzy C-Means (FCM) algorithm. Two of the most important parameters in FCM are fuzziness parameter, m and optimal number of cluster, c. This paper shows the determination of the suitable fuzziness parameter through observation of experimental result of the cluster validity indexes value. Cluster validity indexes were used to determine c. Three cluster validity indexes were discussed in this paper. They are Xie-Beni index, Non-fuzzy index and Davies-Bouldin index. Objectives of this paper are to obtain groups of TLPs by using FCM clustering and to determine the suitable value of the fuzziness parameter, m. The data used in this project are obtained from Tenaga Nasional Berhad (TNB).

Consumer Load Profiling using Fuzzy Clustering and Statistical Approach

Load profiling present useful tool for onitoring typical load shape for a group of customers, which can be performed from past or current day data. In a deregulated energy environment, consumers can purchase electricity from any provider regardless of size and location. As a result, there is a growing interest in understanding the nature of variations in consumers consumption. This information can be used to facilitate electricity supplier in their marketing strategy. Many techniques for load profiling have been reported in the past. The techniques include applications of statistics, unsupervised clustering technique and methods based on frequency domain approach. This paper compares the application of fuzzy clustering with statistical method in load profiling. K-means has been chosen as the statistical approach employed in this study. These two approaches have the same objectives i.e. to recognise similarities, clusters and classify the individual load profiles of different customers to one of the identified categories. The paper evaluates the performance of each method and discusses the strength and weaknesses of both approaches based on the simulated results.

FastICA techniques for load profiles estimation

The objective of this research is to employ Independent Component Analysis (ICA) for electrical load profiles estimation to ensure proper power usage measurement of the customers. ICA technique is able to separate the mixed signal into their source signals. Using this method, the load profiles on feeder distribution can be estimated without any knowledge of the network topology and electrical parameters. In addition, a real-time load profiles on feeder distribution can be established instead of load modeling technique by using incoming distribution feeder data profiles. The estimation quality is verified by using error measures of the load profiles. Simulation results and errors of estimations are discussed in this paper.

Multiobjective quantum-inspired evolutionary programming for optimal load shedding

This paper present new technique namely Quantum-Inspired Evolutionary Programming (QIEP) to determine the optimal location and optimal amount of load to be shed for undervoltage load shedding schemes. This approach is based on the concept of quantum mechanics in the Evolutionary Programming (EP). Quantum-Inspired is implemented according to three levels defined by quantum individuals, quantum groups and quantum universe in order to improve the speed of the algorithm. The QIEP is employed to search for the best location and amount of load to be shed based on multiobjective functions which are power loss minimisation, voltage profile improvement and power interruption cost minimisation. The effectiveness of multiobjective QIEP optimisation technique is illustrated in IEEE 33-bus distribution test system, IEEE 69-bus distribution test system and 141-bus distribution system. The results were also compared with other techniques in terms of fitness values and computation time.

Economic load dispatch via an improved Bacterial Foraging Optimization

An economic dispatch is one of the most important research areas in power system. This is because an optimal power dispatch will contribute to a sound power system load management. As a result, the solution can ease the cost of fuel without ignoring the power operation constraints and system losses. Adaptive Tumble Bacterial Foraging Optimization (ATBFO) and Adaptive Mutation Bacterial Foraging Optimization (AMBFO), which are originally from basic Bacterial Foraging Optimization (BFO), are considered as alternative algorithms to minimize fuel cost. Thus, the two techniques are compared under the same parameters to determine the best optimal result. After several analyses on the results obtained, it was found that ATBFO outperformed AMBFO. Both of these adaptive bacterial foraging optimization techniques are tested on 26 bus Reliability test system using MATLAB R2009b on MS Window 7.

Bacterial foraging optimization algorithm for load shedding

This paper presents an optimization technique for optimal load shedding in power system. The technique is known as Bacterial Foraging Optimization Algorithm (BFAO). Load shedding is done by removing a certain amount of loads at appointed locations of a bus system. By doing so, the stability of the system can be improved, as well as the total power losses and total costs of shedded loads. The objective functions of total power losses, voltage stability index values and also total cost of shedded loads are used in determining the optimal load shedding in that particular system. In this research, the technique is implemented into IEEE 30-bus bus system. Simulations of BFAO proved that it can give a better result when compared to the base case values of total power losses and voltage stability index values of that particular bus system.

Optimal Reactive Power Planning for load margin improvement using Multi Agent Immune EP

Systems loadability is very much depends on the reactive power support in the network. Lack of reactive power support may results in a serious problem to occur i.e. voltage collapse. This paper presents a method to increase the systems loadability and hence the voltage stability margin of a system by utilizing the Optimal Reactive Power Planning (ORPP). A newly developed optimization technique; namely Multiagent Immune Evolutionary Programming (MAIEP) is introduced to obtain the optimal solution to the problem. The concept of MAIEP is developed based on the combination of Multiagent System, Artificial Immune System and Evolutionary Programming. The proposed MAIEP based ORPP was tested on the IEEE-26 reliability test system in order to realize its performance. The results obtained from the proposed ORPP using MAIEP has successfully improved the load margin and at the same time the total system losses and cost of generation were reduced.

Classification of transient in power system using support vector machine

In this paper, application of SVM to classify disturbances in power quality is discussed. Power system transient can pose a serious threat to the reliability of power system apparatus and sensitive loads. There are numerous causes of power system transient namely short circuits, capacitor bank switching, switching of large inductive loads that include motors and transformers as well as lightning. Firstly, an IEEE 30 bus system is modeled using the PSCAD software to generate the different type of transient data caused by capacitor switching and lightning. Feature extraction is performed using wavelet technique. Next, the wavelet coefficients specifically the minimum and maximum values of the wavelet energy served as inputs for the SVM for classification purpose. Initial results showed that SVM is capable to classify the transient source with Radial Basis Function (RBF) as the kernel