Optimal least squares support vector machines parameter selection in predicting the output of distributed generation
- 1 December 2014
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 152-157
Abstract
This paper presents a novel technique to optimise the least squares support vector machines (LS-SVM) parameters in predicting the output of Distributed Generation (DG) in a distribution system. In LS-SVM, the accuracy of the prediction is depends on the selection of kernel parameters. Unfortunately, there is no systematic methodology for selection of their optimal values. Therefore, a novel hybrid Quantum-Inspired Evolutionary Programming - Least Squares Support Vector Machine (QIEP-SVM) is developed for accurate prediction. In QIEP-SVM, Quantum-Inspired Evolutionary Programming (QIEP) is developed to optimise selected parameters for the LS-SVM which are gamma and sigma. QIEP is combining Evolutionary Programming (EP) with quantum mechanics concepts such as interference and superposition in order to enhance classical Evolutionary Programming (EP). The optimal output of DG is first generated using multiobjective Quantum-Inspired Evolutionary Programming (QIEP) at various loading condition according to 24-hours load profile. The data from the simulations are then used as the inputs to the Least-Squares Support Vector Machine (LS-SVM). There are three inputs which are active load (MW), reactive load (MVAR) and minimum voltage (p.u). Whereas, there are five outputs that represents the output of DG at five buses. The objective function for the optimisation process is to minimise the mean square error between predicted and targeted output. The performance of QIEP-SVM is then compared to classical LS-SVM using cross-validation technique and hybrid Artificial Neural Network-Quantum-Inspired Evolutionary Programming (QIEP-ANN). The results of QIEP-SVM model had shown better prediction performance as compared to classical LS-SVM and QIEP-ANN. All simulations in this study were carried out using IEEE 69-bus distribution test system.Keywords
This publication has 11 references indexed in Scilit:
- Quantum-Inspired Evolutionary Programming-Artificial Neural Network for prediction of undervoltage load shedding2013
- Using Hybrid Genetic-Particle Swarm (HGPSO) Optimization to Distributed Generation Allocation to Improve Steady State Voltage Stability of Distribution NetworksJournal of Modeling, Simulation, Identification, and Control, 2013
- Microwave Characterization Using Ridge Polynomial Neural Networks and Least-Square Support Vector MachinesIEEE Transactions on Magnetics, 2010
- Using support vector machines for determining voltage unstable areas in power systems2008
- Optimal Placement and Sizing Procedure for PV Systems on Radial Distribution Systems2006
- State Estimation for Delayed Neural NetworksIEEE Transactions on Neural Networks, 2005
- Classification of power system stability using support vector machines2005
- Analytical Approaches for Optimal Placement of Distributed Generation Sources in Power SystemsIEEE Transactions on Power Systems, 2004
- Application of an inverse input/output mapped ANN as a power system stabilizerIEEE Transactions on Energy Conversion, 1994
- Selecting a classification method by cross-validationMachine Learning, 1993