| Abstract: |
Many techniques for solving reactive power planning problem have been presented. In the past, numerous methods such as successive ordinary load flow, optimal power flow, and sensitivity techniques have been proposed to solve the problem. After that compact methods such as linear programming, nonlinear programming, integer programming, and quadratic programming are applied for solving it. Now some of the Artificial Intelligence (AI) teclmiques such as Fuzzy, and Artificial Neural Network (ANN), are applied for solving the reactive power planning problem. Most of the previous techniques is a single path search which means that local optimal may obtained instead of global optimal. Genetic algorithm which was applied in this research with its multi path search capabilities endorse its applicationIn this research, a two stage technique has been developed for solving Reactive Power Planning Problem. Initially, to reduce the computational time only a few numbers of buses are selected as candidates. The Modal analysis technique is applied to compute the smallest eigenvalues and the associated eigenvectors of the reduced Jacobian matrix of the system. The magnitude of the smallest eigenvalue gives a measure of how close thesystem is to the voltage collapse. Then, the participating factor can be used to identify the weakest buses in the system associated to the minimum eigenvalue. Buses with high bus participation factors are considered as candidate buses. Buses with low voltage levels are added to the set of candidate buses. As a second stage GA is applied to identify the location and size of the new V AR compensators.Single objective functions IS considered by applying the proposed technique to active power loss, voltage deviations, and V AR cost as the objective function. The results obtained from the previous cases leads to the use of multi-objective function in which active power loss, voltage deviations, and V AR cost are combined to form the objective function. The proposed technique was applied to multi-objective function under different weighting factors of voltage deviations term of the objective function. This technique is applied to IEEE 14 bus, IEEE 30 bus, and IEEE 57 bus systems and succeeded in detennining the optimal locations and sizes of the installed compensators, which reduced the cost of the active power loss, and enhance the voltage profile of the system.Saving time is the main advantage of applying the two stage teclmique. Applying such teclmique reduces the number of variables in the objective function and so, reduces the nwnber of generations required to obtain convergenceGenetic algorithms are very robust search and optimization technique, especially when little or no knowledge of search space are available.Fitness function is the heart and driving force of genetic algorithms. The function variables should be chosen precisely to fit the objective.The proper choice of population size is essential in genetic algorithms. A small population may result in a local convergence, while a large population will result in a slow convergence. To choose the suitable population size, several trials are needed.Genetic algorithms are very sensitive to weighting factor of voltage deviation term of the objective function. The sizes of compensators vary with any small change in weighting factor of voltage deviation.The two stage technique is applied to the IEEE 14_Bus System, IEEE 30_ Bus System, and IEEE 57 _Bus System and succeeded in fmding the global optimal sizes and locations of new compensators in each case. So, It is may be suitable to be applied to any Power System.The results of tIlis research show tile promise of the applied technique to the reactive power planning problem employing genetic algoritIuns. Some of the recommended future ideas that could be implemented in the same direction are listed hereunder.• The proposed technique could be applied to larger scale power systems. In tills case, it is recommended to use larger computer sets to get faster calculations.• The effect of hannonics in selecting the best locations and sizes of installed var compensators may be studied• Number of control variable may be extended to include the tap positions of power transfonners.• The best size of population in genetic algoritIun toolbox could be detennined according to the number of variables of the objective function.• The best value of the weighting factors of voltage deviation term in the objective flillction may be detennined in an accurate way. Different weighting factors for each bus can be used.
|
|
|