OPTIMAL REACTIVE POWER DISPATCH ON POWER SYSTEM USING HYBRID PARTICLE SWARM OPTIMIZATION

• Main Author: Kanata, Sabhan
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/49277
• Institution: Institut Teknologi Bandung

Increasing dependency on electricity will require companies to supply electricity, improve efficiency, quality, and safety when operating an electric power system. An increase in active power losses will have an impact on the reduced supply of power sent by the electricity company to consumers. That is, the active power generated by the power plant will simply disappear because it is considered unsold. In this case, the electricity supply company will suffer losses because it produces power at a large cost but does not get financial benefits from the sale of power. Besides, for electricity to provide good voltage quality to consumers, electrical energy supply companies must maintain a constant voltage, especially at the end of the line. The indicator can be seen by maintaining the voltage on the load bus so that it is always at a value equal to or near ideal (usually 1.0 p.u). For this performance to be carried out, it is necessary to optimize the control variables that affect the active power losses and the total voltage deviation on the load bus. This research in the scope of optimization is called optimization of reactive power and voltage regulation which is often known in the world of global research, namely Optimal Reactive Power Dispatch (ORPD). ORPD is a problem that has not been solved until now. The strategy is carried out by optimizing control variables that affect reactive power supply and voltage. The control variables referred to in this study are the magnitude of the output voltage of the generator, tap transformer, and reactive power compensators. This study proposes several methods for solving ORPD problems with singleobjective and multi-objective optimization. For single-objective, this study proposes 5 (five) algorithms, namely Hybrid Artificial Neural Network and Time-Varying Particle Swarm Optimization (HANNTVPSO), Time-Varying Particle Swarm Optimization (TVPSO), Genetic Algorithm (GA), Hybrid Time-Varying Particle Swarm Optimization and Genetic Algorithm (GA). HTVPSOGA) and Hybrid Genetic Algorithm and Time-Varying Particle Swarm Optimization (HGATVPSO). The initial value generation in the HANNTVPSO algorithm uses guided random values. The random value is vi derived from the ANN output. Whereas the other four algorithms use random initial value generation. The five algorithms are tested on a small scale power system namely the IEEE 14-bus power system. As for multi-objective optimization to minimize active power losses and total voltage deviation, this study proposes 7 (seven) algorithms, namely Multi-objective Particle Swarm Optimization (MOTVPSO), Multi-objective Non-dominated Sorting Genetic Algorithm - III (MONSGA- III), Multi-objective Hybrid Particle Swarm Optimization and Genetic Algorithm (MOHTVPSOGA), Multi-objective Ant Lion Optimization (MOALO), Multi-objective Dragonfly Algorithm (MODA), Multi-objective Grey-wolf Optimizer (MOGWO), and Multi-objective Multiverse Optimizer (MOMVO). All of the algorithms are tested on the IEEE 14- bus, 57-bus, and 118-bus power systems. For the comparison of the performance of the seven algorithms, the same algorithm parameter values are used. The complexity of MORPD problems in the IEEE-14 bus and 57- bus power system that involves a control variable consisting of a combination of continuous and discrete variables. Whereas the IEEE 118-bus power system considers all control variables to be continuous. The complexity of the IEEE 118-bus power system lies in the use of reactors and capacitor banks with a large number of control variables. For single-objective optimization, the HANNTVPSO algorithm is can improve the initial value of TVPSO which is 12,365 MW. The value is better when compared to the TVPSO algorithm when using the random initial generation of 13.44865 MW. But the initial guided generation has a weakness that is different weight values for each different case as well. This problem will make it difficult to determine the ANN output value as the initial value of TVPSO. So that the four algorithms in the second study used random initial generation. The four algorithms are TVPSO, GA, HTVPSOGA, and HGATVPSO. These algorithms are tested to solve ORPD problems by determining the output of the optimization based on statistical test criteria namely Best Objective Value (BOV), Worst Objective Value (WOV), and Mean Objective Value (MOV). The comparison results show that the best BOV value is generated by the GA algorithm. The best WOV and MOV values are generated by the HTVPSOGA and HGATVPSO algorithms. From the results of simulations in the 1st and 2nd studies, the five proposed algorithms can reduce active power losses better than previous studies based on references in this work. Besides, the two combination algorithms can find the optimal solution faster, namely the 4th iteration. As for multi-objective optimization in the 3rd and 4th studies, based on the statistical test criteria of the seven proposed algorithms, the MOHTVPSOGA algorithm has a very dominant contribution compared to other algorithm proposals and previous studies in this work. But the algorithm has a disadvantage of the amount of computational time used. Unlike the MOMVO algorithm which has efficient computing time. But the algorithm has a disadvantage of the quality of the vii solution. However, when the MOHTVPSOGA algorithm is tested on a 500 kV electrical system, the algorithm can produce quality solutions with a very significant contribution and time-efficient computing when compared to the central regulatory PT. PLN. This is because although the algorithm uses a maximum number of iterations and a small population, the algorithm is still able to make a better contribution. The results show that the MOHTVPSOGA algorithm has a very promising opportunity when applied to a real electricity system.