Research On Particle Swarm Algorithm And Its Application In Optimization Dispatch Of Wind-Thermal Power System

Particle swarm optimization(PSO), as a typical optimizing method in swarm intelligence, has become an effective approach and research focus to solve all kinds of complex optimization problems. It is a simple and efficient global optimization algorithm. It has some advantages of implementation, parallel computing and less control parameters and is widely used in many fields. However, the existing PSO algorithm has some shortcomings. For instance, the convergence analysis is not enough complete, population size parameter is lack of adaptability and the algorithm is easily to present premature, stagnation and slow convergence speed. Therefore, based on the analysis of the current existing PSO algorithm, the convergence of dynamic interaction for PSO was studied. Dynamic particle population PSO algorithm, the optimal information sharing of PSO gravitational search algorithm and the parallel hybrid method of PSO with GSA were proposed in this paper. These proposed algorithms were applied to functions optimization, parameters identification and optimization scheduling problems of power systems and have significant theoretical significance for enriching the PSO algorithm. The main contribution of this dissertation are summarized as follow.1. For the convergence of PSO, the dynamic interaction memory characteristics of the PSO algorithm are considered in the evolution process. The convergence analysis of dynamic interactive operation under consideration has been proposed in terms of Z transform domain method in the linear system theory, and the reasonable range of parameters have been obtained. Moreover, particle position and velocity of the PSO algorithm in the process of evolution show certain regularity of volatility when the algorithm parameters satisfy certain conditions.The experimental simulations have been conducted based on some benchmark test functions, and the correct conclusions have been proved.2. In terms of population size parameter setting, dynamic particle population PSO algorithm was proposed based on the idea of adaptive evolution for number of biological species. The total number of population size is determined by population age model, and two strategies such as particle reproductive strategy and particle vanishing strategy are taken. The proposed algorithm converges the optimal solution by use of random functional theory. Results show that the proposed is superior to the fixed size and hard assigned particle size of PSO algorithm based on benchmark functions.3. For some shortcomings such as local optimum and slow convergence rate of gravitational search algorithm(GSA), the improved GSA(IGSA) by use of individual optimal memory strategy of PSO and infinite collapses mechanism of chaos was proposed. The individual trajectory stability and global convergence of IGSA have been performed by linear system theory and stochastic process. The simulation experimental results show that the proposed algorithm possesses high accuracy and fast convergence rate based on benchmark functions of CEC2005.4. For some shortcomings such as similarity iterative process and algorithm structure of PSO and GSA, hybrid PSO and GSA(HPOS-GSA) was proposed by use of parallel hybridization design. The acceleration equation of GSA is directly embedded into the update equation of PSO algorithm, and the proposed HPSO-GSA has been presented to enrich the information search behaviors of the algorithm. Based on the Markov chain of stochastic process theory, the global convergence analysis of HPSO-GSA was conducted. The algorithm is carried out for adaptive IIR filter parameters optimal design and results show that the HPSO-GSA can better balance the ability of global detection and local searching.5. Finally, the economic and emission dispatch modeling of conventional thermal power units with various practical constraints and wind-thermal power system with wind farms were established. The proposed parallel hybrid approach of PSO and GSA was carried out to solve the above-mentioned modelling. Results of classical power system cases show that the proposed algorithm has better optimizing performance compared with other approaches. The economic cost is saved and the pollution emission is reduced for the tested power systems.