Swarm intelligence is a group of non-intelligent body that can communicate directly or indirectly to each other. This group of subjects can work together to solve problems and show the characteristics of intelligent behavior. Swarm intelligence algorithm is a heuristic search algorithm based on the group behavior to search for a given target. The optimization process reflects the characteristics of random, parallel and distributed. Therefore, when facing to the complex practical problem without centralized control and global model, it provides the possibility to find out solutions. Kennedy and Eberhart proposed PSO(Particle Swarm Optimization) algorithm in 1995. Due to its characteristics of simple calculation, easy implementation and less control parameters, it has attracted the attention of many scholars in related fields at home and abroad. Based on deep study on convergence behavior of individual particles of PSO and inspired by quantum physics, Sun J proposed QPSO(quantum-behaved particle swarm optimization) algorithm. Compared to PSO algorithm, QPSO algorithm has the characteristics of far less control parameters in practical applications, high speed of convergence and strong global search ability and other features. It is a major improvement branch of PSO algorithm.QPSO algorithm is studied and the mechanism of QPSO algorithm is analyzed, the improved QPSO algorithm with different search strategies which have different attraction points. For the dependence of QPSO on contraction expansion factor, it puts forward a parameter control strategy for reinforcement learning and adaptive learning. It analyzes the local search ability of algorithm and proposes four different search strategies of local search. Also, it studies the QPSO and its improved algorithm dispatch optimization in power system and application method in RZWDM2 model parameter optimization. Specific contents are as follows:(1) Research the backgrounds and developments of SI algorithms. The current research situations of two typical SI algorithms are introduced in detail, which are Ant Colony Optimization ACO(Ant Colony Algorithm) algorithm and PSO algorithm.(2) The leading mechanism of QPSO algorithm in the evolution process is analyzed, which adopts average optimal value of whole swarm as its global attraction point. We can find that the best individual of swarm playing a decisive role in the process of leading social development in intelligence swarm. Therefore, an evolution strategy is proposed, which adopts best individual in swarm as its global attraction point, and optimization ability of algorithm can be improved. The comparison is studied between global best individual and global average best point which are as its global attraction points. The decision-making mechanism of the intelligence swarm is analyzed and we find that individuals have different rights to participate in the decision-making process according to their fitness values, and then the paper proposes the nonlinear weight operator calculation in QPSO algorithm, which improves the solving ablility and optimization efficiency of QPSO algorithm.(3) In QPSO, the algorithm performance is strongly depending on the contraction expansion factor. A variety of parameters control strategies in the parameter control of evolutionary process simultaneously are applied. Reinforcement learning method is introduced in QPSO algorithm and it is improved with this method. What’s more, the paper analyzes the evolution learning mode of QPSO algorithm and points out that the algorithm has only studied for swarm particle itself. Due to lack of discussion of the optimization problem, especially in the process of evolution, different complex problems are not treated differently, it proposes adaptive learning QPSO algorithm and improves the performance of algorithm.(4) Local search ability is an important research part of swarm intelligence algorithm. According to different search characteristics, four search strategies which improve the local searching abilities are introduced in QPSO algorithm, and QPSO algorithm is proposed with local search strategy. In the QPSO algorithm, the information is shared by swarm particles through dimension information, evaluation is through whole information of one particle, which is easy to lose better dimension information of worse particles in swarm. For these phenomena, it introduces crossover operator, which can increase the probability of preserving dimension information and improve algorithm performance, and the QPSO algorithm is proposed with crossover operators.(5) The scheduling problems of active power in power system for power system optimization is analyzed and optimal power flow method is adopted to abstract set into a mathematical model. For the constraints of power system, it adopts penalty function approach in algorithm optimization process. QPSO algorithm and its improved algorithms are used in scheduling problems of power system. Meanwhile, it conducts simulation test for three typical unit systems. Through a comparative analysis of simulation results, it proves that improved QPSO algorithm could obtain better optimization result in scheduling prblems of power system.(6) For the parameter optimization of root stream model RZWQM2 in agricultural system, most of the scholars still use the traditional trial and error method. The pros and cons of model parameters are largely dependent on the user’s experience; they propose to use QPSO algorithm to optimize the model parameters. Root mean square deviation with the weight coefficients is used to design objective optimization function and they conduct classifications of objective function according to evaluation value of production, drainage flow and loss amount of NO3-N. It designs single-objective optimization function and multi-objective optimization function; the result shows that QPSO algorithm could be better applied to the optimization of the model parameters.Finally, the paper gives a summary of works and main study findings and puts forward a further study. |