Study On Complex Problem Optimization Algorithms Based On Swarm Intelligence And Its Applications

"Innovation, coordination, green, open and sharing" becomes the new idea which guides the scientific development of energy and electric power industry in the China's 13th national five-year plan and even in the medium and long term. A lot of technical problems need to be innovated and optimized proliferate during the process of promoting innovative and green development, and these technical problems become large and complex with the increasement of electric system scale and technical requirement. Based on two scientific research projects on construction and operation of electric power system, the maximum power point tracking (MPPT) under complex environmental conditions of large-scale photovoltaic system, as well as the dynamic optimization of operation and maintenance task of electric meter (OMT-EM) are discussed in this study. In terms of the above two problems, a kind of complex optimizing problem with the characteristics of large-scale, multi-modal and variable coupling is described and researched, and models of these problems based on swarm intelligence are also built in this study. Besides, the swarm intelligent methods are researched in order to overcome different difficulties of the problems, and are utilized for the real world engineering applications. The main contents and results of this study are as follows:The swarm intelligent algorithms often fall to find the global optimum when optimizing multi-modal problems, this weakness is caused by the restriction of local optimum and is always described as the "premature phenomenon". The reason of "premature phenomenon" in PSO is discussed, and some strategies to overcome the "premature phenomenon", as well as the human-simulated PSO (HSPSO) and its further improved algorithm HSPSO-FI are proposed by increasing the intelligent level of each particle. The proposed algorithms introduce some properties of human brain in an attempt to improve the abilities of particles to get rid of the restriction of local optimum. Experimental results show the efficacy of the human simulated properties, and particles can overcome the effect of local optimum, as a result, performance of the algorithm is improved significantly.In terms of the complex large-scale optimizing problem, complexity of the problem increases exponentially with the increasing of variables. As a result, most of the ordinary optimizing algorithms may lose their efficacy when solving these large-scale problems because of the "curse of dimensionality". Computing complexity of the problem becomes extremely complex especially when the large-scale problems are combined with the characteristics of variable coupling. In order to expand the application area of swarm intelligence and improve its performance in large-scale problems, a common adaptive multi-context cooperatively coevolving (AM-CC) framework is proposed in this study, and the AM-CCPSO which is formed by applying the AM-CC for PSO, is also discussed. Experimental results show that the AM-CC framework performs well on separable and non-separable large-scale problems with up to 1000 dimensions. The AM-CC framework provides a common and effective solution of solving large-scale problems with swarm intelligence, especially the variable coupling large-scale problems.Based on the theoretical research, modeling and computing method of a typical complex optimizing problem in electric power system construction is researched. Specifically, the swarm intelligence based method in solving "hot spot effect" and MPPT problem of large-scale photovoltaic system under complex environmental conditions is discussed. The "hot spot effect" affects the stable operating of photovoltaic system under partial shading environmental conditions, and the existing methods also have the weaknesses of additional loss in output power, high cost and difficulty of application in large-scale system. A novel topological structure based on the photovoltaic module control device (PMCD) and branch voltage stabilization device (BVSD) is proposed to solve this problem, and make the photovoltaic module level (PVM-level) and minimum control unit level (MCU-level) MPPT possible. Besides, the large-scale MPPT model described as a large-scale global optimization problem is also built, and the algorithms proposed in this study are utilized to solve this MPPT problem. Experimental results show the efficacy of the topological structure, MPPT model and swarm intelligence based algorithms proposed in this study, as a result, each photovoltaic module (or minimum control unit) works on its own MPP under complex environmental conditions, "hot spot effect" is overcome effectively and output power of the large-scale photovoltaic system is maximized.Modeling and computing method of a typical complex optimizing problem in operation of electric power system is also researched. Specifically, in terms of the dynamic optimization problem of OMT-EM, the requirements of power grid company in its daily management are discussed, and the dynamic optimization model based on swarm intelligence is also built in an attempt to response to the dynamic requirements, including the number of tasks, real time traffic, attribute and number of the workers, preference of the decision maker. The swarm intelligent algorithms proposed in this study are utilized to solve this problem, experimental results show that the proposed model and algorithm can response to the requirements and real time changes effectively, as a result, efficiency of operation and maintenance task of power grid company can be improved significantly due to the real-time and dynamic optimization of OMT-EM with high dimensionality.