| In recent years, our country has undergone a rapid development in power industry, and has stepped into the top list of the world in terms of the installed capacity and annual generated energy for power generation. For the large-scale enterprises like power industry, it will bring enormous economic efficiency to endeavor to achieve operation optimization. Therefore, the research on the optimization operation problems in electric power system is displaying increasing importance and urgency, which has enormous economic and project significance to improve and explore the advanced optimal dispatch models and algorithms with high convergence speed and adaptability on the basis of existing achievements.Recently, swarm intelligence computation and multi-agent technique are becoming an important topic in the field of artificial intelligence, and are successful applied in a lot of fields, which are indicating fairly great potential for development. This dissertation focuses on the theory of swarm intelligence computation and multi-agent technology, and employs them to make systematic and thorough studies on the optimal operation of electric power system. Some novel ideals and methods with the combination of the two kinds of technique have been put forward. The dissertation consists of the following six chapters.Firstly, on the basis of describing swarm intelligence computation and multi-agent technique, the current research situation of their applications in the field of electrical power system is summarized. Also, the potential significance of applying swarm intelligence computation and multi-agent technique in optimal operation of electrical power system is given.In the second chapter, based on clonal selection algorithm and standard particle swarm optimization algorithm, two improved swarm intelligence algorithms are proposed given the deficiency of the two optimization algorithms. By means of convergence analysis, it is known that if the control parameters of the two algorithms are chosen appropriately, they can obtain remarkable improvement in computational speed, precision and robustness. Through the test for a series of typical standard functions, the proposed algorithms demonstrate high quality solutions and the high convergence speed.The third chapter applies separately the improved immune algorithm (ⅡA) to the optimal power flow (OPF) and the improved particle swarm optimization (IPSO) algorithm to power system unit commitment problem. In solving OPF problem, the ⅡAwith incorporation of non-stationary multi-stage assignment penalty function can significantly improve the convergence and gain more accurate values. While in the process of handling unit commitment problem, the zero-one variables are relaxed at the first step and then unit commitment problem is transformed into a non-linear programming problem with continuous variables by penalty function, making it suitable for using the proposed optimization algorithm. For different optimal dispatch problems, both of the two kinds of algorithms have shown stronger optimization ability and practical value.The fourth chapter presents a solution to reactive power dispatch problem with a novel multi-agent systems based on particle swarm optimization algorithm (MAPSO). This optimal computation approach integrates multi-agent system (MAS) and particle swarm optimization algorithm (PSO). An agent in MAPSO represents a particle of PSO and a candidate solution to the optimization problem. All agents live in a lattice-like environment. Each agent competes and cooperates with its neighbors, and learns from its own knowledge so as to converge on the global optimal solution quickly and accurately. Meanwhile a simple and practical "cutting" method is put forward to solve the problem of discrete variables in reactive power dispatch problem without influence on optimization process. Analyses of two different scale power systems show that the methodology has great heuristic significance in solving power system optimization operation problem with highly complex constraints.In the fifth chapter, the dissertation introduces a cooperative evolutionary theory, and puts forward multi-agent computation system based on cooperative evolution to solve the large-scale electric power system optimization operation problem, which is decomposed into a series of interactive sub-systems. It accomplishes the whole evolution and finally acquires the solution to the optimization problem through mutual coordination and cooperative evolution among these subsystems. During the process of individual representative selection in cooperative evolution, a hybrid individual representative selection strategy is proposed which fully considers the advantages of the optimum selection method and the random selection method. The multi-agent computation system applied for optimal reactive power dispatch is evaluated on four various scale power systems. Simulation results show that higher quality solutions are obtained in a shorter time, and the proposed computation system with a hybrid strategy is very suitable for solving large-scale power system optimization operation problems.At last, the dissertation further studies the multi-period optimization operation problem — optimal dynamic reactive power dispatch. Inspired by the multi-agent technique used to solve the workshop production dispatch problem, a multi-agent system is brought forward in which dynamic reactive power optimization problem is regarded as a workshop dispatch problem, reactive power optimization problem of each period is regarded as a production task, and the allowable operating times of the control devices are regarded as resources waiting for assignment. Under the guidance of the strategy of switching time of control devices, the methodology applies contract network to the negotiation mechanism of each agent, which solves the conflict among agents, and the reasonable solution is obtained finally. The analysis of an actual power system shows that this methodology is suitable for solving dynamic reactive power optimization problem.
|
PDF Full Text Request