Power System Load Modeling Based On Genetic Programming

The digital simulation based on the mathematical model of power system is themain method of research and analysis of power system behavior. Obviously, themathematical model of generator and its control system, transmission network, powerload components is the basis of power system digital simulation. Therefore, study themathematical model of the equipment has the vital significance. Relative to themodeling technology of generator and its control system and power transmissionnetwork, Power load has the characteristics of time-varying, complexity andrandomicity, which causes the difficulty in modeling, and it never, has a majorbreakthrough. This is a big problem in the field of electric power research.Nevertheless, considering the important influence of load models on power systemsimulation, the study of its exploration never interrupted.The common method to modeling includes two aspects of content, one is thedetermination of model structure, and the other is a model parameter optimization.Among them, the model structure reflect the shape of the model, reflects the loadcharacteristic, also is the core content of the whole modeling work. Relatively, thestudy of model parameter optimization method is mature. In previous modelingpractice, due to lack of identification of model structure theory guidance is usually themodeler artificial selection model based on experience. This method with muchsubjectivity, in the face of tens of thousands of possible combination model structurescheme, selection optimization by artificial method of model structure is difficult.Genetic Programming (GP) is a new evolutionary algorithm based on geneticalgorithm, which has self-adaptive, self-organizing, self-learning and otheradvantages. As a kind of automatic programming technology, it tries to study thecomputer how to solve the problem according to the objective environmentautomatically without human intervention. It adopts the principle of natural selectionand evolution of biological theory, expressed in hierarchical tree structure, startingfrom the initial group of randomly generated, with fitness to measure the quality of the individual, the individual for each generation is used such as reproduction,crossover and mutation operation, after the evolution of several generations to obtainthe optimal solution of a given problem, namely the optimal individual fitness.Power system load modeling based on genetic programming need not todetermine the specific model structure in advance like the traditional method, it canaccord the input and output data directly evolved function relationship betweenvariables. Overcome the limitations of traditional modeling method, and be able toone-off determine the structure and parameters of function at the same time, makesthe model generation process tend to be more intelligent, automation, solve thelong-standing problems of automatic identification model structure in the powersystem.Finally, according to the load data measured in dynamic model experiment, byusing the genetic programming to achieve the model structure of automaticidentification, through the simulation contrast found that the mathematical modelestablished by this method has high fitting precision. Thereby, verify the feasibility ofthe power system load modeling by genetic programming.