Research Of Power System Load Model Based On Improved Genetic Programming

Power system digital simulation is one of scientific tools to analyze and research power system. It has the advantages of economy and convenient. Power system is comprised of generator, excitation regulation and power load. The accuracy of element models will have influence on result of Power system digital simulation.Element models play an important role in the operation planning of Power System,even in some situation, it will get the opposite characteristics.The load modeling have to fully use the knowledge and experience of the modeling people at present, because the load model have the feature of dispersivity randomicity and multiformity, the result is inaccuracy. With development of power system analysis, the load model have more challenge, we should pay more attention to it and intensive study.The methods of load modeling mainly includes component-based method,post-disturbance simulation method and measurement-based modeling method. The measurement-based modeling method have the advantages of simple and less knowledge dependency. But the measured data and model structure should be known when we identify the parameter.Compared with the traditional load modeling which need to determine the specific model structure at first time, Power system load modeling based on genetic programming need not to determine the specific model structure in advance like the traditional method, it can according to the input and output data directly create the fitting function, and find one have the best fitness. The Genetic Programming reference from natural selection and genetic mechanism. It’s based on the basic idea of Genetic Algorithm but using the hierarchical tree-structure. Each individual has the fitness to evaluate its performance. According to the principle of survival of the fittest,the next generation created by crossover variation and replication. Finally, The individual which has the highest fitness is the result.In this paper, We use the improved method of GP to static and dynamic loadmodeling respectively. Firstly, The precision and efficiency has been improved by optimize the adaptation of the calculation process. Secondly, The effectiveness and feasibility of the static load model based on improved genetic programming is verified by comparing with the traditional load modeling method. Finally, The dynamic load model shows that the dynamic modeling based on improved genetic programming is realizable.