Application Of Improved Gene Expression Programming In Load Modeling

With the rapid development of computers, digital simulation technology is widely used in power systems. Power system simulation accuracy has a close relationship with power generators, control systems, transmission network and power load model. Since the model of generator and control systems are more accurate,while the load model is still rough that hinders the power system simulation calculation further enhance. However, because of power load has randomness 、time-varying、dispersible features, it is difficult to establish an accurate load model.But it is possible to establish a more realistic and practical engineering load model to make it more rational and effective, and it is the goal of many researchers and engineers efforts.With the accelerated process of China’s power grid construction, promotion of new energy technologies, diversification of energy mix, the increasing complexity of grid, the demand of simulation accuracy is increasing. Based on past experience, the simulation results are largely influenced by the load mode. Therefore, the traditional method of determining load model in accordance with the experience is no longer suitable for constantly updating and developing power system. The introduction of new methods and algorithms is the key to solve the problem of load modeling, and it is also the development trend of power load modeling.GEP(Gene Expression Programming) is a new algorithm derived from the genetic algorithm family, this algorithm combines the advantages of genetic programming(Gene Programming) and genetic algorithm(Genetic Algorithm).Comparing with traditional genetic algorithm, GEP has advantage of 2 to 4 orders of magnitude, so GEP algorithm has a good performance in data mining and searching model that other algorithms do not have. However, with the in-depth study of GEP algorithm, we find that GEP is lack of diversity and GEP has "premature" phenomenon. In order to prevent these phenomenon, it is needed to improve the GEPalgorithm, the main improvements include: initial population, genetic operators,constant treatment etc. The improved GEP algorithm can increase the diversity of the population of individuals in large part, so that it can prevent "premature" phenomenon.The using of GEP algorithm in the load modeling, it does not need to determine the form of the model structure in advance, and it does not need to use the method of least squares algorithm optimization to identify model parameters. The algorithm can search function between data automatically according to the input and output data, so it can achieve modeling automatically. In terms of programming, this article choose to write Java program in eclipse platform, then according to the results of the program,fitting contrast simulation in Matlab software, finally we get feasibility of improved GEP.