Research On Classification And Synthesis Methods Of Load Characteristics For Load Modeling

Digital simulation is an important means for planning, design, analysis and control of power system, and the accuracy of the load model has a significant influence on the results of the simulation. Along with the advancement of Chinese regional power grid interconnection, the electric power system becomes more complex, accurate load model is particularly important to improve simulation precision of power system. At the same time, the load components of power system are constantly changing, all kinds of nonlinear load increase massively, a large number of distributed power are integrated into the distribution network, it is difficult for the traditional load model to describe the different load characteristics, accurately describing the load characteristic becomes more complex. Therefore, to strengthen the research of load characteristic and to establish the accurate load model reflecting the actual load characteristic have important theoretical value and practical significance. In this paper, classification and synthesis of dynamic characteristics of the load are discussed deeply. Main works are included as below:(1) This paper studies the dynamic characteristics of the load, especially the load classification of active load components. A load clustering algorithm based on lifting wavelet packet is used to analyze the IEEE-14 node distribution network with active load. Firstly, the wavelet packet transform is carried out on the load disturbance response data. And then extract the dynamic characteristics of the load based on the energy of wavelet transform coefficients and its distribution. Finally, complete the load classification by clustering algorithm.(2) In view of random time-varying problem of the load characteristics, a new method of load characteristics classification and synthesis is proposed based on Markov Monte Carlo in this paper. This method is based on extracting feature of dynamic load characteristics, and determines whether the variation of load sequence in the load categories has the Markov property. All load characteristics data are segmented by mean time, and the probability transfer matrix of Markov chain is established based on the maximum likelihood for each segment. The load characteristics data in each time period are clustered according to the corresponding digital features of matrix. And then calculate the transition probability matrix of load characteristics data in each time period. The load characteristics data are interpolated and extrapolated in each time period to generate the load characteristics sequence. Hidden Markov model is used to obtain more accurate load characteristics sequence in order to accomplish classification and synthesis.(3) Considering different influence factors on the dynamic load characteristics in active distribution network, the case study is set up to classify and compare the test results. And it simulates the time-varying load with different dynamic characteristics in a period of time. The case study has completed the feature extraction and classification of the dynamic characteristics data. The case study results show that the clustering results are consistent with the similarity among the samples dynamic characteristics, which further verifies the effectiveness of the lifting wavelet packet transform for the load dynamic characteristics. Simulation verifies the effectiveness of the load characteristics classification and synthesis methods which is based on Markov Monte Carlo.