Study On The Parameter Identification And Classification Algorithm Of Electric Load Model

With the constant expansion and more complex structure of power grid, people gradually realized the importance of electric load model in power system analysis, operation and control. Since measurement-based modeling approach is direct, realistic, practical, it was quickly favored by most researchers. But it also has some problems, such as riding index of parameters and practicability of load model. Based on the above issues, the parameter identification and load model classification are researched.The two main problem of measurement-based load modeling is model structure based on measured load characteristics data and the model parameters identification. First, three existed load models are introduced, and then point out that either the static model or dynamic model is not a good interpretation of the measured load. Finally, the TVA synthesis load model is selected, and the corresponding parameters are identified.Identification algorithm has a significant impact to the parameter riding index. If the algorithm do not have a good global convergence and a good control of the parameters dispersion, the identified model parameters will be unpractical.Contraposing to the Ant colony algorithm easy to fall into local optimum and the Chaos's ergodicity and randomness, the chaos is introduced into the ant colony algorithm. Then a new algorithm of the chaos ant colony optimization is proposed for the load model parameter identification, overcoming the dispersive of model parameters, reducing the identification error, and improving the robustness and precision of identification results.The classification of load models based on identified parameter space is researched, which is an important issue in practical load model. In this paper, the identified parameter is used as feature vector of the load model cluster analysis; HCM and FCM are used as the classification algorithm. The example results indicated that FCM algorithm is more convenient for the classification of load models in sensitivity of initial conditions and veracity of results.In addition, since some initial value may lead to incorrect classification with FCM algorithm, an improved FCM algorithm is proposed in this paper and it selects the initial cluster center using the largest Euclidean distance. The results by exemplification expatiate that this method avoids the error classification and reduces the number of computing, which is excellent for the classification of load models.