The Studies On Identification Of Electric Load Model

Some important problems are studied in the thesis about electric load modeling of measurement-based modeling approach, such as the characteristics clustering and synthesis of electric dynamic loads, the dispersing of the load model parameters and the electric nonlinear load modeling. These studies are very important to solve the difficulties such as the time and composition variation of the loads in the load modeling. The main works are as following:The dispersing of load model parameters existing in the load modeling of measurement-based modeling approach is further studied in the paper. The results show the inappropriate selection of the load model structure and the noise lead to the dispersing of load model parameters, either for the static load model or the differential equation load model. In order to solve the dispersing of model parameters, an effective method, the characteristics synthesis of loads, is presented. Many examples of load modeling by using the field test and the simulation test data indicate the validity and feasibility of the method. With the development of measurement-based modeling approach, the characteristics clustering and synthesis of electric dynamic loads arise. In this paper, a new algorithm based on Kohonen self-organization neural network is firstly presented for the characteristics clustering of dynamic loads. Many sets of load data measured from North China Power System in three years(1996-1998) have been dealt with using the method. The results show load characteristics have rule though they are random and time-varying.In accordance with the problem that some parameters can be identified easily and other parameters are difficult to be identified in the load modeling of induction motor, this paper elucidates the relations between the parameter sensitivity and the parameter identifiability through the analytic sensitivity analysis. The problem that deals with the effects on parameter sensitivity of the intensity of excitation is also studied. Finally, the parameter identification strategy that is meaningful and valuable in modeling is advanced. The simulation and field data modeling examples prove that this strategy is feasible.When the voltage and frequency change greatly, the nonlinear characteristic of the electric load is considerable, especially during the long-term dynamic process of power system. Since the single load model could not describe the nonlinear characteristics, the global load model (T-S fuzzy model) is first presented. The detailed algorithm to obtain the global load model is also developed. The example of load modeling by using the simulation test data indicates the validity and feasibility of the method.ANFIS (Adaptive Neural Fuzzy Inference System) is first presented to obtain the global load model for describing the nonlinear characteristics of the electric load in the paper. ANFIS automatically produces the If-Then rules, whose premise parameters and consequent parameters are adjusted by using a Back-Propagation algorithm in combination with the least squares method. The simulation results also show the feasibility of the method.