Parameters Identification For Power Load Models Based On Improved Particle Swarm Optimization Algorithm

As an important part of power system, electrical load has a great influence on the sta-ble operation of power system. As the researchers involve their deep efforts in the field of power load modeling, the mathematical models of electrical load have been mature. However, it’s quite difficult to obtain the parameters of the models due to the intrinsic complexity of load, while the accuracy of the parameters dramatically affects the effec-tiveness of load model. Consequently studying the parameter identification of load model is of great significance.Static and dynamic load models are analyzed, numerical simulation for the start and disturbance transient process of the induction motor is executed with time domain si-mulation method. The result proves the feasibility of the applied numerical algorithm and mathematic model. On the basis of studying particle swarm optimization, a muta-tional PSO with S-curve inertia weight is proposed to overcome premature and poor er-godic property of PSO with linearly decreasing weight and PSO with constriction factor. Several benchmark functions are used to test the performance of the proposed PSO. The results indicate that the proposed algorithm could be applied in deriving parameters of load models due to its outstanding converging speed and solution quality. The characte-ristics of power function and binomial static load models are analyzed, and the cause of non-uniqueness of binomial static model parameters identification is theoretically ex-plained. It is concluded that there is a link between the uniqueness of identification and the structure of load after the validation based on the measurement data. Then the power function model is selected as the static model, which is identified with improved PSO. The limitation of static model description capability is proved by means of fitting the actual mutation load with static model. Meanwhile, the composite load model consti-tuted by ZIP static model in parallel with induction motor is researched. To solve the differential equation of dynamic model, the initial states of arguments in dynamic ma- thematic model are achieved, and the fourth-order Runge-Kutta is employed to deduce the iteration difference equations. Combining with improved PSO, the calculation pro-cedure of dynamic model is designed. And parameters identification of dynamic load model is realized based on the measured data.The results demonstrate that the fitting data of the mutational PSO with S-curve iner-tia weight is closer to measured data. Compared with conventional methods, the im-proved algorithm which validates its accuracy and feasibility, has better convergence speed and higher identification precision, and can be applicable to estimate the parame-ters of load models.