Load modeling and parameter identification based on random fuzziness clustering

For the generally larger fluctuation range of reactive loads than active loads in practice, along with the time and composition variation of power load, a new method for load modeling and parameter identification is proposed. Firstly, to describe the mentioned characteristic of reactive power, an improved linearized-GNLD reactive model is introduced. Then, a new random fuzzy clustering method is presented by applying the effectiveness function to the random fuzzy clustering, thus achieving the optimal clustering center number and the corresponding clustering membership grade simultaneously without prior classification number. Finally, a new method for load modeling and parameter identification is obtained by combining the method proposed, the improved linearized-GNLD model and the identification strategy of clustering-before-parameter identification. Compared with other methods, the model obtained makes the practical application more convenient even though the load model number is not given in advance. Moreover, the application of the identification strategy proposed guarantees that each type of load models gained is based on a similar load curve (i. e., the collected load curve groups composed of the node voltage, the active power and the reactive power) and the model obtained has higher identification precision and is better for promotion. Example results demonstrate the effectiveness and validity of the method proposed. © 2013 State Grid Electric Power Research Institute Press.