Recursive Bayesian-Based Approach for Online Automatic Identification of Generalized Electric Load Models in a Multi-Model Framework

Electric loads are essential for power system dynamic simulation. However, load modeling is one of the most challenging topics due to the diversity and time-varying behavior of the load. When considering the intervention of rapidly developing distributed generation (DG), load modeling becomes more difficult. In this paper, a new solution for determining the unknown generalized load model is proposed. The radial basis function (RBF) neural network-based sub-models of generalized load are stored in the form of a sub-model bank. A recursive Bayesian approach is used to identify the sub-models and then merge them into one generalized load model according to their probabilities. The proposed method can be implemented online and adapt to describing the diversity and time-varying behavior of the generalized load. Numerical studies are carried out using both simulation data and actual measurements. The comparisons with other load modeling methods verify the advantages of the proposed method.