Modeling Non-Linear Load Currents by Mixture Distribution Separation

The paper presents an algorithm for modeling the harmonic currents of non-linear loads. Non-linear load models are necessary to plan the non-sinusoidal states in power networks to control the power quality. The measured non-sinusoidal state harmonic parameters are the source data for modeling. The analysis of these data shows that the harmonic parameters are stochastic. The non-linear load model is a set of values of active and reactive harmonic currents with a given probability. These are calculated after identifying the functions that govern the distribution of series of random harmonic current values. The algorithm identifies such distribution functions that are not described by the known laws and are essentially a mixture well-known distributions. The algorithm uses a method of mixture distribution separation. The paper gives examples of the algorithm being used to identify the distribution function to and calculate an active/reactive 5-th harmonic currents of an aluminum plant.