Blind Source Separation Based Analysis of Correlated Harmonic Sources in a Micro Grid Environment

Harmonics are a persistent problem in power systems starting from the simple to any complex natured power system network. Blind source separation using Independent Component Analysis (ICA) is applied to estimate the harmonic sources with least prior knowledge on the topology of power systems. The limitation of the method is that it assumes that the harmonic sources are statistically independent. However, in simple microgrid system powering a small region, it becomes very likely that the harmonic sources like two housing colonies,etc to have some degree of dependence which makes them correlated to a minimal extent. Under such circumstances, it is observed that the traditional ICA algorithms like FastICA breakdown. In the work, two algorithms suitable for statistical dependence namely the eigBSE(Eigen value based Blind source Extraction) algorithm and the maxNG (maximum non Gaussianity based) algorithm discussed in literature are applied to a simple four bus microgrid model. The separation quality and performance indices of the two algorithms are explored for various correlation coefficient values and it is found that in majority of the cases maxNG algorithm is more precise. More details of the theoretical issues of statistical dependence and the algorithm behaviour needs to be workedout.