A Tensor Decomposition Approach for Contingency Screening and Coherency Identification in Power Systems

Contingency Screening and Coherent Identification are two fundamental parts of power system planning and operation. A common characteristic among these two methods is the need to analyze multiples contingencies. However, most of the current work existing in the literature is based on the paradigm of analyzing one contingency at a time, using 2D arrays (matrices) for the event representation. The drawback with this type of representations is the impossibility to consider multiple contingencies simultaneously. In this paper a reformulation of the problem using 3D arrays (tensors) is presented. Then, the extraction of the information is carried out using PARAFAC2. With this information, a severity index for contingency screening is proposed and identification of the coherent areas is accomplished. The approach is validated in the IEEE NETS-NYPS test system. The results confirm that the proposed approach allows to extract more information than in the traditional form.