A Machine Learning Toolkit for Power Systems Security Analysis

Machine Learning techniques have been extensively used in Power Systems Analysis during the last years. A Machine Learning Toolkit, i.e. a versatile software environment incorporating multiple interoperable tools facilitating experimentation, can be a valuable asset during power systems analysis studies. In this paper our experience with building such a Toolkit, incorporating a repository of data (Data Warehouse), is described. A number of machine learning and analysis tools have been built and applied on collected structured data in order to perform steady state security assessment of a power system. The Toolkit (UMLPSE), the data collection process, the analysis performed and the tools used are the subject of this paper. In particular application of the developed toolkit in contingency analysis is demonstrated. A series of system indices and metrics are stored in the data warehouse describing the effects each contingency is expected to have on each operating point. The system indices and metrics are subsequently filtered through statistical procedures with qualitative measures and criteria to be used as training data for machine learning tools (neural networks and decision trees). The performance indices and the fine tuning of the parameters of these machine learning tools are then considered for the screening and ranking of the contingencies for any given electrical network operating point. It is argued that the proposed approach and tools can be applied to many similar power systems analysis studies.