Colloquium Computer Science - J.Yamnenko (Kyiv Polytechnics)

Titel: Use of machine learning in MicroGrids

Abstract:

Modern electrotechnical objects with distributed generation (alternative and renewable energy sources) rise serious demands for data processing including climatic, technical, economical and subjective parameters.

Operating with large massive of different data, its processing, analysis and generation of control signals in real time could be achieved by use of spectral and wavelet analysis, big data processing and machine learning. These modern methods allow increasing of energy efficiency, quality of energy and quality of services.

Separate aspect is detection of anomaly human behavior. MicroGrid is characterized by the presence of a person, which leads to the need to take into account subjective factors - the presence of a person, its interference in the functioning of electrical equipment and the impact of his actions on the working modes of whole system. Research directed to revealing of non-typical (abnormal) behavioral models is topical for the systems of psycho-physiological monitoring of elderly people, persons who need constant observation and care, persons under post-stress or extraordinary conditions. As the key parameters for the simplest case of detecting abnormal behavior, the average power consumption at five-minute intervals, as well as the number of triggers of the motion sensor were considered.