Application of state estimation and machine learning techniques for the energy management of microgrids

• Grant number: 20/03069-5
• Support Opportunities: Scholarships in Brazil - Doctorate
• Start date: June 01, 2020
• End date: May 31, 2025
• Field of knowledge: Engineering - Electrical Engineering - Power Systems
• Principal Investigator: Luiz Carlos Pereira da Silva
• Grantee: Byron Alejandro Acuña Acurio
• Host Institution: Faculdade de Engenharia Elétrica e de Computação (FEEC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Associated research grant: 16/08645-9 - Interdisciplinary research activities in electric smart grids, AP.TEM
• Associated scholarship(s): 22/16881-5 - Application of state estimation and machine learning techniques for the energy management of microgrids, BE.EP.DR

Abstract

Microgrids are systems that integrate renewable energy sources, that can operate connected or disconnected to the main electrical network. Traditionally, microgrids operate with Distributed Energy Resources (DERs) such as photovoltaic, wind, storage and thermal generation systems. The patterns of renewable generation in the microgrids are highly intermittent, creating stochastic problems by its nature that challenge real-time control and reliability systems. In this context, it is necessary to detect and process these dynamic and random events that are out of human control, which can be done through machine learning and state estimation algorithms, to calculate and predict the conditions of the microgrid components in different operating scenarios. Thus, state estimation in microgrids is used to convert metering data and other available information into a reliable estimation of unmonitored variables, or alternatively it can be used to filter noise and gross measurement errors. On the other hand, machine learning can be used to deal with microgrid's stochastic parameters. For example, in short-term forecasting of renewable energy, demand, electrical market, and storage systems under uncertainty. Hence, the objective of this research project is to develop and evaluate the performance of machine learning and state estimation algorithms to perform essential microgrid's functions, such as data-driven energy management and short-term data forecasting of renewable generation and demand. Two software prototypes will be developed and tested in the microgrid laboratory of UNICAMP, specifically in the microgrid pilot named as LABREI, which is being implemented at UNICAMP through the project R&D MERGE - Microgrids for Efficient, Reliable and Greener Energy, and through the thematic project FAPESP 2016/08645-9. The microgrid pilot LABREI is the only one of its kind in Brazil, because has technologies for analysis of microgrid's operation with DERs, real-time controls and instrumentation, advanced telecommunications and data acquisition systems, which are essential for a practical validation of the proposed algorithms. (AU)