1
A MACHINE LEARNING (ML) BASED APPROACH FOR SMART ENERGY MANAGEMENT IN MICROGRIDS José Antonio Araque Gallardo 1 - PhD student , Juan Carlos Burgillo - Rial 2 - PhD, Thesis advisor 1. School of Telecommunications Engineering , University of Vigo, Department of Electronic Engineering , University of Sucre - Colombia, [email protected], [email protected] 2. Associate Professor , Department of Telematic engineering , University of Vigo, [email protected] Motivation C onventional centralized energy system : ✓ It is strongly dependent on fossil fuels and other non-renewable sources. ✓ Poor energy efficiency. ✓ It contributes to air pollution [1]. N ew approach: ✓ Power generation locally. ✓ Non-conventional and renewable sources: Wind power, solar photovoltaic cells, etc. ✓ Integration into the utility distribution network [1]. C hallenges: ✓ Uncertainty and complexity. ✓ The system needs to monitor, predict, schedule, learn and make decisions. [2], [3] Can we do smart energy management in microgrids? Is the ML a suitable approach to optimize the energy management in microgrids? The aim of this research is to optimize the energy management in microgrids applying Machine Learning (ML) techniques. For this, we need: ✓ Extract knowledge of the microgrid and learn the patterns and habits of energy consumption. ✓ Make accurate predictions of consumption and power generation in the microgrid. ✓ Optimize the schedule of distributed energy resources in the micro-grid through smart decision making. Thesis objectives Next year planning A1: Literature review/state of the art. A2: Simulation model. A3: Study, select and describe the optimization targets. A4: Investigate and select the ML algorithms for consumption and power generation. A5: Implement and test the prediction algorithms on the model. A6: Study the ML algorithms for knowledge extraction. A7: Implement and test the knowledge extraction algorithms on the model and other available datasets. A8: Develop the smart decision-making algorithms based on ML and test it on the model. A9: Write and prepare the Ph.D. thesis. WP1: Write paper 1 (state of the art). WP2: Write paper 2. (Simulation and validation of the model) WP3: Write paper 4 (knowledge extraction and learning from microgrid) WP4: Write paper 3 (prediction of consumption and power generation in microgrids). WP5: Write paper 5 (smart decision making in microgrids) Research plan References [1] S. Chowdhury, S.P. Chowdhury and P. Crossley. Microgrids and Active Distribution Networks. London: The Institution of Engineering and Technology, 2009. [2] E. Mocanu. “Machine learning applied to smart grids”. PhD. Thesis, Technische Universiteit Eindhoven, Eindhoven, 2017. [3] V. Francois. “Contributions to deep reinforcement learning and its applications in smartgrids”. PhD. Thesis, University of Liège, Belgium, 2016. Identify and carry out an in-deep literature review and state of the art. Obtain a reliable and precise simulation model of the micro-grid. Write paper 1 (state of the art).
