engleski

Optimal Control of a DC Microgrid with Renewable Energy Sources

Microgrid is defined as a cluster of distributed generation sources, storages and loads that operate together and thereby improve reliability and quality of the power supply and of the power system. Microgrids are expected to significantly reduce power transmission losses, and through distributed energy storage to enable integration of a large share of renewable energy sources. The main goal of this doctoral dissertation is to develop a control algorithm for a residential grid-connected DC microgrid with renewable energy sources, that achieves optimal techno-economic microgrid operation. To that aim, the dissertation addresses several important aspects: (i) modeling of all systems included in the microgrid ; (ii) prediction of local production and consumption profiles for one-day ahead by means of machine learning and artificial intelligence, which is especially challenging due to strong correlation with atmospheric conditions ; (iii) power flow management based on convex optimization and model predictive control scheme with receding horizon principle, which is developed in a stochastic framework to account for local production and consumption profiles prediction uncertainty ; and (iv) DC link (bus) voltage control with power flow reference tracking, whereas optimal power references are generated by the power flow management control loop. The control (decision) variables in the considered system are charge and discharge profiles for local energy storages, which serve as energy buffer that improves system stability and enables trading on decentralized electricity market. In the simplest scenario, microgrid would buy electricity during low electricity price intervals, and sell electricity during high price intervals, in that way reducing microgrid’s operating costs. However, a decision when to buy and sell energy to the utility grid and in which amount, i.e., when to charge and discharge storages, is a complex function of the predicted microgrid load, power production, current storages state of charge, and of the predicted electricity price, which altogether make the central subject of the conducted research. It should be noted that all concepts developed in this dissertation could also be applied to AC microgrids, with an exception of microgrid’s bus voltage control. The considered residential DC microgrid is formed in the Laboratory for Renewable Energy Systems (LARES), at the University of Zagreb Faculty of Electrical Engineering and Computing (UNIZG-FER), Croatia. The developed concepts are verified by experiments in LARES based on real meteorological and electricity price data, i.e., they are tested in real operating environment. Predictions of relevant meteorological variables, which are used as inputs to predict local production and consumption profiles in the microgrid, every hour for one-day ahead, are provided by the Croatian Meteorological and Hydrological Service (DHMZ).

dc microgrid, smart grid, renewable energy systems, distributed storage, photovoltaic system, machine learning, artificial intelligence, load prediction, battery state of charge estimation, Kalman filter, ultracapacitor, fuel-cell, power žflow management, convex optimization, linear program, model predictive control, stochastic control framework, voltage stability

nije evidentirano