Learning based cost optimal energy management model for campus microgrid systems

The introduction of microgrids has enabled an efficient energy management in the system with installation of renewable energy sources. As one of the representative models of microgrid, various studies on campus microgrids (CMGs) have been conducted. In operation of CMG, various energy consumption resources and renewable energy are considered to minimize overall cost or peak power in the system. However, most of conventional researches only deal with performance analysis in terms of simulation by collecting data from the different environments. In this case, there are lack of consideration in power regulation or electricity cost which make it difficult to apply the researched energy operation technology to the actual power system. To solve the problem, this paper build a test-bed in an actual CMG environment and collect dataset through the various IoT sensors. In addition, uncertainties that occur through the various power resources are analyzed and used to derive net energy consumption scenarios. In this way, we propose a new cost optimal energy management model with the detailed analysis of power generation and consumption using various auxiliary IoT devices. Based on the real-world datasets from the implemented CMG, we show that the proposed analytical models and energy management model are feasible for actual environments. With satisfying the constraints, we show that the daily electricity cost could be reduced up to 2.16% and peak power is reduced up to 3% compared to the case without considering the uncertainties in CMG.