A novel data-driven intelligent computing method for the secure control of a benchmark microgrid system

Shunjiang Wang, Yan Zhao, Jiang He, Rong Chen, Lina Cao
2020 Computer Science and Information Systems  
Microgrid is a small-scale cyber-physical system, and it generally suf-10 fers from various uncertainties. In this paper, we investigate the secure control prob-11 lem of a benchmark microgrid with system uncertainties by using data-driven edge 12 computing technology. First, the state-space function of the benchmark microgrid 13 system is formulated, and parameter uncertainties are taken into consideration. Sec-14 ond, a novel data-driven intelligent computing method is derived from the
more » ... 5 based reinforcement learning algorithm, which only requires system data instead 16 of system models. By utilizing this computing method, the optimal control policy 17 can be obtained in the model-free environment. Third, the Lyapunov stability theory 18 is employed to prove that the uncertain microgrid can be asymptotically stabilized 19 under the optimal control policy. Finally, simulation results demonstrate the control 20 performance can be improved by tuning the parameters in the performance index 21 function. 22 Keywords: edge computing, microgrid system, secure control, reinforcement learn-23 ing. 24 a benchmark microgrid [5,13,14,23]. This benchmark microgrid consists of three main 30 parts: power generation, loads and distributed energy storages. The power generation in-31 cludes regular generation (microturbine), and supplies energy for the demands of various 32 loads. 33 However, due to the intermittent power injection from photovoltaic arrays and sud-34 den change of load demands, the unbalance between power supply and demand may oc-35 cur, which will cause the frequency fluctuations and threaten the security of the entire 36 microgrid. Thus, we incorporate distributed energy storages (electric vehicles) into this 37 microgrid to compensate the unbalance. The system data can be measured by sensors and 38 transmitted to the management center through the communication module. The whole mi-39 crogrid is controlled by using the edge computing technology [2,3,4,24]. The schematic 40 diagram of edge computing for the benchmark microgrid system is shown in Fig. 1. 41 2 Shunjiang Wang et al. Different from the traditional automatic control, edge computing is more like an in-1 telligent control method which is based on computing and information, and it mainly 2 concerns the control strategies for dispatch and optimization. In [2], by means of road 3 networks, the problems of frequently moving vehicles and network connectivity were an-4 alyzed, and then a modified greedy algorithm for vehicle wireless communication was 5 proposed for network optimization. In [3], a holistic framework to attack the QoS pre-6 diction was developed in the IoT environment, and authors designed a fuzzy clustering 7 algorithm which was capable of clustering contextual information. In order to fully uti-8 lize hidden features in edge computing environment, the work [24] presented a new ma-9 trix factorization model with deep features learning via a convolutional neural network. 10
doi:10.2298/csis190912023w fatcat:cdxxgfag5vhcjg5msv22htndoa