**1. Introduction**

In recent years, environmental pollution and the current depletion of fossil energy have become more and more serious. The energy structure should be transformed and upgraded urgently. It is the current trend of the energy field to pursue the efficient use, clean environmental protection, and sustainable development of energy [1]. As a further extension of micro power grids, a micro energy grid could realize the coordinated planning and unified scheduling of multiple energy sources (electricity, heat, cold, and gas, etc.) through new energy technologies and internet technologies, which could effectively improve energy efficiency while achieving local production and consumption of energy [2]. The China Development and Reform Commission also put forward the Guiding Opinions on the Development of "Internet + Promoting" Intelligent Energy, pointing out that it is essential to strengthen the construction of multi-energy synergistic integrated energy networks, and the coupling interactions and comprehensive utilization of different energy types, such as electricity, gas, heating, and cooling [3].

In recent years, micro energy grids have attracted widespread attention and practical development. The smart polygeneration microgrid (SPM) project of the University of Genoa for the Savona campus involves multiple electric, thermal, and combined heating and power generators [4]. Cassel University integrated a wind turbine, PV, biogas power station and hydro power plant into a micro energy grid (MEG) [5]. In 2014, the Xiaozhongdian wind-photovoltaic (PV)-hydro distributed demonstration project of the China National Electric Power Group Corporation successfully connected to the grid in Yunnan province [6]. The La Plata University in Finland is developing a 40 m diameter straight-leaf vertical shaft wind turbine that drives an oil-fired heating system for greenhouse heating [7]. The pilot projects of wind power heating in Linxi County and Zha'rutqi of the Inner Mongolia Autonomous Region have been operational for three years, and the annual consumption of wind power is approximately 149 million kW·h [8].

The operation mechanism and scheduling operation of micro energy grids have always been hotspots for research, both at home and abroad. On the premise of meeting the security constraints, the micro energy grid uses different objectives to rationally arrange the operation of energy equipment within the micro energy grid. Zhang et al. [9] constructed a micro energy grid operation optimization mode taking the minimum daily operating cost as the target. Du et al. [10] introduced modeling, planning and optimization methods for a regional integrated energy system. Luo [11] established a model for minimizing the sum of various costs and analyzed the integrated energy system with P2G. The above reference considered the objectives and related constraints of the micro energy grid, but the proposed models were used to analyze the uncertainty of the distributed energy. In fact, the uncertainty will affect the coordinated supply of multi-energy loads, such as electricity, heating, cooling, and gas, in the grid.

The uncertainty of distributed energy in a micro-energy network is mainly reflected in the volatility of wind power plants (WPP) and PV output power. A key issue for MEG operation is how to use units, storage devices, electric vehicles, and load, to balance random changes of wind and solar units, to guarantee the steady output of the MEG. Peik-Herfeh et al. [12] used the two-point estimation theory to take an estimate point on both sides of the forecast value to represent the unit output variability. Yang et al. [13] gained distribution parameters based on the characteristic that the wind speed obeys a Weibull distribution. Zamani et al. [14] used a stochastic program to handle electricity price uncertainty and studied a virtual power plant bidding model considering the uncertainty. Tan et al. [15] constructed an economical dispatching model considering the output power volatility of clean energy, based on a chance-constrained program. The above related research mainly considered uncertain variables as random variables and constructed stochastic dispatching optimization models by using stochastic modeling methods, such as stochastic programming and robust optimization. The validity of the method was verified by actual cases.

The above research shows that the existing research about micro energy grids focuses on system modeling, optimal operation, and uncertainty analysis. However, in uncertainty analysis, a stochastic program describes the uncertainty with stochastic variables. Based on the probability distribution of stochastic variables, system constraints are described as opportunistic constraints [16]. However, whether DERs with a small capacity and large quantity have statistical characters needs to be checked. The accurate establishment of information collection and a probability distribution function is difficult. The optimal solution sets of robust optimization have a certain degree of restraint on the effects. Adjusting the size of a coefficient can determine the dispatching scheme, which can restrain the influence of uncertainty to different degrees [17]. At the same time, the existing research results are more focused on the processing of constraints with uncertain variables, lacking consideration of the objective function processing method with uncertain variables. Conditional value at risk (CVaR) can quantitatively represent the uncertainty risk of the objective function. By combining it with robust stochastic optimization theory, a relatively complete risk decision model can be constructed. According to the above analysis, an optimal dispatching model for a MEG is put forward. The main contributions are as follows:


The structure of the paper is as follows: Section 2 designs a core structure for the micro energy grid and establishes an operation model of the equipment and DR operation model. Section 3 constructs the basic scheduling model of the micro energy grid without considering uncertainty, which takes maximizing the operational benefit as the optimization objective. Furthermore, a risk aversion model of the micro energy grid is established on the basis of CVaR and robust stochastic optimization theory in Section 4. Finally, Section 5 selects the China Jining Xinxiang Active Distribution Network Demonstration Project as an example object to verify the effectiveness and applicability of the model. Section 6 outlines the contributions and conclusions.
