5.2.3. Scheduling Results of Case 3

This scenario analyzed the validity of robust stochastic optimization theory in controlling uncertain variables of constraints. Compared with scenario 2, this scenario introduced both the CVaR method and robust stochastic optimization theory, which further strengthened the constraints on uncertainty. In order to avoid WPP and PV uncertainty risks, the MEG reduced scheduling of WPP and PV. The output of WPP and PV decreased to 9602 kW·h and 6287 kW·h, respectively. Correspondingly, the MEG operating income also reduced. Table 4 shows the dispatching results of the micro energy grid of Case 3.


**Table 4.** Scheduling results of micro energy grid of Case 3.

Furthermore, the output distribution of the micro energy grid at different times was analyzed. Compared with Figure 5, when considering robust stochastic optimization theory, the WPP and PV uncertain variables in the constraints could be described. The corresponding output power of WPP and PV decreased, especially during peak times. The MEG will buy more power from UPG, thus reducing operational risks. Because part of WPP and PV are converted into heating, which creates uncertainty in the heating load supply, the MEG will buy some energy from UHG and UCG to realize a reliable supply of heating and cooling, to reduce the risk to the heating and cooling supply. This means that the introduction of the CVaR method and robust stochastic optimization theory can control the energy supply risk of electricity, heating and cooling loads simultaneously, and take into account both operational benefits and risks, to achieve the optimal safe and steady operation of the MEG. Figure 7 is the output distribution of the micro energy grid of Case 3.

**Figure 7.** Output distribution of the micro energy grid of Case 3.

In addition, to study the applicable space of the risk aversion model, a sensitivity analysis of the robustness coefficient and confidence was carried out. It can be seen that when 0.7 < Γ ≤ 0.95, the increase of β will result in a larger improvement of CVaR, which means that the MEG operation scheme will change when considering WPP and PV uncertainty. When Γ ≤ 0.7, the increase of β will result in a lower increase of CVaR. The decision makers will pay attention to the operation benefits and risks of the MEG at the same time, so the operation of the MEG is relatively steady, but the overall CVaR value will increase with the increase of Γ. When Γ ≥ 0.95, the increase of β will bring a great increase of CVaR. In this case, the decision makers are extremely risk averse. A smaller uncertainty will bring greater operational risk. Figure 8 is the analysis result of the MEG operational risk under different robustness coefficients and confidence levels.

**Figure 8.** The analysis result of the MEG operational risk under different robustness coefficients and confidence levels.

In general, the CVaR method and robust optimization theory can better describe the uncertain risk of MEG operation. To realize the optimal operation of the MEG, decision makers should set reasonable confidence and robustness coefficients, considering the operational risks and benefits of the MEG at the same time.

#### *5.3. Results Analysis*

According to the above three scenarios, the impact of demand response and MTEA on MEG operation is further analyzed to establish external key factors of MEG operation.

(1) DR optimization effect analysis

DR includes two response modes: PBDR and IBDR. PBDR indirectly leads terminal users to use energy reasonably by implementing a differentiated time-of-use price. IBDR directly controls the terminal users' load by signing a pre-agreement with them. Figure 9 is the load curve of power, heating and cooling before and after PBDR.

**Figure 9.** Load curve of power, heating and cooling after demand response (DR). PBDR: price-based demand response; IBDR: incentive-based demand response.

Compared with original load curve, the maximum load of peak time after PBDR decreases, and the minimum load of valley time increases. The maximum load reduction effect after IBDR is stronger than that of PBDR, but the valley load enhancement effect is weaker than that of PBDR. At the same time, after the application of PBDR and IBDR, the peak load decreases more, the valley load increases more, and the load curve becomes smoother. Table 5 shows the dispatching results of the micro energy grid before and after PBDR.


**Table 5.** Dispatching results of the micro energy grid before and after PBDR.

According to Table 5, if PBDR is considered, the WPP and PV grid-connected power are increased by 565 kW·h and 370 kW·h, respectively. As the load curve becomes smoother, the output of PS, HS, and CS decreases, indicating that the peak load regulation demand of the MEG for WPP and PV decreases. Similarly, after PBDR, the load in the valley period increases and the convertible power of P2G decreases. From the operational objective function, the economic benefits, carbon emissions and

CVaR values of MEG operation after PBDR are all optimized. Further, the output distribution of MEG at different times after PBDR was analyzed. Figure 10 is the output distribution of the micro energy grid after PBDR.

**Figure 10.** Output distribution of the micro energy grid after PBDR.

According to Figure 10, the output distribution of the MEG at different times was analyzed. Due to PBDR the load curve became smoother, and the WPP grid-connected power increased in the valley period. The PV grid-connected power increased in the peak period because IBDR can provide a peak-shaving service. In terms of power load, the power supply structure was cleaner and lower in carbon due to the increased power generation of WPP and PV. In terms of heating and cooling load, the smoother load curves reduced peak-shaving demand, make full use of IBDR, which can provide a peak-shaving service, which optimizes the meeting efficiency of the heating load and cooling load. In general, PBDR can optimize the output structure of the MEG and improve the operational efficiency of the MEG.

(2) MTEA sensitivity analysis

For the MEG, the main carbon emission sources include CGT and UPG. Therefore, the setting of MTEA will directly affect the power supply of CGT and UPG to the MEG. Therefore, this section outlines the sensitivity analysis we carried out on MTEA to construct the optimal dispatching strategies with different MTEAs. Table 6 shows the dispatching results of the micro energy grid with different META.


**Table 6.** Scheduling results of the micro energy grid under different META.

According to Table 6, with the increase of META, the power generation of WPP and PV gradually decreased. This is because the MEG is more inclined to buy power from UPG to avoid uncertain risks. It also reduces ES and IBDR output, meaning that the demand of the MEG for reserve services is reduced. When META is small, the carbon emissions of CGT and UPG power generation will reduce their power generation advantages, while more WPP and PV are scheduled to satisfy the load demand. Generally speaking, a reasonable META needs to be set to enhance WPP and PV grid-connected space and realize the optimal operation of the MEG as a whole.

#### **6. Conclusions**

In order to improve the sustainable development of distributed energy such as wind and solar, this paper emphasized optimization of the operation of the micro energy grid aggregated by multiple distributed energy sources. Because of the strong uncertainty of WPP and PV, the CVaR method and robust stochastic optimization theory were applied to describe the uncertainty of the objective function and constraints, and a risk aversion dispatching model of the micro energy grid considering the demand response and maximum total emission allowance was constructed. Finally, this paper selected the Xinxiang Active Distribution Network Demonstration Project in Jining, China as an example. The following conclusions were reached:

(1) The micro energy grid can make the most use of the complementary characters of different energy, such as WPP, PV, and CGT, and can make use of a variety of EC equipment (P2H, P2C, H2C, P2G) and ES equipment (PS, HS, CS, GS) to achieve optimal satisfaction with various loads types, such as electricity, heating, cooling and gas. On the one hand, clean energy has both environmental and economic characteristics, which can improve the economic and environmental benefits of MEG operation. On the other hand, the cooperative operation of various EC and ES can effectively handle the strong uncertainty.

(2) The proposed risk aversion dispatching optimization model with the CVaR method and robust stochastic optimization theory can describe the impact of uncertain variables in objective function and constraints, and provide a basis for decision makers who have different attitudes. When Γ ≤ 0.7, β increases and results in a lower increase of CVaR, and decision makers are operating in risk-free conditions. When 0.7 < Γ ≤ 0.95, β increases and results in a larger increase, the decision makers are risk-averse. When Γ ≥ 0.95, β increases and results in a lower increase, and the decision maker becomes extremely risk averse. Thus, it is possible to formulate the optimal scheduling strategy for decision makers who have different attitudes by setting reasonable confidence and robustness coefficients.

(3) DR can smooth the energy load curve of electricity, heating, cooling, and gas. MTEA can enhance the market competitiveness of the clean energy market, thus promoting grid-connected power of clean energy, such as WPP and PV, and optimizing the multi-energy supply structure of the MEG. On the one hand, PBDR has a better "valley filling" effect, and IBDR has a better "peak cutting" effect. The synergistic operation of the two can maximize the "peak cutting and valley filling" effect. On the other hand, MTEA determines the supply space of CGT and UPG. When MTEA is low, the MEG gives priority to WPP and PV to satisfy the load demand, which will meet the maximum carbon emission constraint.

**Author Contributions:** Conceptualization, P.L.; methodology, W.F.; software, L.J.; resources, N.L. and F.Z.; data curation, F.Z.; writing—original draft preparation, X.F.; writing—review and editing, X.F. and H.L.; visualization, L.J.

**Funding:** This research was funded and supported by the Project funded by China Postdoctoral Science Foundation (2019M650024), the National Nature Science Foundation of China (Grant Nos. 71904049, 71874053, 71573084), the Beijing Social Science Fund (18GLC058) and the 2018 Key Projects of Philosophy and Social Science Research, Ministry of Education, China (18JZD032).

**Conflicts of Interest:** The authors declare no conflict of interest. It should be noted that the whole work was accomplished by the authors collaboratively. All authors read and approved the final manuscript.

#### **Nomenclature**



