Two‐level scheduling of an active power distribution network by considering demand response program and uncertainties

This study investigates a hierarchical approach for modeling the mutual impacts of distribution network (DN) decisions and microgrids in a multi‐microgrid system under the cover of a two‐level problem. Due to the conflicting interests of decision‐makers in an active DN, the optimization problem has a hierarchical structure, with distribution companies (Discos) at the top attempting to maximize profits and microgrids at the bottom attempting to reduce costs. This study examines the relationship between the unpredictability of renewable energy sources and power demand. In this mechanism, the influence of the demand response program (DRP) is also considered an important aspect of intelligent systems. The two‐level nonlinear optimization problem has been transformed to a one‐level linear problem using dual theory and KKT conditions. According to the findings of the studies, even though cost reduction is the sole objective of scheduling in centralized mode, applying DRP to the system results in a 16% reduction in overall costs for the two‐level scheduling method, and a 22% reduction in daily costs when compared to two‐level scheduling without DRP.

voltage (LV) and medium voltage (MV).Power demand is increasing as a result of population increase (particularly in metropolitan areas) and the rate of industrialization.The possibility of developing a platform for local power generation and consumption, as well as better and more efficient exploitation of renewable energy resource (RES) capacity, has increased interest in MGs and their broader implementation.However, expanding microgrids (MGs) is difficult due to the fact that RESs offer a considerable percentage of MGs' power.Modeling the uncertain behavior of renewable energy sources and their reliance on environmental circumstances such as wind speed and solar radiation has become one of the necessary concerns in the efficient exploitation of the DN.The advent and widespread use of MGs may alter the distribution system's roles in the future.The new systems may be the only structural link between a group of MGs which configure a multi-microgrid distribution network (MMGDN).Energy management systems (EMSs) typically handle the optimal operation and management (O&M) problem in the MMGDNs, which is a lot more complicated and difficult than a O&M of a single MG.The O&M problem becomes more complex due to the divergent objectives of MGs and DN operators as well as the growing integration of RESs caused by the addition of more MGs to the system structure. 2he wind turbine units (WTs) and photovoltaic units (PVs) generate electricity based on weather conditions.As a result, the presence of renewable energy sources in MGs poses issues such as a lack of continuity and stability in these units' generation capability.On the other hand, network power demand is increasing.To meet the increased demand, one option is to add micro-sources to the networks.This method raises concerns and issues, such as their optimal placement.Another short-term remedy that generates environmental difficulties is purchasing power from the upstream network.Based on this, programs to manage consumers' power demand are an optimal solution to deal with the growth in demand.The DSM is a set of solutions that move the behavior of systems from "demand-oriented" to "generation-oriented." 3,4 The DRPs have recently been expanded in theory and technology so that customers may become more active in the energy market while also serving as the quickest, cheapest, and most reliable choice for resolving power market challenges.DSM activity is the primary choice in all energy strategies to establish a stable system. 5Customers who participate in DRPs alter their consumption profile to balance generation and consumption, improving the system's reliability.Also, responsive loads can be thought of as a form of support for RESs uncertainty. 6esearchers have long been interested in finding the best strategy to manage and schedule a single MG. 7,8In Reference 7, a mixed integer linear programming (MILP) method with a risk management approach is examined.Also, reserve market optimization considering uncertainties is conducted for an islanded MG. 8 Since they can increase system resilience and allow for more efficient resource sharing, MMG systems have been increasingly popular in recent years. 9,10he optimal management of a multi-MG system has been described and researched as a linear problem with an objective of enhancing the system's self-healing capacity. 11The effects of the distribution system, the wholesale market, and the impact of demand-side management technologies are not taken into account in this article.The hierarchical structure has the advantage of creating a better model of the ruling business in the electrical market and allowing for a larger number of market participants and players to be addressed compared to the centralized method.Delegating all day-ahead or real-time scheduling duties to an entity like independent system operator (ISO), it is subjected to a significant number of computations. 12However, for smaller systems, such as several residential MGs in retail electricity markets, centralized scheduling and decision-making may be acceptable. 13The optimal operation problem of an active local electricity market (ALEM) has been investigated in the form of a bi-level problem by Kim et al. 14 In this study, despite considering the topology of the DN, various aspects of intelligent DNs such as DRPs have not been considered.Sheikhahmadi et al. analyzed local energy markets and their modeling in DN decision-making with high penetration of distributed energy resources. 15n example DNs, the provided two-level model is implemented.However, despite taking into account key details such as the uncertainty of RESs and modeling it with the information gap decision technique, and considering system's power losses and power flow constraints, various aspects of smart grids such as demand side management mechanisms, and the ESSs, that play important roles in the system stability and facing the uncertainties of RES, have not been included in this study.The importance of charging and discharging strategies of batteries and their owners' roles, and the key role of ESSs in the operation and optimal participation of MGs in wholesale markets, have been explored in a valuable research by Bahramara. 16Afshan and Salehi have modeled the problem of optimal operation of the active DN considering the uncertainty of PV units' generation and the effect of ESSs.In this paper, the problem is modeled as a single-level problem. 17he optimal stochastic bidding strategy in joint energy and ancillary services market is studied in Reference 18.The coordination between transmission, distribution, and DER aggregators that interact in a real-time market is modeled and investigated in Reference 19, where the authors applied a risk-based two-stage stochastic model to address RESs uncertainties, and the conditional value at risk method to model the DER aggregator's risk level.The management of MMGDN with the approach of optimal exploitation of the common coupling point of MGs has been studied in Reference 20.In Reference 9, the effects of DRPs on the changes in operating and management costs of DNs consisting of several MGs have been studied from the perspective of technical and economic indicators.The scheduling and operation of multicarrier systems that include energy hubs with diverse types of power sources, using the Cournot competitive game strategy and taking into account the DRP and the ESSs, is of the recent studies in the area of energy systems management. 21ikmehr et al. looked at the effects of DRP in interconnected MGs. 22While the study does account for the decentralized energy management mechanism, it does not take into account DN exchanges with the upstream network.Additionally, the problem has been solved using metaheuristic approaches, which do not ensure the achievement of globally optimal answers.
Different gaps are seen in the prior models reported in various studies.Scheduling a multi-MG system using metaheuristic optimization methods to minimize additional mathematical calculations associated with simplifying the provided models is one method for addressing the challenge of optimal energy system scheduling. 23The model described in this article focuses on forced load interruption, leaving out the involvement of the DN and many elements of current MGs.The metaheuristic problem-solving technique has also been applied in a multi-island MG system that takes into account local and retail markets.Distribution network company (Disco) and similar innovative MG technologies are not taken into account in this study. 24However, reaching the optimal global response with metaheuristic and population-based methods is not guaranteed.
A bi-level optimization method is presented in this work to represent the collaboration between Disco and MGs, as well as the local power market considering DRPs and ESSs.This paper's proposed model results in a non-linear bi-level optimization problem that is transformed into a single-level linear problem utilizing Karush-Kuhn-Tucker (KKT) conditions and dual theory.The influence of the DRP is investigated at the problem's second level, which is the MGs decision-making level.The scenario tree technique has been used to adjust the uncertainty of wind speed and power demand.The centralized decision-making mechanism is compared to the suggested two-level model with and without DRP.
The contribution of this study can be summed up in the following way, in brief: • To design the scheduling optimization problem of an electrical DN consisting of several MGs with distinct characteristics, taking into account the wholesale market, the uncertainty of RESs and the power demand of MGs, as well as the DRP.
• To combine the two approaches of increasing profits at the second level (MGs) and reducing costs at the first level (price-taker Disco) in a two-level problem.
• To convert the two-level ALEM optimal scheduling problem to a single-level linear problem in order to facilitate obtaining the global optimal solution, in the condition that the MGs' load can be shifted during daytime.
• To investigate the impact of demand side participation on how to schedule and operate the system.
The rest of the paper is organized as follows: Section 2 discusses the multistage decision-making procedure at various stages of the electrical energy system.Section 3 investigates the bi-level model between second and third level decision makers (Disco and MGs).The nonlinear two-level model is linearized in Section 4. In Section 5, numerical studies on an ALEM are performed, and lastly, in Section 6, the final result is provided.

PROBLEM STATEMENT
The three levels of decision-making in a power system are shown in a general scheme in Figure 1.According to Figure 1, the generating companies are the top-level decision-makers in this framework.MGs are considered third-level decision-makers, while Discos, markets, and big consumers are second-level decision-makers.More details of the decision-making process and the flow of energy and information between different parts of an electricity system, including the wholesale and local energy markets, are illustrated in Figure 2. MGs, which are at the second level of the bi-level problem in a local electricity market, may have unique characteristics, but their objective is to maximize profit by engaging in the most effective interacting with Disco.In an ISO-regulated wholesale electricity market, both top-and bottom-level decision-makers work together.The Disco sends its price bids to the day-ahead energy market, which is settled by the ISO.After clearing the market, the power exchange of the Disco with the market is determined.Other participants may fit into any of the levels of the framework depending on the nature of the work they do. 25

F I G U R E 1
The three levels of decision-making in a power system and their interactions.

F I G U R E 2
Overview of the decision-making mechanism considering wholesale and local energy markets and local energy markets.

F I G U R E 3
Overview of the decision-making mechanism considering wholesale and local energy markets.
In local power markets, cooperation between second and third-level decision-makers is examined.The local market price is the same for all MGOs due to the Disco's consistent price signaling.The MGO chooses the best timing for DERs and the best exchange of energy with the Disco after clearing the local energy market in accordance with the aggregators' and Disco's bids. 16he governing mechanism at an ALEM can be defined as a two-level optimization problem when the decision-makers at the second and third levels are considered.The presumptive model has decision-makers at two levels: the upper-level decision-maker leads the problem, while the lower-level decision-makers adhere to the leader's actions.Disco and MG are regarded as the model's leaders and followers, respectively.The Disco, a price-taking actor in the wholesale energy market who purchases power at a specified price and interacts with MGs in the local electricity market, is regarded as the upper-level decision-maker (leader) of the problem, as illustrated in the Figure 3.The MGs receive local electricity tariffs from cooperating with Disco.By obtaining these pricing, MGs can also plan their resources.Then, they coordinate the power exchange with Disco.The Disco also makes decisions about the power that is bought on the wholesale energy market.Each MG schedules the amount of load shedding, the charging and discharging of batteries, and the power generation of the DGs.Depending on the characteristics of the resource and the demand, the MGs can participate in the local electricity market each hour as either producers or consumers.Consequently, interactions between the second and third levels have no discernible effect on wholesale market pricing. 15

BI-LEVEL PROBLEM MODELING
This section illustrates the two-level problem in an ALEM described in the previous section.It should be emphasized that because this greatly expands the problem's dimensions, modeling in this case has neglected the implications of the DN.As a result, the impact of power losses is not considered.The objective function of the first level is to maximize Disco's overall profit, and it is represented as follows: Also, the constraints related to the first level of the problem are as below: 0 ≤ P M t ≤ P max , ∀t. ( For second-level decision-makers, the objective function with the profit maximization approach will be as follows: The constraints related to the second level of the problem are as follows: In the objective functions expressed by Equations ( 1) and ( 5), the vector Y and X represent the decision variables of the problem in each of the first and second levels.In  The term c D t P D j,t,s appears in the objective functions of both levels and serves a purpose in tying the two levels together and explaining how the economy interacts.The direction of the power exchange between the jth MG and Disco is indicated by the sign of P D j,t,s in the first term of (1).The selling of power to MG j is shown by the positive values of P D j,t,s , and the purchase of power from MG j is indicated by the negative values of P D j,t,s .If there is no power exchange, P D j,t,s will have no value.Equations ( 2) and (3) illustrate, respectively, the amount of power exchanged between Disco and the wholesale market and the price range of electricity exchanged between Disco and MGs.The cost of exchanging electricity with Disco, the cost of DG power generation, and the cost of load shedding comprises three parts of (5), which defines the objective function of each MG.Here, any MG may form a contract with a consumer.In order to do this, a bid-based load shedding process is being developed, in which customers submit their bids for the maximum amount of load shedding and the corresponding hourly pay.Equation ( 6) represents the expected value of the MG power exchange with Disco; this variable is also used in the first level.Equations ( 7) and ( 8) express, respectively, the power generation limit of the DGs included in the MG and the power exchange limits of each MG with Disco.One of the important influential variables in deciding the local prices is the rate of change of DGs' power production; the corresponding constraints are provided in ( 9)- (12).The allowable load reduction (interruption) value is displayed in Equation (13).
The charge and discharge rates of the ESS, which is referred to here as a battery, are constrained by ( 14) and ( 15).The battery cannot be discharged beyond a certain or charged above its capacity, according to (16).Each time interval's battery energy is a function of the energy stored in the preceding interval, as indicated in (17), and (18).Also, constraint (19) has been applied to determine the validity of studies over a 24-h period.It should be mentioned that it is impossible to charge and discharge the battery simultaneously.Given that the load points in an MG are close to one another in the current model, each MG is considered as a bus whose power balance equation is defined as (20).According to (20), the entire generation of DGs and WTs, plus the injected power of the ESSs and the electricity acquired from Disco, equals the power demand after nonvoluntary load shedding and the power required to charge the battery.In addition to the nonvoluntary load interruptions included in the planning process, the price-based DRP (load shifting) is taken into account here.According to (21), MGs can raise or reduce their power usage by Inc or Dec per hour according on the price of energy.It should be emphasized that, the total power consumption will be constant for a set period of time (24 h), as shown by (22).In ( 7)-( 21), the variables  1 j,t,s - 21 j,t,s are dual variables of the second level.

LINEARIZATION OF THE TWO-LEVEL MODEL
The problem formulation is done in the previous section.The current two-level problem is nonlinear due to the existence of the c D t P D j,t,s term.One method for solving the problem is to calculate the optimal value of c D t in the first level and input it as a parameter in the second level.Because the problem of the second level is linear, continuous, and convex in this case, the optimality conditions of KKT can be applied.However, the single-leveled problem utilizing KKT conditions is mixed-integer nonlinear.Linearization of the problem can be done using dual theory and considering that the answer is the same at the optimal point of the original and dual problems.Details of the linearization of the equations using the KKT optimality conditions are given in Appendix.Therefore, the linearized single-level problem will be as follows:

Wind speed uncertainty
Uncertainty in wind speed can be modeled using the Rayleigh PDF.The relevant PDF is broken into five parts for this purpose as shown in Figure 4. 4 Each scenario's wind speed is equal to the average value of the function in that phase.The probability of each wind speed scenario  WT s is also estimated by integrating the aforementioned PDF, defined by (43).
The scale factor of the PDF and the wind speed, respectively, are represented by the v and c in the Equations ( 42) and (43).Additionally, the scale factor c can be determined as follows assuming the average wind speed is known: Finally, the predicted production power of the wind turbine in each P WT s scenario is calculated as follows: In ( 45), the values of v i , v o and v r correspond to the cut-in, cut-out and rated speed of the wind turbine, respectively.Also, P r represents the rated power of the wind turbine.

Demand uncertainty
The normal PDF is divided into seven sections, and these sections are used to generate power demand scenarios. 26The following equation can be used to determine the probability of each scenario.
The average power demand predicted for each interval is represented  d in (16).Additionally,  d indicates the probability distribution function's SD, which in this case is taken to be 5%.Figure 5 shows the normal PDF of the demand for each MG.
The set of S j for MGs without WT simply contains scenarios pertaining to the uncertainty of the electric load.However, the scenario tree method is used to apply a total of 5 × 7 = 35 scenarios for MGs that contain WT.It should be highlighted that the solution to the problem is not contingent upon a particular scenario's execution; rather, the problem's final results can only be observed during actual operation.Therefore, it is necessary to determine the planned values of the variables after solving the optimization problem so that we face the minimum cost for future corrective actions (buying more power from the wholesale market, producing more DGs, etc.).Consequently, minimizing the expected value of the objective function for all potential scenarios will be an acceptable and basic approach to solve the problem, taking into account that the probability of each scenario reveals the degree of relevance and feasibility of its realization.

NUMERICAL STUDY
The proposed strategy is put into practice in this part on a DN that includes three MGs.MG 1 has DG, wind turbine, and battery storage while MG 2 and 3 just have DG units.Table 1 details the MGs' power sources and unit information.Figure 6 depicts each MG's power requirements.The maximum non-voluntary load shedding per hour for each MG is equal to 10% of the MG's electricity demand.Figure 7 depicts the wholesale market's hourly electricity price curve and the price of load interruptions.Disco and MGs can exchange power for no more than $90 per MWh.The maximum allowed power transfer between MGs and Disco is 10 MW, while the highest allowed power purchase from the wholesale energy market by Disco is 50 MW.Using GAMS software and the CPLEX solver, the suggested bi-level method is implemented as a mixed integer linear problem (MILP).
The investigations are initially conducted without consideration of responsive loads.The results of two-level and concentrated modes are contrasted with one another and then, in the third scenario, DRPs for MGs are taken into account.

Bi-level without DRP
The power balance diagrams of each of the MGs are shown in Figure 8. Also, Table 2 shows the local electricity prices per hour.Looking at Figure 5, it is clear that the cheapest electricity rates occurred during hours 2 and 3, when there were the greatest forced power outages.The negative value for P ex indicates the sale of power, and its positive values indicate the purchase of power by the MG from Disco.MGs and Disco trade power between the hours of 13 and 16 at the highest price permitted.In this case, MG 3 is compelled to purchase power from Disco due to the fact that it utilizes the maximum DG capacity (which has a lower price than the wholesale market pricing).However, MG 2 has the best chance of selling power to Disco at the highest price due

F I G U R E 9
The charge/discharge state of the battery using bi-level method.
to its maximal low-cost DG generation.In general, all three MGs operate their DGs at full capacity between 13 and 16 h.In general, all three MGs operate their DGs at full capacity between 13 and 16 h.Considering that Disco seeks to maximize its profit, it is obvious to provide power at the maximum possible price.The cost of the load reduction is matched by the local power price at 12 and 17.This suggests that the ideal plan is to implement power outages during these hours due to the rise in power consumption because there is a chance of a power outage (according to the 10% allowed ceiling).Each MG will attempt to sell power to Disco during the hours when the local price of energy is higher than, or at most equal to, the price at which the MG generates electricity.Each MG will attempt to sell power to Disco during the hours when the local price of energy is higher than, or at most equal to, the price at which the MG generates electricity.For example, in hours 3-5, the local price of electricity is equal to the price of MG DG generation 1.During these hours, MG 2, which has the lowest for a DG production unit, is at its maximum capacity, and more power demand can provided by the DG of the MG 1.
Figure 9 shows the status of charging, discharging, and stored energy in the MG 1 battery.It is clear that after optimization, energy storage has been done in cheaper hours and power injection into the network to provide power in more expensive hours.This issue can also be seen by paying attention to Figure 8.

Centralized without DRP
In the centralized model, the Disco is responsible for the entire scheduling process.Disco simply does this without considering the distinct objectives of MGs.In such a case, according to Figure 6, the prices of the power cut contract in hours 2 and 3 are lower than the price of electricity in the market, so we see load shedding in all three MGs according to Figure 10.Furthermore, there is a strong correlation between the wholesale market price of electricity (Figure 7) and the power purchase chart (Figure 11).For instance, during the first 3 h of the day, when electricity costs are lowest, Disco makes more wholesale purchases.Also, in hours 6-9 and 23-24, Disco buys almost no power.In addition, the Figure 12 shows that the charging and discharging status of the battery is also directly affected by the price of electricity in the wholesale market.According to the results of studies of 2 two-level and centralized mechanisms, the maximum energy injected by the battery in the centralized mechanism is 3.023 MW and in the two-level mechanism it is 4.03 MW.Therefore, storage plays an important role in optimal utilization in the two-level mechanism.The centralized mechanism relies less on storage capacity, and its planning is more focused on the price of electricity.Also, the total load cut in both cases has a slight difference.So that the total load shedding in hours 2 and 3 in both modes and for all MGs is about 1.2 MW.The total production power of DGs in two-level mechanisms in MGs 1, 2, and 3 is equal to 79.6, 85.6, and 124.6, respectively.These values in the centralized mechanism are equal to 71.5, 73.75, and 123.94 MW, respectively.As expected, the third MG's DG, which has the lowest power generation unit price, recorded the highest power generation in both situations.

Bi-level with DRP
From looking at two-level and one-level mechanisms, it is clear that in the two-level structure, each MG can use more of its own independent resources.Also, the total power purchased by MGs from Disco has decreased, which shows the independence of MGs and their less reliance on the upstream grid.However, in the two-level mode, the total operating F I G U R E 10 MGs scheduling results from centralized mechanism.

F U R E 11
Power purchased from wholesale market by Disco in centralized mechanism.
costs are $12,212.83,but in the centralized operating mode, the total operating costs are $11,282.7.This is due to local prices generally being higher than wholesale market prices.In this scenario, the DRP for MGs is considered.So that MGs 1 and 3 can move (decrease or increase) load up to 8% and MG 2 up to 5%.The power demand curves before and after applying load response are shown in Figure 13.Also, the power balance curves of MGs are shown in Figure 14, and charge-discharge states of the battery is shown in Figure 15.According to the results summarized in Table 3, the production of DGs of MGs 1 and 2 has increased by more than 25% and 47%, respectively.On the other hand, the total power purchase of MGs from Disco has decreased to 18.53 MW.The significant

F I G U R E 12
The charge/discharge and energy state of the battery using centralized method.

F I G U R E 13
Power demand curve of microgrids.Solid: initial power demand, dashed: power demand after applying demand response program.

F I G U R E 14
The charge/discharge and energy state of the battery using bi-level method and demand response program consideration.

F I U R E 15
The charge/discharge and energy state of the battery using bi-level method and demand response program consideration.

F I G U R E 16
Load curtailments of the microgrids over a 24-h period in Cases 3 and 4. increase in involuntary load shedding is important in applying DRP.In other words, using responsive load capacities, MGs have shifted their power usage to times when load shedding is more cost-effective based on existing contracts.However, in this case, the total operating costs of the system reached $9455.97.In other words, the costs of the system have decreased by 22% compared to the case where there is no DRP.Also, using DRP, the two-level method has lower operating costs than the centralized operating mechanism.In this situation, by applying only 8% of the possibility of load shifting in MGs 1 and 3 and 5% in MG 2, the operating costs have been improved by 16% compared to the centralized planning mechanism.An interesting point to note in this situation is the situation of local electricity prices.By applying DRP, C D j by changing the power consumption pattern of MGs is within the limit of the price of power exchange ($90/MWh) at all hours.In this scenario, while purchasing power from the wholesale market is reduced to 18.53 MW, we see an increase in the sale of power to the upstream network to 59.34 MW.

Analyzing the effects
In this case, the potential for load shifting in MGs has been increased by 15% while maintaining the problem conditions described in Section 6.3.Here, studies indicate that the total cost of operating the system is $9628.30.Although this amount is approximately 21% better than when the DRP is not considered, it is greater than when the possibility of load shift is considered for MGs 1, 2, and 3, which is equal to 8, 5, and 8%, respectively ($9455.30).The expenditures associated with the penalty of load curtailment can explain this discrepancy of less than 2%.Since MGs seek to maximize their profits, it stands to reason that they would schedule the greatest possible amount of forced load reduction in conjunction with the load shift, taking advantage of the available capacity for the load shift during the times of day when the selling price of electricity is highest.Figure 16 depicts how load curtailment occurs over the course of 24 h for two scenarios that include the DRP.While the load interruption pattern in both scenarios matches the price change curve in Figure 6, there are minor variations in the load curtailment values at various hours.

CONCLUSION
This paper investigated and implemented a two-level approach for ALEM scheduling.Uncertainty in wind unit production and power demand, the presence of energy storage, and the consequences of DRP were all investigated.According to the findings of the studies, total operational costs have increased in two-level planning when the objective function of MGs is with the minimization method compared to the centralized mechanism.However, using the DRP with a ceiling of 5%-8% of the load shift, the total daily operating costs of the system improved by about 26 percent compared to the two-level mode without the DRP.An important issue in applying the DRP is increasing the peak to average ratio of the power demand in MGs.This issue is related to load shedding price contracts.In the second level of the problem, MGs manage their power demand to reduce their total costs.Despite the considerable rise in involuntary load interruption, the 16 percent reduction in daily operating costs compared to the centralized mechanism indicates that the system's behavior has improved from the perspective of the objective function.This shows the critical impact of load shedding prices.Also, with the implementation of the DRP, the total power sale to the upstream network is increased by about three times and has reached 59 MW from 20 MW.Further research is needed to determine precise and appropriate boundaries for the demand side participation ceiling, which was shown by comparing the impacts of the DRP with the potential of greater free load shift.

F I G U R E 4
Wind speed Rayleigh probability distribution function.

F I G U R E 5
Normal probability distribution function of the microgrids' demand.TA B L E 1Microgrids' (MGs) resources characteristics.27 the first level of the problem Y = ]are the decision variables related to Disco and MGs, respectively.
Power demand of the microgrids.Power exchange price.MGs scheduling results from two-level mechanism.Local electricity price with bi-level method.
F I G U R E 7 TA B L E 2 Comparisons of various operation approaches of multi-microgrid distribution network.