*6.2. Comparison Analysis*

The U.S. plans to reduce CO2 emissions from fossil fuel power plants by 32% before 2030 [39]. Thus, in this study, we assumed that emissions should ideally reduce by 32% when compared with an optimal operation situation of DG without consideration of emissions. Table 4 shows the optimal operation results when operating the MG with only DG units. The total operation costs and pollutant emissions are \$101,065.60 and 5064.76 kg, respectively. Thus, maximum emission is set at 3444.04 kg, which is 32% of the pollutant emissions from DG units.

#### **Table 4.** Results for DG.


To reduce operation costs and to avoid the violation of the maximum emission constraint, the proposed EBDR and WU-ABC algorithms are used for solving the CEED problem. The penalty factor should be set taking generation costs and the emissions of each generator into account. Various types of penalty factors are calculated and listed in Table 5. These factors vary according to the characteristics of each generator. In multiobjective problems, the penalty factor affects the relative importance of different objective functions, because it affects the optimal solution. In other words, a penalty factor that is excessively large may overestimate the effect of one objective function in a multiobjective problem. Thus, a min-max relationship with a small average penalty factor is chosen to prevent predominant effects of only one objective function, and the penalty factor is iteratively updated from the initial value to find the best solution that satisfies all constraints.

**Table 5.** Various price penalty factors of each DG unit.


In this study, three cases are considered to validate the proposed approach. Case 1 is a conventional CEED proposition without a DR program and penalty factor update. Case 2 involves a conventional economic DR program but does not include penalty factor updates. In Case 3, the proposed EBDR is used and the penalty factor is updated through the WU-ABC algorithm. All cases are simulated under the same constraints, except for the DR programs and penalty factor update.

Figure 4 shows the hourly scheduling of optimal power generation for all cases. To reduce operation costs, the MGO sells power for utility when market prices are high, in all cases. Although the generation cost of G1 is higher than that of the two other generators, it generates the most power due to low pollutant emissions per MWh. As shown in Figure 4a, Case 1 does not shift the peak demand; operation mainly occurs through the power generated by DG units, because no DR program is considered. Compared to Case 1, Case 2 utilizes a conventional economic DR program and demonstrates demand shifts in consideration of elasticity according to market prices. As shown in Figure 4b, the MGO uses the DR program to shift demand from periods of high market price to periods when power is relatively inexpensive. Thus, the power generation of DG units in the MG is reduced, and the operation costs are reduced despite considering incentives for the DR participants. However, in Case 2, no maximum emission constraints are imposed; to remedy this problem, penalty factor update and EBDR are considered in Case 3. The results for Case 3 are shown in Figure 4c. Compared to Case 1 and Case 2, it can be noted that the period taken by the MGO to sell power is reduced and the number of participants in the DR program is increased. Moreover, the power generation capacity of G2 and G3 (which have relatively large pollutant emissions per unit power generated) is also decreased. Case 3 could satisfy the maximum emission constraint.

Table 6 shows the optimal CEED results for each case. Compared to the results in Table 4, which are essentially optimal operation propositions without environmental considerations, it can be seen that Case 1 has marginally higher operation costs (by 1.1%) but shows significant reduction in pollutant emissions (by 26.1%). These results violate the maximum emission constraint and there is no significant reduction in operation costs. For Case 2, the MGO reduced the operation costs of DG by using an economic DR to reduce peak demand. Thus, additional DR incentive costs are incurred, but the total operation costs are reduced by selling more power during high market price periods. The operation costs are reduced to \$98,187.99, and pollutant emissions are decreased to 3571.14 kg. These results show that scheduling under a DR program for MG operation can have a positive impact, leading to reductions in peak demand, operating costs, and pollutant emissions. Compared to Case 1, Case 2 reduces both operation costs and pollutant emissions, but it does not satisfy the maximum emission constraints. Therefore, the MGO pays \$805.81 as a violation fee, which is proportional to the excess pollutant emissions. In Case 3, the proposed EBDR and penalty factor update method are used to satisfy the maximum emissions constraint. Compared to Case 2, the operation costs of each DG unit are reduced due to the reduction of the power generation capacity of the units. The quantity of power that is sold to the utility decreased. The cost of participating in the DR program increased with the use of EBDR. Consequently, operating costs increased by approximately 0.05% compared to Case 2, but the maximum emission constraint is satisfied, curtailing emissions within 3444.03 kg. As can be seen from these results, the proposed approach is effective in solving the CEED problem with a maximum emission constraint.

**Figure 4.** Optimal hourly scheduling in the MG for each case.


**Table 6.** Operation costs and pollutant emissions in each case.

Figure 5 indicates the penalty factor and pollutant emissions with respect to the number of updates. The penalty factor is updated proportionally to the excess pollutant emissions; consequently, the penalty factor update is performed 11 times, and the penalty factor is finally determined to be 23.55\$/kg. Table 7 summarizes the optimal CEED solution for all cases. Case 1 without a DR program presents a greater total operation cost and higher emissions than the other cases. Meanwhile, for Case 2 with the conventional economic DR program, the MGO is able to reduce operation costs and pollutant emissions relative to Case 1 but pays a violation fee for exceeding the maximum emission constraint. In Case 3, the maximum emission constraint is satisfied through the use of penalty factor update and the EBDR program; although it slightly increases the operating cost compared to Case 2, the constraint violation cost of \$805.81 is not incurred. Therefore, it can be concluded that the proposed approach exhibits the most promising results for the CEED problem, when considering a violation fee.

**Figure 5.** Penalty factor update and emissions.

**Table 7.** Summary of results for all cases.


To demonstrate the superiority of WU-ABC, performance tests have been conducted through comparison with various algorithms [40,41]. For a fair comparison, the maximum emission constraints and violation fee are equally considered in grid-connected MG operation. A total of five algorithms are considered, such as ABC, improved multilayer artificial bee colony (IML-ABC) [41], weighting update genetic algorithm (WU-GA), weighting update particle swarm optimization (WU-PSO), and proposed WU-ABC. WU-GA and WU-PSO are the addition of the weighting update method to the conventional GA and PSO. The simulation process for all algorithms is repeated 10 times, and then the results are calculated as average value. Table 8 shows the CEED results of each algorithm. It can be observed that WU-ABC reaches a best solution compared with other algorithms. The algorithms without the weighting update method do not satisfy the emission constraint, and as a result, the total cost is high due to imposing violation fees. In addition, WU-ABC indicates the lowest cost compared to other algorithms considered weighting update methods and takes the least CPU time. Therefore, these results reveal that the proposed WU-ABC algorithm is appropriate for solving CEED problems considering maximum emission constraints.


**Table 8.** Comparison results for various algorithms.
