**4. Case Study**

In this section, the effectiveness of the proposed method is verified through case studies employing the Korean power system. As the Korean power system does not have any interconnection with neighboring countries, its frequency control is controlled only by its own operating reserves, including the governor responses and AGC frequency controls. However, their effective operation has increasingly become a concern with the increasing penetration level of RES.

The dynamic data of the Korean power system are obtained from the planning database of the power utility [17], which is maintained in the format of PSS/E. The operating conditions of the power system are assumed to be a set of typical values used in the planning database, where the total electricity demand is 94 GW, provided by 276 generators. Among the 276 generators, all the generators except for 27 nuclear generators provided the governor response, and 143 generators provided the AGC frequency control.

#### *4.1. Verifocation of the Simulation Model for AGC Frequency Control*

This case study reviewed each process function of the simulation for the AGC frequency control according to Figure 5 to verify the effectiveness of the proposed simulation model. Figure 10 shows the filtered frequency, which is used to calculate the ACE in the proposed model assuming a generator trip.

**Figure 10.** Comparison of frequency and filtered frequency using low-pass filter.

An efficient and stable response of the AGC frequency control can be achieved by using the filtered frequency for calculating the ACE.

The function of the dynamic deadband is verified by comparing the operating results of the AGC frequency control with and without it. Figure 11 shows the simulation results for the comparison in terms of the ACE and controlled frequency.

**Figure 11.** AGC frequency control with and without dynamic deadband.

As shown in Figure 11, the dynamic deadband delays the activation of the ACE when the frequency drops. However, the performance of the frequency control is not deteriorated by the delayed activation of the ACE because the nadir of the controlled frequency is the same as that in the case where the ACE is instantly activated without the dynamic deadband. Therefore, it would be efficient to operate the AGC frequency control with some degree of delay by using the dynamic deadband during a transient period, as even an instant response of the AGC frequency control would not be able to contribute to the frequency control owing to the overlap with the governor responses immediately after the disturbance. The detailed simulation results of the proposed model are shown in Figure 12.

**Figure 12.** Verification of dynamic deadband activation.

As illustrated in Figure 12, the ACE is not activated immediately after the frequency drop because of the dynamic deadband. However, if the frequency is not recovered until its ACE exceeds the decreasing dynamic deadband, the ACE is activated and its amount is adjusted by the control gain depending on the control mode. Once the frequency is recovered and its ACE is smaller than the static deadband, the dynamic deadband is initialized. The activated ACE is allocated to each generator as the AGC frequency control target. Figure 13 shows an example of ACE allocation using the proposed simulation model.

**Figure 13.** *Cont*.

**Figure 13.** Example of area control error allocation using the proposed simulation model (**upper**) and comparison of between the total allocated area control error and pre-allocated area control error (**lower**).

As shown in Figure 13, the amount of activated ACE depends on the frequency drop, and it is allocated to each generator using the proposed simulation model.

In the AGC frequency control operation, the allocated ACE is received by each generator through its PLC so that the governor response could be prioritized in the frequency control. In this study, the PLC of the proposed model is verified by applying the contingency of the generator trip. Figure 14 illustrates the simulation results of the PLC function.

**Figure 14.** Revised AGC target by the plant-level controller function.

As illustrated in Figure 14, once the measured output of the generator is different from the received AGC target, the PLC is activated to revise the received AGC target by adding a governor response to it. Therefore, the output of the generator follows the revised AGC target rather than the received target.

Once the AGC target is received by each generator, it should be met with a response from the generator even considering its PLC function. The performance of the proposed simulation model is verified by comparing the output of the generator and the AGC target in terms of the appropriateness of the AGC allocation and generator dynamics. Figure 15 shows the AGC response of the generators with the highest and lowest amounts of ACE allocation.

**Figure 15.** AGC frequency control response of GEN#1 (**upper**) and GEN#2 (**lower**).

As shown in Figure 15, the amount of ACE allocated to GEN#1 is smaller than that allocated to GEN#2. This is because GEN#1, a coal generator, has a lower performance than GEN#2, a gas generator. As the output of the generators could satisfy the revised AGC target, the appropriateness of the AGC allocation could be verified by considering the dynamics of the two generators.

To validate the proposed model, the simulated results are also compared by the selected actual frequency responses in the Korean power system. The real-event data was recorded when a nuclear generator of 1 GW tripped in 2013. Since there must be differences in the detailed operating conditions of the real event and the simulated case, those should be compared only in terms of AGC frequency control. Figure 16 compares ACE calculations of the measured and simulated case to the same frequency recorded in the real case.

**Figure 16.** Comparison of the measured and the simulated ACE.

As shown in Figure 16, the simulated ACE was matched with the measured ACE and this validates that the proposed model could be used to simulate real ACE calculations by appropriately setting the control parameters.

In addition, AGC command dispatched to a specific generator and its response is compared using the same event as shown in Figure 17.

As shown in Figure 17, the time delay in the AGC activation that occurred in the measured data could be also found in the simulated results. This validates that the proposed model could simulate the time response of the real AGC operation.

**Figure 17.** *Cont*.

**Figure 17.** Comparison of the AGC activation time between the measured (**upper**) and simulated (**lower**).

#### *4.2. Analysis of Tuning AGC Frequency Control*

In this section, case studies using the proposed simulation model for analyzing the effects of setting the deadband filter of the AGC frequency control, which should consider both the dynamics and AGC operation of the power system, are described. As the proposed model simulates the frequency performance of the power system synthesizing the governor response and AGC operation, the effect of parameter setting can be analyzed by considering not only the effect of the parameter itself but also the coordination with the governor response.

The deadband filter has three control parameters: variable time, the boundary of the control mode, and the control gain at each mode. The variable time delays the activation of the AGC frequency control for coordination with the governor response. Figure 18 illustrates the simulation results of the frequency control with different variable times of the deadband filter in the AGC operation.

**Figure 18.** *Cont*.

**Figure 18.** The area control error activation time (**upper**) and the frequency response of generator (**lower**) depending on variable times of the dynamic deadband.

As shown in Figure 18, the ACE activation time is controlled by the variable time of the dynamic deadband, and it affects the frequency response of the generators and the frequency of the power systems. Although the frequency recovered faster when the variable time of the dynamic deadband was reduced from 32 to 8 s, the AGC frequency control could not improve the nadir frequency. This appears to be because the generator power was saturated with the governor response, as shown in the lower figure in Figure 18. Accordingly, the proposed model can be used to set the variable time of the dynamic deadband appropriately in AGC operations.

The boundary conditions of the control modes are used to determine the status of the power system depending on the amount of ACE. In this case study, three cases are assumed for the proposed simulation model with different sets of boundary conditions, as summarized in Table 1. As one goes from cases 1 to 3, the range of each mode becomes wider and the emergency mode appears later. For analyzing the effectiveness of boundary conditions of the control modes only, the variable time of the dynamic deadband is set as 32 s, and the control gains are set as 1.0 in all three cases.


**Table 1.** Bounds of the control mode (ACE: Area Control Error; Static DB: Static Deadband).

Figure 19 shows the simulation results using the proposed model for the three cases and the distribution of the control mode determined every 2 s during the simulation.

**Figure 19.** Comparison of frequency response with different bounds of control mode (**upper**) and the distribution of the determined control mode (**lower**).

As shown in Figure 19, the activation of the AGC frequency control is also delayed by setting the ranges of the control modes wider. Among the three cases, the boundary condition of case 2 appears to be relatively more reasonable considering the frequency drop because the frequency could recover earlier without the emergency mode. In the case of 1, the nadir frequency could not be improved even by the more frequent emergency modes.

Accordingly, the proposed model can be used to set the boundary conditions of the control modes appropriately in the AGC operations. It can also be observed from Figure 19 that the nadir frequency is the same in all three sets of boundary conditions, but the frequency recovers earlier with narrower ranges of the control modes.

The control gain of each mode determines the strength of the AGC frequency control. In this study, three cases are assumed with different sets of proportional control gains, as summarized in Table 2 where the boundary conditions of the control modes are assumed to be the same as those in case 2 of Table 1. As one goes from cases 1 to 3 in Table 2, the proportional control gain of each mode increases. The boundary conditions of the control modes are assumed to be the same as those in case 2 of Table 1.


**Table 2.** Gain setting of each control mode.

Figure 20 shows the simulation results using the proposed model for these three cases.

**Figure 20.** Comparison of frequency responses with different control gains.

As shown in Figure 20, the frequency recovers faster with higher control gains of the AGC frequency control. However, the high control gain in the AGC frequency operation can cause an overshoot of the power system control. Accordingly, the proposed model can be used to set the control gains of each mode appropriately in the AGC operations.

As shown in the results of the case studies, the proposed simulation model can be used to tune the control parameters of the AGC frequency control considering both the dynamics and AGC operations of the power system. This is one of the benefits of developing a simulation model for the AGC frequency control based on the dynamic models of the power systems.
