5.1. Case Description
An integrated version of the IEEE 118-bus power system and a 10-bus natural gas system is employed to demonstrate the proposed method and analyze the impacts of natural gas system integration on the operation of the power system [
34]. In this paper, the acceptable maximum risk level (probability) is specified as 0.001, and the required minimum healthy state probability is specified as 0.99 [
25]. Two simulation scenarios with different GT installation proportions are arranged: (1) the considered power system contains 34 generating units (25 conventional units, six GT units, and three VRES units); (2) the considered power system contains 38 generating units (19 conventional units, 16 GT units, and three VRES units). The allocation of generating units and power loads is depicted in
Figure 4. L1–L4 are the residential natural gas load, whereas L5–L20 are consumed by GT units when allocated. The installation capacities of different generating units can be found in
Table 1. The natural gas load curve, electricity load curve, and total VRES generation forecast curve are identical in both scenarios, as shown in
Figure 5. The daily natural gas production limit is 1100 MMcf, and the equivalent hourly natural gas transmission limit is 50 MMcf. The contingency gas upper limit is set at 55 MMcf. In 2015, the average heat content of natural gas was about 1032 Btu per cf [
35]. The installation proportions of GT units in the two scenarios are (1) 11.33% and (2) 30.21%; the efficiency ratio of a GT unit is assumed to be 50%. The WBA is conducted on a typical day (24 h). The period lead time
T is set as 1 h, and the probability of a GT unit start-up failure is set to 10%. The ORR of conventional units is two failures per year, and the transfer matrix of the GT unit is demonstrated in
Table 2.
5.2. Results and Comparisons
The fact that GT units are employed as low-cost generating units keeps them from being competent reserve providers. In fact, both the probabilities of risk and healthy state do not meet the required levels, no matter whether qTC is considered or neglected in scenario (1). The results confirm that the power system cannot rely on a small number of GT units as reserve providers without specific planning. In scenario (1), the GT installation capacity is actually sufficient for the required reserve, but the GT units will be committed prior to most conventional units as there is always surplus natural gas. When the vast generation capabilities of GT units are scheduled by unit commitment, the equivalent reserve capacity they can provide is thus limited. When the natural gas production rate varies, the GT consumption margin will be further compressed. In the worst case scenario, when the natural gas production rate is extremely low and the overall residential natural gas load is high, the consumption margin available for the GT units may not even be sufficient for the reserve requirement. Therefore, the production and consumption rates of natural gas should be examined for a power system that depends on GT units for reserve provision, and other reserve sources should be equipped if necessary.
In scenario (1), the GT units’ functions as low-cost generating units and as peak/reserve units will conflict due to their limited capacities. When GT units are taken as reserve providers, one way to guarantee the sufficient reserve capacity is to keep a certain number of GT units from the energy commitment. The probabilities of risk with different numbers of reserved GT units are shown in
Table 3. It can be found that the probability of risk decreases as the number of reserved GT units increases. The probabilities of health with various number of reserved GT units are shown in
Table 4. From
Table 3, it can be found that, regardless of the number of reserve units, the risk probabilities are distinct at some moments, especially for hour 6. This is because the specific energy production and consumption conditions at hour 6 make the unit commitment at that moment different from at other hours. From
Figure 5a, the residential natural gas consumption rebounds after hour 5 and the bounce rate is the highest at hour 6, which indicates that the committed GT units are likely to shut down so that the natural gas consumption of the residential load can be guaranteed. However, they are still able to provide reserve services under the contingency offers if there is a reserve gap threatening the system’s well-being. At the same time, more conventional units will be committed as the electric load is also going to rise according to
Figure 5b, which strengthens the total reserve capacity of the system. Thus, the risk metric is considerably lowered at hour 6 due to the start-up of plentiful conventional units, large reserve capacities provided by GT units, and the relatively low load level. However, this does not indicate that the reserve is unnecessary at hour 6. In fact, the health metric at hour 6 does not meet the requirement unless five GT units are providing reserve services. In general, the abnormal probability value at hour 6 is an occasional event that is the result of many factors and cannot guarantee the well-being performance even for minor fluctuations. In scenario (1), the
qTC constraint does not affect the WBA results, which indicates that the fuel scarcity and transmission constraints will not influence the operation of GT units when the installation proportion of GT units is low and the interconnections between the electric power and the natural gas system are loose.
When five GT units are reserved, the well-being states of the power system will meet the proposed requirements, as shown in
Table 5. It should be mentioned that when the actual output of VRES decreases by 85.56 MW from the forecasted one, the results in
Table 5 remain the same, indicating that the reserved GT units are sufficient to cover the impact of the intermittent VERS without extra reserve cost.
Table 6 shows the results of WBA with and without the natural gas contingency constraints considered in scenario (2). In this case,
qTC is of great significance to the well-being of the power system. The differences in WBA results reflect the impacts of fuel scarcity and transmission limit on the operation of GT units. For certain operation periods (e.g.,
t = 11, 12, 18…), the results in
Table 6 present significant gaps, indicating that the fuel scarcity and transmission limit is hindering the ordinary operation of GT units. Limited impact of fuel constraints can be found at periods such as
t = 8, 10, 15 and 16, whereas the impact becomes trivial at periods such as
t = 1, 2 and 7. The availability of GT units is another factor that affects the well-being requirements of the power system. It is worth noticing that, when
qTC is neglected, the well-being performance still cannot be completely fulfilled throughout the observation period, which is due to the high percentage occupation of installed GT units as electricity suppliers. Though the GT penetration level under scenario (2) is significantly higher than under (1), the system operator may still run out of reserve capacity during the original valley of the natural gas demand, as the installed GT units are scheduled due to the natural gas surplus. Similar to the analysis for scenario (1), the system well-being can be improved through discounting some economic benefit for risk aversion. Therefore, both the fuel insufficiency and the availabilities of GT units should be taken care of to simultaneously reduce the energy supply risk and enhance the system’s well-being level.
To enhance the well-being level of the energy system operation, the conventional unit commitment planning should be revised considering the installation capacity of GT units and the fuel constraints. Denote the reserved unit number as
nr, the suggested generation capacity of GT units for unit commitment denoted as
as shown below, will replace
ρ1.
Obviously, the total generation cost will increase if
<
ρ1, but the improvements in well-being level are far more significant. By setting
nr = 4, the suggested GT capacities for unit commitment are illustrated in
Figure 6. The corresponding system well-being indicators are shown in
Table 7.
As can be observed from
Table 7, the acceptable risk level (below 0.001) can be fully reached, and the required healthy state probability (above 0.99) is also accomplished. The system well-being performance will not be affected by a VRES fluctuation of three times the standard deviation as well. As the failure rate and transition probability of GT units are much higher than other generators, the risks they introduce (especially when the installation capacity is high) must be dealt with carefully. From the above analyses, proper revision of GT unit commitment schedules will make the GT units competent not only to mitigate their own risks, but also to cover the demand of system reserve capacity.