Chapter 3: Methodology
3.3 Model Construction and Preliminary Analysis
important to emphasize that the North American electricity market is profoundly different from Taiwan. Each region within the US has a different configuration from that of Taiwan.
Some regions have entirely privatized, market based electrical supplies. Others, like the Tennessee Valley Authority are fully federally owned. In Washington State, there is publically owned transmission, mixed generation, and separate ownership of distribution. The most unique aspect of the North American grid compared to Taiwan is its size and that all grids are interconnected. When there is a shortage in one grid, they can buy power from
neighbors through interties. Policies such as operating reserve and grid composition are set by each region and utility in accordance with local laws. In spite of these differences, GADS provides a rich data source that we can use as a basis for comparison.
Taipower’s Power Dispatching Department keeps figures on historical availability of all Taipower sources for many years. Those these data sets are not public; they are
available on request. Some of this data can be used to do a comparison with GADS.
These reliability indicators can be compared with regional indicators from the US.
The North American Electric Reliability Corporation maintains extensive records on the reliability of US generation through the Generating Availability Data System (GADS) - a data source which is available to the public. The author can then compare the
3.3 Model Construction and Preliminary Analysis
Taiwan’s long term power projections may be misleading. Although the MOEA 2014 report on long term electricity growth suggests that installed capacity will continue to grow, most of the growth the report documents will be in renewables with extremely low capacity factors. Taipower’s Long Term Growth report also suggests a similar, if more serious power shortfall. The full data sources for these reports are not available.
‧
To answer the question, “how much reserve does Taiwan need?” I have developed a prototype metric - Average Individual Availability - derived statistic for evaluating reliability of individual generation systems. Its formula is simple:
Equation 3.3.1
Average Individual Availability =
%
This metric tells us, on average, how much capacity is available according to Taipower’s metrics of availability. Those metrics often drastically exceed capacity factors - especially for wind power - so we can conclude that availability refers to maintenance, hydrology, transmission and operational limits, and faults. It may also include environmental restrictions. Taipower’s availability statistic is calculated as follows:
Equation 3.3.2
可用率＝〔１－（影響供電量／參考容量／全期時數）〕Ｘ 100 ％
1 / / 100%
Where: IGC=Impact to Generating Capacity, NMC=Net Maximum Capacity, PH=Period Hours
This metric has a number of serious flaws, but it still enhances our understanding of Taiwan’s electrical supply reliability by providing us with a baseline probability of when power is available from traditional sources. In phone interviews with the Taipower Electrical
Dispatch section, their officers confirmed Taipower’s Availability Factor is impacted by situations like maintenance, thermal variations, hydrology problems (for hydro), breakdowns, transmission problems, fuel problems, or other technical issues that could influence supply
‧
availability. They did not explain exactly how this was calculated for each source, and this will remain a major weakness of this study.NERC uses a collection of statistics, but the most important for the purposes of this study is the Effective availability factor. Although it has a different calculation procedure which is much more transparent than Taipower’s number, it still provides an availability percentage which accounts for both planned and unplanned outages:
Equation 3.3.3: Effective Availability Factor
100%
EAF: Effective Availability Factor AH: Available Hours
EPDH: Equivalent Planned Derated Hours EFDH: Equivalent Forced Derated Hours EMDH: Equivalent Maintenance Derated Hours ESEDH: Equivalent Seasonal Derated Hours
Source: GADS Data Reporting Instructions, Jan 2015, NERC
These are all derived from GADS data. While it is unclear how exactly Taipower’s numbers are calculated, they are assumed equivalent to the GADS data for the purposes of this study. This assumption will require further study if this technique is to be widely
implemented as a comparative model. This assumption only affects this study where direct comparisons are made between NERC GADS data and Taiwan’s data, and does not affect the evaluation of Taiwan’s long term forecasts.
Capacity factors alone are an insufficient means of assessing reliability of high uptime sources and dispatchable sources because they never exceed demand, nor do they unambiguously tell us about the underlying reasons for lack of production. In my proposal, I provided evidence of days when Taipower had serious losses of capacity due to
maintenance and other unforeseen issues. This metric quantifies those events and provides general probabilities of failures based on historical data from Taiwan’s own experience.
‧
This metric has a number of weaknesses: It is extremely to apply to intermittent sources like wind and solar. For wind, this metric only evaluates that wind turbines are not malfunctioning and able to harness any wind resources available. It does not allow us to know when the wind blows to generate power. For solar, Taipower does not keep availability statistics, and this metric cannot be calculated at all. Even so, solar suffers the sameproblems that wind does: the availability of the power plant does not mean the availability of intermittent and uncontrollable resources. This metric also distorts pumped storage. Pumped storage facilities need to be charged, and the availability metric provided by Taipower only indicates that the facilities are operational, not when they are charged. For these sources, capacity factors will remain indispensable for power system planning. Though the initial analysis assumed 100% availability factors, but upon further consideration, using capacity factors may be more accurate. In the datasets, I call this average individual generation, and it is used when availability is thought to be unrepresentative.
Equation 3.3.4 Average individual generation was calculated as follows:
100%
A further problem comes from IPP facilities and Cogeneration. Taipower does not keep facility level statistics on private power plants. For the purposes of this study, I assumed that the reliability of these private thermal plants was comparable to those of Taipower facilities using the same source, and extrapolated the values based on the combined average individual availability for all Taipower thermal plants on all co-gen and IPP plants. This will no doubt distort the result, but due to limited time, assuming these facilities to have similar repair records still provide a useful analysis.
‧
Based on the author’s June 2015 analysis, Taiwan had an average availability of 37,960.8 MW. That means that at any given time, we could assume that just under 38GW of generating facilities would be “available” - that is, not broken, under maintenance, or
suffering other technical problems. This is overly optimistic because solar and wind sources have artificially inflated numbers. Capacity factors would be better analyses for these sources. Without wind and solar, the amount is 36,958.0 MW.
This means that, barring other factors such as political ones, Taiwan should have a consistent source of power up to this peak level. It may be possible to reach higher outputs, but based on this historical period, it is not likely, nor could it be relied upon. I think of it as a maximum theoretical reliable supply.
Availability data of any kind of IPP and Cogeneration is not available to this author's knowledge. In places where it is necessary, Availability rates for these sources are assumed to be comparable to Taipower’s availability for the same sources. This is both highly
generous, as Taipower’s availability rates are unusually high, and therefore also potentially inaccurate, skewing the IPP availability on the higher side. This is particularly true in the case of cogeneration, as the cogen system is not really firm power and produces only a small fraction of its reported installed capacity.
For IPP capacity factors, they are calculated only using yearly averages, since monthly data is unavailable. The total IPP generation and capacity factors were derived in several steps. First, the total IPP generation and total LNG generation were retrieved from the MOEA Handbook, then Taipower’s total LNG generation was subtracted from the LNG total. The remaining LNG figure represents IPP LNG generation for the year. From there, that is subtracted from the total IPP value. The remaining IPP value should contain coal and IPP solar. IPP solar values - acquired from data.gov.tw - are then subtracted, leaving us with the IPP coal value. The MOEA Energy Statistics Handbook value for coal is not used
‧
because Co-Gen might also be counted, and co-gen installed capacity is extremely difficult to nail down.Using a similar process, we can derive IPP solar and wind installed capacity and, with data.gov.tw availability can derive their capacity factors.
The true measures of reliability for Cogeneration have proven to be extremely difficult.
Cogeneration refers to boilers for industrial use - usually coal fired - which sell their surplus steam as wholesale electricity to Taipower. The total generation is listed by Taipower and data.gov.tw, but installed capacity is hard to know. Not all of these facilities are designed for power generation, and installed capacity, availability, and capacity are unclear from easily accessible data. The MOEA handbook implies that there may be more than 7GW of installed available, while Taipower’s real time data system lists less than 0.6GW - but by this
measurement they had capacity factors of more than 100% during 2015. In the quantitative analysis, this thesis calculates all cogen capacity factors twice - once with the MOEA installed capacity, and once with Taipower’s installed capacity. The difference in capacity factors is startling. Using the higher MOEA installed capacity results in average Cogen capacity factors around 12%, while the Taipower installed capacity ranges from 120%-190+%. Since the generated capacity is likely accurate, and GADS has no comparable data, we will have to settle for incomplete data in this regard. GADS does record co-generation capacity and availability data, but it is unclear how much installed capacity there may be, as GADS records a number of retirements for cogeneration, and no new installations as of 2014.
For the availability factor, since Co-gen is primarily coal-based, Taipower’s coal availability factors were used in the initial analysis, however, this proved to be
unrepresentative. Therefore, for this calculation this thesis has opted to use the same Average Individual Generation - based on capacity factor and installed capacity - similar to the metric for renewable source. However, without more information on how cogeneration is implemented, there is no way to tell which would be more appropriate. Further studies and data about cogeneration will be needed for a more thorough analysis. This method is, in the author’s opinion, the single greatest flaw in this analysis. However, without more accurate
‧
data about exactly how much installed capacity there is, and what kind of grid availability to expect, there is no way to make a more precise analysis.Finally, there is no availability data for solar - IPP or Taipower. Availability data therefore uses solar capacity factors instead. Since solar-PV panels themselves have no moving parts, they are unlikely to be damaged easily. No data was found on whether these systems have sun-tracking capabilities, which could conceivably require more maintenance.
Capacity factor may be a more appropriate measure for wind and solar, since, unlike conventional sources, their output cannot be controlled. Additionally, it is reasonable to assume that renewable sources will be used whenever they are available, since they have no fuel cost.
Taipower’s wind systems do provide availability factors. However, those figures are not representative of available capacity. This analysis will therefore use capacity factors for both wind and solar. Using availability factor for wind does increase numbers somewhat, but this author feels that since that does not represent actual generating availability, this
measurement is not useful in a power system with limited storage.
3.5 Capacity Factor Vs Availability Factor
Capacity factors demonstrate different things for different sources. For dispatchable sources - those sources whose generation can be controlled (conventional sources, usually) - it demonstrates usage. A coal plant that is available 80% of the time, with a capacity factor of 40% tells us that this facility is able to produce more if needed - up to its nameplate capacity 80% of the time (or perhaps some other combination of time and capacity, depending on regulation and maintenance). However, for wind and solar, capacity factors generally represent their total availability. In most cases, power from these sources is used whenever it is available, since they have no fuel cost. In Taiwan’s case, its wind farms have availability factors well in excess of 80%, but very low capacity factors. These likely either reflect the low quality of wind resources, or the inability of the grid to accept them. Due
‧
to the flexibility demonstrated by Taiwan’s LNG-fired plants, this author is inclined toward the former - low quality resources.3.6 Discussion
From here, calculate all figures for summer months. Do an evaluation of spinning reserve and operating reserve? Compare this data to an EPRI region in the United State using GADS. Evaluate Taiwanese future demand and supply growth using these figures to counter balance traditional operating reserve. Devise minimum safe operating reserve based on this metric, along with a recommended reserve. Evaluate how Taiwan could develop renewables given this metric. Evaluate DPP, Taipower, and MOEA projects based on these figures.
3.7 Problems with Taipower’s Numbers
Taipower’s data has several errors. The first is that in all cases where percentages are totaled, either for availability or capacity factors, they are done by a simple average (total percentage points/number of generators) and are not weighted based on the nameplate capacity of the facility. This substantially changes the numbers. All calculations introduced in this paper have corrected for this error, and whenever percentages are used that represent multiple generators in aggregate, they will be weighted based on the installed or nameplate capacity of the generators involved.
Second, Taipower does not explain specifically how it’s reliability and capacity factors are calculated. NERC and GADS distinguish between net generation - the power that is put into the grid - and gross generation - the power that comes out of the generator. This difference may not be apparent immediately. To clarify, many power plants consume some of their own generated electricity to use for instrumentation, control systems, or any other number of needs. Therefore, net generation is generation after any usage internal to the plant, while gross generation includes both grid transmitted power AND power that was consumed inside the facility. The reason this is useful is to understand maintenance and efficiency procedures inside plants. Taipower, being essentially a monopoly, is responsible
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
38 for all consumption and most generation on its network, and so it may not be efficient or effective to track this data. Further study may be warranted, but this is well outside the realm of the paper.
Third: There were a number of cases where data was either corrupted, or data was input in an incorrect form. Specifically, in several sources many of the numbers included white spaces, which cause spreadsheet programs to treat the cell as text, rather than as numbers. Although this study corrected for that, it is possible that other studies would have been unable to do so.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
39 Figure 3.7.1: Taiwan Capacity factors by source, 2015. Data sourced from Taipower,
Data.gov.tw, and MOEA Energy Statistics handbook 2015
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
40 Figure 3.7.2. Taiwan Average availability by source (In KW) - 2015. Note that this uses Average Individual Generation instead of average individual availability for wind and solar sources.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
41 Figure 3.7.3; Taiwan Average Availability (KW) by source, using AIG for wind, solar, and cogen
‧
The first notable observation is that availability is generally higher during the summer.
Much higher. We can speculate that Taipower shifts it maintenance to winter months when demand is lower. Not surprisingly, capacity factors also peak during the summer months.
During 2015, much of Taiwan’s nuclear capacity was impacted due to maintenance problems. Taiwan lost an average of nearly 3 GW of generating availability due to nuclear outages. Nuclear Plant 1 Reactor One has been out of commission since December 2014 due to a minor malfunction - a loose handle on a fuel cask. The legislature has so far refused to accept the atomic energy council’s report on the incident that would allow the plant to resume operation. (Chen, Wei-han, 2016) What is perhaps most shocking is that during July and August, even though availability spikes slightly, likely due to maintenance planning, to 36,220 MW, peak demand during this time was 35,385 MW. That small difference of less than 1GW represents two large generators or one nuclear power plant.
This 3GW of nuclear outage represents the entirety of Taipower’s reserve for
outages of all causes. Any additional failures would result in power shortages at peak hours.
4.2 International Comparison
When compared with aggregate GADS data from the US, Taiwan’s power system reveals two major observations. The first is that Taiwan’s power system has, on the whole, a higher level of availability than similar systems in the US. The only exception was diesel generation, which was slightly lower than US generation availability. However, this may not be viable. Diesel generation in Taiwan only occurs as backup, on-site supply for large plants (which does not feed into the grid) or as generation for outlying islands. In the US, only
‧
companies operate grid connected diesel generators, and the sample size may be so small as to not be useful comparison. See Table X. Note that the most recent GADS data is 2014.However, US electricity consumption has remained relatively stable for some time as the US economy’s energy intensity
The capacity factors are somewhat more interesting. When compared to US figures we see that, with the exception of nuclear, ALL Taiwanese capacity factors are much higher than those of the US. Even considering the availability problems with Taiwan’s nuclear installations, they still contribute an impressive share to Taiwan’s capacity factors. Overhauls have allowed them to generate in slight excess of their original nameplate capacities.
In every field, Taiwan dramatically out produces the NERC facilities. In the case of Taipower’s coal facilities, they had a 90.25% capacity factor, and a 90.31% availability. That means that Taipower’s coal facilities were running at nearly maximum capacity for the entire time they were functional. Nuclear availability was only 54%, but generated at 66% capacity because of overhauls. The coal IPP facilities also run at extremely high capacity above 70%.
Taipower LNG runs at well above 60%, compared to the US 10%. IPP LNG runs over 51%.
This ordering suggests a clear cost of generation for Taipower, which is played out in Taipower’s 2015 generating cost breakdown - NTD1.15/KWh for nuclear, 1.22 for Taipower coal, 2.05 for IPP coal, 2.68 for Taipower LNG, and 3.29 (on average) for IPP LNG.
Cogeneration was also quite cheap at 2.21/KWh, but without a better understanding of how
Cogeneration was also quite cheap at 2.21/KWh, but without a better understanding of how