Introduction
The conventional gas terminal at Datan power plant is planned to start operation by late 2022, supplying 2 CCGT with a total of 2 GW of maximum capacity.
However, the gas terminal requires the construction of a new port, under which a coralline algae ecosystem might be irreversibly damaged. In addition, conventional gas terminals will not be suitable for importing green hydrogen in the long term future, thereby they are subject to becoming stranded assets during the energy transition.
A referendum has been initiated by environmental groups in Taiwan to halt the project; this assessment report investigates what could happen to near term coal reduction and power system dispatch if such referendum is passed.
Coal Reduction Impact Estimation
Without the terminal, there will be an annual shortage of 1.41 million tonnes of conventional gas supply in Taiwan by 2025.
Assuming this supply gap will be met by additional coal consumption in the electricity sector, this will result in 3.28 million tonnes of additional coal consumption, or 4.36 million tonnes of additional carbon emissions annually.
Source of which the value 1.41 million is obtained:
《能源統計月報》，Bureau of Energy，Feb 2021
《溫室氣體排放係數管理表6.0.4版》，Environment Protection Agency
Power System Dispatch Impact Estimation
To investigate the impact of delay / cancellation of the terminal construction will have on the power system dispatch, we will solve a minimum emission dispatch problem with a power system dispatch model of Taiwan by 2025.
Dispatch Problem and Model
Dispatch Model: Assumptions
- Electricity demand in Taiwan will grow by 10% by 2025, compared with 2020.
- Gas supply constraint: in the congestion free case, we assume daily gas consumption of the power system can not exceed 80% of the import capacity (estimated from power system data in 2020) of Taiwan; in the congestion case we assume no additional gas transmission capacity is available for Datan power plant, so it can only operation 7 CCGT at full load at most at any given time.
(The actual gas pipe transmission constraint must be in between the 2 extreme cases; information on the gas pipe transmission capacity and gas demand profile would be needed to make assumptions that can better reflect reality.)
https://munchkuo.blog/2021/03/20/藻礁公投成案後-開發三接的謬誤與出路/
It seems that the reality the congestion constraint would be looser than the congestion case (needs verification). In the congestion case the transmission capacity gap is around 3100 \(\mathrm{MW}_{th}\), while in reality by 2025 it would probably be 1450 \(\mathrm{MW}_{th}\), so reality is most likely half way between congestion free and congestion case.
- Demand and generation of electricity is balanced at any given time.
- 10% of the residual load of control reserve capacity must be kept at any given time.
- Technical constraints (start-up time, ramp limits, minimum compliant load, etc.) of the power plants must be honored.
- Fuel consumption variations due to operational efficiency change are neglected, so the problem remains a mix-integer linear problem.
Dispatch Problem: Objective Function
$$ \begin{equation} \min \sum_{t,i} e_i \cdot cpp_i(t) \cdot \Delta t + e_{u,i} \cdot u_i(t) \end{equation} $$Where
- \(e_i\): emission factor of running conventional power plant i
- \(cpp_i(t)\): power output of conventional power plant i at time t
- \(e_{u,i}\): emission factor of starting conventional power plant i
- \(u_i(t)\): a time series of binary variables indicating when start-up events of conventional power plant i occurs
Dispatch Problem: Supply and Generation Constraints
-
Balancing constraint
$$ \begin{equation} \sum_{i} cpp_i(t) \geq RL(t) \:\forall t \end{equation} $$ -
Gas consumption constraint
$$ \begin{equation} \sum_{i \in \{Gas\}, \:j \in \{GT_i\}, t} \frac{cpp_{i,j}(t)}{\eta_i} \cdot \Delta{t} \leq Gas_{max} \end{equation} $$ -
Gas transmission capacity constraint
$$ \begin{equation} \sum_{i \in \{Datan\}, \:j \in \{GT_i\}} cpp_i(t) \leq Datan_{max} \:\forall t \end{equation} $$
Where
- \(RL(t)\): residual load at time t
- \(\eta_i\): thermal efficiency of power plant i
- \(Gas_{max}\): maximum available gas within a time interval
- \(Datan_{max}\): maximum possible power output at Datan
Dispatch Problem: Control Reserve Constraints
-
Headroom constraint
$$ \begin{equation} flex_{u,i}(t) \leq s_i(t) cpp_{\max, i} - cpp_i(t) \:\forall i,t \end{equation} $$ -
Upward flexibility constraint
$$ \begin{equation} flex_{u,i}(t) \leq ru_{\max,i} \:\forall i,t \end{equation} $$ -
Positive control reserve constraint
$$ \begin{equation} \sum_i flex_{u,i}(t) \geq \max\{flex_{u,\min}, \:0.1 \cdot RL(t)\} \:\forall t \end{equation} $$ -
Footroom constraint
$$ \begin{equation} flex_{d,i}(t) \leq cpp_i(t) - s_i(t) cpp_{\min, i} \:\forall i,t \end{equation} $$ -
Downward flexibility constraint
$$ \begin{equation} flex_{d,i}(t) \leq rd_{\max,i} \:\forall i,t \end{equation} $$ -
Negative control reserve constraint
$$ \begin{equation} \sum_i flex_{d,i}(t) \geq \max\{flex_{d,\min}, \:0.1 \cdot RL(t)\} \:\forall t \end{equation} $$
Where
- \(flex_{u,i}(t)\): upward flexibility of power plant i at time t
- \(cpp_{\max, i}\): maximum available power output
- \(ru_{\max}\): maximum ramp up rate
- \(s_i(t)\): binary variable indicating the operation state of power plant i at time t
- \(flex_{d,i}(t)\): downward flexibility of power plant i at time t
- \(cpp_{\min, i}\): minimum available power output
- \(rd_{\max}\): maximum ramp down rate
Dispatch Problem: Technical Constraints
-
Ramp up constraint
$$ \begin{equation} cpp_i(t+1) - cpp_i(t) \leq ru_{\max} + u_i(t) \cdot (cpp_{\min, i} - ru_{\max}) \:\forall i,t \end{equation} $$ -
Ramp down constraint
$$ \begin{equation} cpp_i(t) - cpp_i(t+1) \leq rd_{\max} + d_i(t) \cdot (cpp_{\max, i} - rd_{\max}) \:\forall i,t \end{equation} $$ -
Power output constraint
$$ \begin{equation} s_i(t) \cdot cpp_{\min, i} \leq cpp_i(t) \leq s_i(t) \cdot cpp_{\max,i} \:\forall i,t \end{equation} $$ -
State Evolution Constraint
$$ \begin{equation} s_i(t+1) = s_i(t) + u_i(t) - d_i(t) \:\forall i,t \end{equation} $$ -
Start-up time constraint
$$ \begin{equation} d_i(t) + u_i(t + \tau) \leq 1 \:\forall i,t=\{0,1,...,T-\tau\},\tau=\{1,2,...,su_{\min,i}\} \end{equation} $$ -
Start-up order constraint for combined cycle steam turbines
$$ \begin{equation} s_{i, ST}(t) \leq \sum_{j \in \{GT_i\}} s_{i,j}(t) \:\forall i \in \{CC\},t \end{equation} $$ -
Power output constraint for combined cycle steam turbines
$$ \begin{equation} cpp_{i, ST}(t) \leq \alpha \sum_{j \in \{GT_i\}} cpp_{i,j}(t) \:\forall i \in \{CC\},t \end{equation} $$ -
Start-up time constraint for combined cycle steam turbines
$$ \begin{equation} \frac{1}{N_i} \sum_{j \in \{GT_i\}} s_{i,j}(t) \leq s_{GT_i}(t) \leq \sum_{j \in \{GT_i\}} s_{i,j}(t) \:\forall i \in \{CC\},t \end{equation} $$
$$ \begin{equation} \begin{array}{l} s_{GT_i}(t) - s_{GT_i}(t-1) + u_{i,ST}(t + \tau) \leq 1 \\ \forall i \in \{CC\},t=\{0,1,...,T-\tau\},\tau=\{1,2,...,su_{\min,i,ST}\} \end{array} \end{equation} $$
Where
- \(d_i(t)\): a time series of binary variables indicating when shut-down events of conventional power plant i occurs
- \(su_{\min,i}\): minimum start-up time for power plant i
- \(s_{i,ST}(t)\): state variable for the steam turbine in a CCGT
- \(s_{i,GT}(t)\): state variable for the gas turbines in a CCGT
- \(\alpha\): coefficient determining the maximum extractable waste heat from the gas turbine
- \(cpp_{i,j}(t)\): power output of a ST/GT in a CCGT.
- \(s_{GT_i}(t)\): binary variables indicating whether any of the GT in a CCGT is operating.
- \(N_i\) : Number of GTs in CCGT i.
Dispatch Problem: Mathematical Constraints
\(\forall i,t\):
-
Binary variables
$$ \begin{align} u_i(t) &\in \{0,1\} \\ d_i(t) &\in \{0,1\} \\ s_i(t) &\in \{0,1\} \\ s_{GT_i}(t) &\in \{0,1\} \end{align} $$ -
Non-negative variables
$$ \begin{align} cpp_i(t) \geq 0 \end{align} $$
Dispatch Problem: VRE Modeling
As a conservative model, we assume the VRE power plants are totally inflexible, i.e. they cannot contribute to the control reserve requirement of the system.
This assumption is probably trivial because curtailment is still not an issue by summer 2025; with enough storage and other flexibility options the minimum residual load is still safely above the system must run. In winter 2025 this might prove to be an issue, but that is beyond the scope of this assessment report.
The effective load carrying capacity for VRE in the annual schedule problem is also a conservative estimation; we assume only during summer and only solar PV has ELCC.
Dispatch Problem: Non-Hydro DRE Modeling
We will assume that bioenergy, waste, and geothermal power plants runs at a constant power output that is derived empirically from the annual average capacity factor of 2020.
This means bioenergy power plants runs at 23.2% of full capacity, waste power plants at 64.2%, and geothermal power plants at 72.5%.
We assume that geothermal will increase to 0.040 GW by 2025; the capacity of other types of DRE remains the same compared with 2020.
Overall this translates into around 0.453 GW of constant non-hydro DRE output. Since the firm capacity factor for waste power plants is 70.3%, this translates into 0.039 GW of additional positive control reserve. Meanwhile, since Taipower assumes 50% effective load carrying capacity factor of bioenergy power plants, this translates into 0.021 GW of additional positive control reserve. In total non-hydro DRE power plants provide 0.060 GW of additional positive control reserve (upward flexibility of geothermal power plants is not considered).
Assuming non-hydro DRE power plants are fully flexible, the constant non-hydro DRE output of 0.453 GW can be added into the negative control reserve.
Dispatch Problem: DR Modeling
Currently Taipower purchased about 1.5 GW of demand response; they are however mostly voluntary programs with no penalties. We therefore do not consider these flexibility resources as reliable control reserve capacity in the dispatch problem; nonetheless we assume that these resources have an effective load carrying capacity factor of 20% when solving the annual schedule problem presented later.
We assume that by 2025 there will be 200 MW of reliable DR for positive control reserve; this is in line with Taipower’s current plan of purchasing 800 MW of control reserve from unconventional resources (batteries, DR, VRE, etc.).
Dispatch Problem: CHP Modeling
About 4.968 GW of coal CHP is currently installed in Taiwan; of which Taipower assumes 1.474 GW of effective load carrying capacity.
We thus model the coal CHP power plants to be a fully flexible generation unit with a peak capacity of 1.474 GW. The thermal efficiency of coal CHP is assumed to be 65%, and the carbon emissions per kWh is calculated accordingly.
Only 0.025 GW of Oil CHP currently exists in Taiwan and they do not account for any effective load carrying capacity, so they are neglected in our model.
Marginal change of coal consumption due to change of CHP electricity generation is actually very complex.
Annual Plant Schedule Problem
Before solving the minimum emission dispatch problem, we need to know which conventional power plants are available in summer and winter; in other words, we need to solve an annual plant schedule problem.
We use empirical monthly peak demand data of 2020 to estimation the monthly peak demand by 2025.
As a conservative estimation of the capacity value of VRE, we assume that they can provide 3 GW of peak residual load shedding during July and August, and 1.5 GW during May and September. For other months the reserve capacity must be fulfilled by DR, storage, conventional and DRE power plants.
Assumptions for the schedule problem
- All conventional power plants need at least 1 month of maintenance in a year.
- For every month, the available capacity must be greater or equal to 110% of the projected peak demand (if this cannot be met, we lower the criteria until a feasible solution exists).
- For air pollution months (October to March), total assigned coal capacity cannot exceed 8 GW. Additional regulatory constraints on Taichung and Hsinta coal power plants are also taken into account.
- In the congestion scenario there will be an additional constraint on the maximum available power capacity at Datan power plant at any given time.
For Taichung coal power plant we assume that only 8 units can be operational between May and September, and 6 units in other months.
For Hsinta coal power plant we assume that only 1 unit can be operational in May and September, and 2 units in July and August.
Model for the schedule problem
$$ \begin{align} & \max Reserve_{\min} \\ & \begin{array} {lcl} \sum_{i} cpp_{\max,i} \cdot u_i(t) &\geq& 1.1 \cdot RL_{\max}(t) \\ \sum_{i} cpp_{\max,i} \cdot u_i(t) &\geq& RL_{\max}(t) + Reserve_{\min} \\ \sum_{i\in\{Coal\}} cpp_{\max,i} \cdot u_i(t) &\leq& Coal_{\max}(t) \\ \sum_{i\in\{Datan\}} cpp_{\max,i} \cdot u_i(t) &\leq& Datan_{\max}(t) \end{array} \:\forall t \\ &\sum_t m_i(t) = 1 \:\forall i \\ &m_i(t) + m_i(t - 1) + u_i(t) \leq 1 \:\forall i, t \\ &u_i(t) \in \{0, 1\} \\ &m_i(t) \in \{0, 1\} \end{align} $$Note that there is an abuse of notation here: in the economic dispatch problem, \(u_i(t)\) indicates whether a start-up event occurrs for a conventional power plant i at time t, while here it indicates whether the conventional power plant is operational.
Annual Schedule Problem Results
If the 2 units of Hsinta coal power plant can be utilized 100% between May and September, the minimum available reserve margin can be increased to 7.04%.
If the 2 units of Hsinta coal power plant can be utilized 100% between May and September, the capacity value of offshore wind power plants becomes more obvious; the minimum available reserve margin can be increased to 7.45%.
Annual Plant Schedule Problem: Results Passing
The solution to the annual plant schedule problem will be passed to the main dispatch problem; it will correspond to the available conventional power plants at the modeled time slice.
RL Profile and Preprocessing
Before solving the minimum emission dispatch problem, we need to model the residual load profile. In this step the dispatch of storage units and hydroelectricity power plants are also modeled, and the residual load profile will be smoothed accordingly.
We use empirical demand and VRE (solar PV, onshore wind, and offshore wind) capacity factor data from 23-24 July, 2020 to generate a 48-hour-long demand load and residual load profiles for the impact study. Loss data points are linearly interpolated.
We assume that there will be 20 GW of solar PV, 0.8 GW of onshore wind, and 5.5 GW offshore wind by 2025.
Hydroelectricity and storage will be scheduled to minimize the maximum residual load during the period.
For hydroelectricity, we will only consider reservoir hydro (1.835 GW): we assume the daily average capacity factor of these power plants are 17.9% (empirical value of July 2020) but the operator can decide when to dispatch the water so long as the daily electricity output does not exceed the capacity factor mentioned above. An power limit of 0.6 GW is empirically observed and will also be set in the problem.
For storage, we assume the SOC to be 50% at the beginning and the end of modeling (we set maximum available to be energy stored after charging continuously at full capacity for 24 hours for pump storage and 4 hours for batteries). We also set maximum charge of PS to 2.5 GW and discharge to 1.8 GW according to empirical data. Power capacity of BAT is set to 0.59 GW. Also, to avoid abrupt change, we impose addition ramp constraints (\(0.1 \cdot P_{max}\)) on storage and hydroelectricity power plants.
Once the optimal peak residual load is obtained, we polish the solution so that the cycling energy loss is minimized. Then, we polish the solution again, so an imaginary price spread (assumed to be the original RL) is minimized.
RL Profile and Preprocessing: Original Problem
$$ \begin{align} &\min RL_{\max} \\ &RL(t) + \sum_{s_i} (P_{s_i,dc}(t) - P_{s_i,ch}(t)) + HE(t) = RL_0 (t) &\:\forall t \\ &SOC_{s_i}(t+1) = SOC_{s_i}(t) - \frac{P_{s_i,dc}(t)}{\eta_{s_i, dc}} + \eta_{s_i, ch} P_{s_i,ch}(t) &\:\forall s_i,t \\ & 0 \leq SOC_{s_i}(t) \leq SOC_{\max, s_i} &\:\forall s_i, t \\ & 0 \leq P_{s_i, dc}(t) \leq P_{\max, s_i, dc} &\:\forall s_i, t \\ & 0 \leq P_{s_i, ch}(t) \leq P_{\max, s_i, ch} &\:\forall s_i, t \\ & 0 \leq HE(t) \leq HE_{\max, power} &\:\forall t \\ & \sum_t HE(t) \Delta t \leq HE_{\max, energy} \\ & RL(t) \leq RL_{\max} &\:\forall t \end{align} $$The original problem minimizes peak residual load in the time interval considered.
Note that there is an abuse of notation here: in the economic dispatch problem, \(s_i(t)\) stands for the operational status of a conventional power plant i at time t, while in the RL profile and preprocessing problem \(s_i\) stands for a storage unit i.
RL Profile and Preprocessing: Intermediate Problem
$$ \begin{align} &\min \sum_t \left( \frac{1}{\eta_{s_i, ch} } - 1 \right) \cdot P_{s_i,ch}(t) + (1 - \eta_{s_i,dc}) \cdot P_{s_i, dc}(t) \\ &RL(t) + \sum_{s_i} (P_{s_i,dc}(t) - P_{s_i,ch}(t)) + HE(t) = RL_0 (t) &\:\forall t \\ &SOC_{s_i}(t+1) = SOC_{s_i}(t) - \frac{P_{s_i,dc}(t)}{\eta_{s_i, dc}} + \eta_{s_i, ch} P_{s_i,ch}(t) &\:\forall s_i,t \\ & 0 \leq SOC_{s_i}(t) \leq SOC_{\max, s_i} &\:\forall s_i, t \\ & 0 \leq P_{s_i, dc}(t) \leq P_{\max, s_i, dc} &\:\forall s_i, t \\ & 0 \leq P_{s_i, ch}(t) \leq P_{\max, s_i, ch} &\:\forall s_i, t \\ & 0 \leq HE(t) \leq HE_{\max, power} &\:\forall t \\ & \sum_t HE(t) \Delta t \leq HE_{\max, energy} \\ & RL(t) \leq RL_{\max} &\:\forall t \end{align} $$Since the optimal solutions to the original problem lie in a highly degenerate space, the intermediate problem finds a subset of the solutions where the cycle loss of storage units is minimized.
RL Profile and Preprocessing: Polishing Problem
$$ \begin{align} &\min \sum_t RL_0(t) \cdot \left( \frac{P_{s_i, ch}(t)}{\eta_{s_i, ch} } - \eta_{s_i,dc} \cdot P_{s_i, dc}(t) \right) \\ &RL(t) + \sum_{s_i} (P_{s_i,dc}(t) - P_{s_i,ch}(t)) + HE(t) = RL_0 (t) &\:\forall t \\ &SOC_{s_i}(t+1) = SOC_{s_i}(t) - \frac{P_{s_i,dc}(t)}{\eta_{s_i, dc}} + \eta_{s_i, ch} P_{s_i,ch}(t) &\:\forall s_i,t \\ & 0 \leq SOC_{s_i}(t) \leq SOC_{\max, s_i} &\:\forall s_i, t \\ & 0 \leq P_{s_i, dc}(t) \leq P_{\max, s_i, dc} &\:\forall s_i, t \\ & 0 \leq P_{s_i, ch}(t) \leq P_{\max, s_i, ch} &\:\forall s_i, t \\ & 0 \leq HE(t) \leq HE_{\max, power} &\:\forall t \\ & \sum_t HE(t) \Delta t \leq HE_{\max, energy} \\ & RL(t) \leq RL_{\max} &\:\forall t \\ & \sum_t \left( \frac{1}{\eta_{s_i, ch} } - 1 \right) \cdot P_{s_i,ch}(t) + (1 - \eta_{s_i,dc}) \cdot P_{s_i, dc}(t) \leq Loss_{\min} \end{align} $$The space of optimal solution for the intermediate problem can still be highly degenerate, so a polish problem is solved to choose a subset of the solutions where the marginal operational cost is minimized.The marginal price of electricity is assumed to be proportional to the default residual load.
RL Profile and Preprocessing: Results
RL Profile and Preprocessing: Remaining Flexibility
After the preprocessing problem is solved, there might be remaining flexibility regarding the power and energy output of hydroelectric power plants. This flexibility will be calculated and passed on to the main dispatch problem.
Since Taipower assumes 87% effective load carrying capacity factor of reservoir hydro, at any moment the upward flexibility of hydroelectricity power plants can be assumed to be \(1.835 \cdot 0.87 = 1.59645\) GW minus the scheduled power output. This can be added to the positive control reserve, so at any moment hydroelectricity power plants can contribute to at least 0.996 GW of positive control reserve.
Results
Results: Dispatch Problem (Summer)
Dispatch Problem Results Summer: Comparison
Dispatch Problem Results: System Reliability of Different Cases
In this assessment report, due to the nature of the modeling only prediction error and power plant outage can be considered. Transmission line outage, gas infrastructure outage, and other sources of contingencies are not considered.
To give a conservative estimation, we assume that unscheduled outage happens at a frequency of 1 power plant per week. We assume prediction error sd of RL is \(0.5 + \max\{0.1 \cdot VRE, 0.5\}\) GW.
Monte Cairo simulations with 1 million samples for each case are conducted to find the reliability statistics.
Mitigation Approaches
To mitigate the coal reduction impact of not building the gas terminal as planned, there are three main aspects policy makers will consider:
- Alternative Gas Supply Options
- Accelerated Renewable Deployment
- Ambitious Energy Efficiency and Conservation Policies
Mitigation Approaches: Comparison
| Alternative Gas Supply | RE+ | EE+ | |
|---|---|---|---|
| Possible Policies or Plans in Near Term | Floating Storage Regasification Unit (FSRU) | Broader and stricter renewable obligation for energy intensive industries and other entities | Carbon pricing; local marginal pricing for industrial areas |
| Corner Solution Requirements | Unavailable about 10 days annually due to typhoons, so to fully mitigate the coal reduction impact the maximum import capacity must be increased by 2.81% compared with the original plan. | To completely fill the gap without addition coal consumption about 10.3 TWh of renewable energy is required, which translates to around 9.03 GW of solar PV or 2.86 GW of offshore wind. If only carbon emission is considered then only 55.2% of this amount is required. | To completely fill the gap without addition coal consumption about 10.8 TWh of gross electricity demand needs to be reduced (around 3.51% of the original projected gross electricity demand by 2025). If only carbon emission is considered then only 55.2% of this amount is required. |
| Advantages | Reduces costs and ecological impacts; can be deployed to other ports according to the overall gas supply situation in Taiwan later. | Non-regrettable (we will need to deploy RE anyway); avoids significant ecological impacts | Non-regrettable (we will need to set these policies eventually); completely avoids ecological impacts |
| Potential Obstacles | Still fossil fuel infrastructure and subject to stranded asset risk; additional safety concern during typhoons. | Opposition from energy intensive industries; corner solution means increasing the planned installation rate of VRE by 45.2% (solar PV) to 52.0% (offshore wind) until 2025. | Opposition from energy intensive industries; price elasticity of electricity highly uncertain. |
Gross electricity demand: the electricity demand on grid + generator self consumption
Net electricity demand: the electricity demand on grid = 95% of Gross electricity demand
For replacement of coal by renewable energy, generator self consumption is assumed to be negligible, so only the net electricity demand is needed to be replaced
Mitigation Approaches: Optimal Pathway
Obviously an optimal mitigation pathway will be a combination of all the possible approaches.
For example, in the end 50% of the coal reduction impact will be mitigated with a smaller FSRU project, 25% with additional wind and solar, 25% with higher carbon tax and local marginal pricing for industrial areas in Hsinchu and Taoyuan.
Further socio-economic assumptions and parameters will be needed to determine the optimal pathway. This is beyond the scope of this assessment report.
Conclusions
-
A rough estimation suggests that the delay / cancellation of Datan conventional gas terminal will result in 3.28 million tonnes of additional coal consumption and 4.36 million tonnes of additional carbon emissions annually.
-
Under the congestion free case, the minimum reserve margin ratio for each month can be kept above 10%, and dispatch modeling suggests that power system reliability will not be affected even if Datan conventional gas terminal is delayed / cancelled.
-
Under the congestion case, the minimum reserve margin ratio for each month will decrease 3.90 % to 6.11% if Datan conventional gas terminal is delayed / cancelled. However, dispatch modeling suggests power system reliability is still within tolerable range during peak residual load time periods in summer.
-
Some of the possible approaches to cope with the potential coal reduction gap due to delay / cancellation of the terminal are shown. The optimal pathway will most likely be a combination of several of these approaches.
Reference
- Latest (Dec. 2020) Energy Statistics in Taiwan:
《能源統計月報》，Bureau of Energy，Feb 2021 - Latest (2019) Carbon Emission Statistics in Taiwan
《溫室氣體排放係數管理表6.0.4版》，Environment Protection Agency - Effective Load Carrying Capacity Factors and CHP Data
Taipower website. - Details of the Conventional Power Plant Fleet in Taiwan
2019 annual reports of the private and public utilities in Taiwan. They were published in March 2020. - Technical Parameters for Conventional Power Plants
- “Generator Technical and Cost Parameters”, ElectraNet, July 2020
- “On Start-up Costs of Thermal Power Plants in Markets with Increasing Shares of Fluctuating Renewables”, Wolf-Peter Schill, Michael Pahle and Christian Gambardella, Deutsches Institut für Wirtschaftsforschung, 2016
- “Flexibility in thermal power plants - With a focus on existing coal-fired power plants”, Agora Energiewende, 2017
LP Solver Used for the Modeling and Original Data
The LP problem in RL pre-processing was solved with lpsolve.
The MILP problems of annual plant scheduling and minimum emission dispatch were solved with GCG - SCIP.
The dispatch problem was too complex to solve so we assumed that all the conventional power plants were on line at any given time; the problem was thus reduced to a LP.
The data for conventional power plant, DRE, and storage fleet by 2025, and the results of the power system dispatch modeling can be found here.