WP EN2018-08
Integrating flexibility constraints in long-term planning models: impact, pitfalls and modeling recommendations K. Poncelet, E. Delarue, W. D’haeseleer
TME Working Paper - Energy and Environment Pre-print submitted to IEEE Transactions on power systems Last update: May 2018 An electronic version of the paper may be downloaded from the TME website: http://www.mech.kuleuven.be/tme/research/
1
Integrating flexibility constraints in long-term planning models: impact, pitfalls and modeling recommendations Kris Poncelet%,+ , Erik Delarue%,+,* and William D’haeseleer%,+ %
University of Leuven (KU Leuven), Energy Institute, Celestijnenlaan 300, B-3001 Leuven, Belgium + *
EnergyVille, Thor Park, B-3600 Genk, Belgium
Corresponding author:
[email protected]
Abstract With an increasing penetration of intermittent renewable energy sources (IRES), it is expected that it becomes more important to account for detailed flexibility constraints in generation expansion planning (GEP) models. In this context, multiple authors have assessed the impact of neglecting flexibility in GEP models, concluding that it has a significant impact on the results. However, in the current literature assessing the impact of flexibility, thermal generators are often considered to be the only source of flexibility and the sensitivity to certain assumptions made have not been investigated. This paper contributes to the existing literature by revisiting the relevance of considering flexibility constraints in GEP models for varying assumptions regarding the ability to provide, and the need for flexibility. The results indicate that when other flexibility providers are considered, integrating technical constraints has a limited impact on both cost projections and investments, with the exception of investments in dedicated flexibility providers. Furthermore, the investments in dedicated flexibility providers are shown to be highly sensitive to the assumptions made regarding the flexibility of thermal generators and the future need for operating reserves. Finally, the analysis alerts for the risk of making overly conservative assumptions when integrating detailed flexibility constraints.
Abbreviations BAT CCGT COALSC CU C GEP GHG NUC P HS PV OCGT UC IRES
Battery storage Combined cycle gas turbine Super-critical coal-fired power plant Clustered unit commitment Generation expansion planning Greenhouse gas Nuclear power plant Pumped hydro storage Photovoltaic Open cycle gas turbine Unit commitment Intermittent renewable energy sources
1
1
Introduction
Due to computational restrictions, generation expansion planning (GEP) models or long-term energy-system planning models typically do not integrate detailed operational constraints. More specifically, the technical constraints faced by individual power plants (e.g., the minimum operating point, ramp-rate restrictions, start costs, etc.), as well as detailed system constraints aiming to ensure reliability (e.g., operating-reserve requirements) are typically not considered. This level of detail has historically been reserved for operational models, such as unit commitment (UC) models. However, in the context of an increasing penetration of intermittent renewable energy sources (IRES), concerns have recently been voiced that neglecting this level of detail in planning models might yield infeasible or sub-optimal solutions to the planning problem (see e.g., [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]). These concerns regarding flexibility in power systems with high penetration levels of IRES have spurred multiple authors to integrate detailed flexibility constraints in planning models. Kirschen et al. [7], Ma et al. [8] and de Sisternes [11] have extended a UC problem to consider investment decisions in individual plants. To maintain computational tractability, they make use of four or five representative weeks to represent an entire year. To make optimal use of the available computational resources, different authors have developed methods to carefully select representative sets of weeks or days (see e.g., [12, 13, 14]). A different strategy, proposed by Flores-Quinoz et al.[15], is to use decomposition techniques to reduce the computational burden. Other authors have instead focused on altering the formulation of the technical constraints. For instance, Palmintier and Webster [5, 16] have developed a clustered unit commitment (CUC) formulation, in which the binary commitment variables for individual generation units are replaced by an integer commitment variable per cluster of identical or similar generation units. This allows reducing the combinatorial state space, the number of variables and the number of constraints, and is shown to significantly reduce the computational cost. A number of authors explored using continuous rather than integer commitment variables to further reduce the computational burden and showed that this has a limited impact in terms of accuracy [17, 18, 19]. Other authors have used even more stylized representations of the technical constraints (see e.g., [6, 20, 21, 22]). Whereas a myriad of different approaches has been developed to integrate technical constraints in planning models, fewer authors have investigated the actual impact of not incorporating detailed technical constraints. By taking the capacity mix resulting from a planning model which does not incorporate technical constraints and reevaluating the dispatch decisions using a UC model, Deane et al. [23] and Poncelet et al. [1] both conclude that neglecting these technical constraints leads to an overestimation of the value of baseload and IRES technologies whereas flexible technologies are not sufficiently valued. In turn, this was shown to lead to an underestimation of the operational costs, and hence the total system cost. Similar findings were found by Belderbos and Delarue [10]. Shortt et al. [3] have investigated the impact of neglecting detailed technical constraints for a broad set of test systems differing in load and renewable time series, the penetration level of wind energy and the type of generators using a UC model. They highlight that the impact of neglecting these technical constraints is highly system specific, but generally becomes higher with an increasing penetration of IRES. Finally, Palmintier and Webster [5, 4] developed a clustered UC formulation which allowed them to directly investigate the impact of neglecting the technical constraints in the investment planning problem. Also they conclude that neglecting these technical constraints has a significant impact on the capacity mix deemed optimal by the model and the resulting generation mix. Additionally, they emphasize that the impact of neglecting these technical constraints not only tends to increase with an increasing penetration of IRES, but also tends to grow as more stringent carbon policies are imposed. In summary, in the recent literature regarding flexibility aspects in long-term planning models, a number of authors have assessed the impact of not incorporating detailed technical constraints and have concluded that neglecting technical constraints has a significant impact on the capacity mix, the generation mix and the cost projections. Recently, the focus of the literature has shifted towards developing methods to tractably integrate these technical constraints. However, the existing literature has a number of limitations: 1. In the literature assessing the impact of neglecting technical constraints, thermal genera3
tors are frequently considered to be the only source of flexibility (aside from curtailment of IRES electricity generation). 2. The sensitivity to certain input data and specific modeling assumptions that need to be taken when integrating detailed technical constraints in long-term planning have not been investigated. More specifically, assumptions need to be taken regarding: (i) the current and future cycling capabilities of thermal power plants and (ii) the amount and type of operating reserves which will be required in future power systems. Regarding the cycling capabilities of thermal power plants, a wide range of values are reported in the literature (see e.g., [24, 25, 26]). Regarding the characterization of operating-reserve requirements, rules of thumb based on current practices or simple extrapolations are regularly used. More details on these assumptions are presented in Section 3.3. The goal of this paper is twofold. The first goal is to revisit the impact of not considering detailed technical constraints in planning models in the light of the above-mentioned limitations of the current literature. The second goal is to assess the sensitivity of the model results to certain assumptions taken when integrating detailed technical constraints. As such, this paper contributes to the existing literature in the following ways. First, as will follow from the remainder of this paper, the assumptions adopted in the current literature can have a strong impact on the conclusions regarding how relevant it is to integrate these technical constraints. This work therefore adds nuance to the results described in the current literature. Second, our analysis exposes critical elements and potential pitfalls when integrating technical constraints in generation expansion planning models, and emphasizes key aspects which are currently underexposed in the literature and require further research. The remainder of this paper is organized as follows. First, Section 2 briefly describes the models used in this paper. Next, Section 3 presents the methodology used to assess the impact of neglecting detailed technical constraints for a variety of assumptions and to assess the sensitivity to these assumptions. Section 4 subsequently describes the case study used in this paper. The results are presented and discussed in Section 5. Finally, the main conclusions are summarized in Section 6.
2
Model description
The model used in this paper is a greenfield GEP model. Two versions of this model are considered. A first model integrates a CUC model. This CUC model groups together identical power plants and replaces the binary commitment variables for individual generators by an integer variable per technology cluster. The following elements are accounted for in the model: ramping constraints, the minimum operating point, minimum up and down times, start-up costs, part-load efficiency losses and operating reserve requirements. For the sake of brevity, the mathematical formulation is not presented here. For a detailed mathematical description of the model, we refer to [27]. A second model is a traditional investment planning model which does not consider technical constraints. In earlier work, the CUC model has been validated by comparing it with a traditional UC model employing binary variables. The errors introduced by clustering identical power plants were shown to be neglegible [28].
3 3.1
Methodology General methodology
To assess the impact of not incorporating detailed technical constraints, the results of the planning model integrating detailed technical constraints (indicated by ’REF’) are compared to the results of the planning model which does not consider such technical constraints. When no technical constraints are being considered, technology-types will be dispatched according to the merit-order list. We therefore refer to the model which does not incorporate technical constraints as the ’MO’ model. The results are compared in terms of the projected total annual system cost and the capacity mix. 4
3.2
Scenarios
To consider the relationship between the penetration level of IRES and the composition of the thermal generation fleet, on the one hand, and the impact of not considering technical constraints, on the other hand, four different scenarios are considered. An overview of these scenarios is presented in Tab. 1. Table 1: Overview of the considered scenarios representing different capacity mixes Scenario
Ther Nuc
Ther No nuc
IRES No nuc
IRES Nuc
0 0 -
0 0 nuclear
30 50 -
100 50 -
GHG tax [EU R/ton] IRES support [EU R/M W h] Technology-types excluded
Two scenarios have a low penetration of IRES and are hence dominated by thermal generators (indicated by ’Ther’), and two scenarios have a considerable penetration of IRES (indicated by ’IRES’). For both the scenarios with a low penetration of IRES and the scenarios with a high penetration of IRES, two variants are created: one with nuclear plants and one without nuclear power plants (indicated by ’Nuc’ and ’No nuc’ respectively). To achieve these different capacity mixes, the tax for greenhouse gas (GHG) emissions and the support for IRES used in the GEP model are varied, as indicated in Tab. 1.
3.3
Cases
In order to analyze the impact of certain modeling decisions, a number of different cases is considered. These cases differ in terms of the assumed flexibility of thermal generators, the availability of other flexibility providers and the assumptions taken for characterizing operating reserve requirements. 3.3.1
Cycling capabilities and other flexibility providers
First, both the assumed cycling capabilities of thermal power plants and the availability of other flexibility providers have an impact on the supply side of flexibility. As discussed in [25], there is a large range of data regarding the cycling characteristics of thermal power plants. In this paper, we consider two sets of cycling characteristics. In the first set, the flexibility of thermal power plants is near the lower limit of the ranges reported in the literature (referred to as the ’Inflex’ case) whereas in the second set, the flexibility of thermal power plants is assumed to be near the upper limit of the ranges specified in the literature (referred to as the ’Flex’ case). The cycling characteristics adopted in both cases are presented in Tab. 2. The range of cycling capabilities of thermal power plants is adopted from [25, 26, 24]. Regarding the availability of other flexibility providers, we consider a case with or without the opportunity to invest in electricity storage-technology types. Two types of storagetechnology types are considered: pumped-hydro storage (PHS) and battery storage (BAT)). Cases in which investments in storage-technology types are and are not considered are indicated by a ’Stor’ and ’No stor’ respectively. An overview of the 4 considered flexibility cases is presented in Tab. 3. 3.3.2
Characterizing operating reserve requirements
Tab. 4 presents an overview of how reserve requirements are characterized in a number of stateof-the-art planning models. From this table, it can be observed that between different models, significant differences exist in terms of the sizing of reserve requirements and the required activation times (and whether or not fast-starting units can provide reserves). In addition to these differences in terms of the required amount of reserve requirements and the required activation times, it can be observed that the methodologies used to endogenously size reserves in most planning models are highly simplified. First, different sources of 5
Table 2: Cycling characteristics of thermal generators Technical characteristic
Case
NUC
COAL SC
CCGT
OCGT
Minimum operating point [%/Pnom ]
Inflex Flex
50 40
40 25
50 30
50 20
Eff. loss at minimum operating point [%pt]
Inflex Flex
5 1.8
2 2
11 3.2
22 9
Ramp rate [%Pnom /min]
Inflex Flex
0.25 5
0.66 4
0.83 10
0.83 25
Minimum up time [h]
Inflex Flex
24 0.25
10 0.25
6 0.25
1 0.25
Minimum down time [h]
Inflex Flex
24 24
10 3
6 0.5
1 0.25
Start-up energy [M W hth /∆M We ]
Inflex Flex
46.7 16.7
3.6 3.6
1.8 1.5
0.0 0.0
Start-up depreciation [EUR/∆M We ]
Inflex Flex
1.7 1.7
70.3 45.1
68.4 24.5
105.0 19.4
Start-up time [h]
Inflex Flex
50 24
8 2
1 1
0.33 0.17
Table 3: Overview of the considered cases representing different levels of available flexibility Case
Inflex No stor
Flex No stor
Inflex Stor
Flex Stor
low no
high no
low yes
high yes
Flexibility of thermal generators Storage available
uncertainty (e.g., demand forecast errors, wind generation forecast errors and solar generation forecast errors) are often treated independently, which can lead to an overestimation of the required reserves. Second, the required reserves to deal with IRES forecast errors are typically assumed to increase linearly with their instantaneous power generation. Whereas this linear relationship might accurately reflect the need for operating reserves in systems with a low penetration of IRES, it can significantly overestimate the need for operating reserves in systems with a high penetration of IRES where there are regular periods with scheduled curtailment. Indeed, if there is scheduled curtailment, this scheduled curtailment directly reduces the exposure to forecast errors and hence the need for operating reserves, i.e., every unit of scheduled curtailment reduces the exposure to forecast errors by a single unit. This is visualized in Fig. 1. Even more so, if the scheduled generation from IRES is below a certain amount that can be guaranteed with a reasonable certainty, IRES could even provide upward reserves to deal with other types of uncertainty (e.g., demand forecast errors). The linear relationship between instantaneous power generation from IRES and the required operating reserves, as adopted in some planning models, does not appropriately account for this reduction of the exposure to forecast errors if there is scheduled curtailment. Further note that in most planning models, only upward reserves are considered. This because ensuring sufficient downward reserves tends to be rather inexpensive[29]. Therefore, in this paper, downward reserves are also not considered. In the simulations presented in this paper, the reference model (REF) adopts the reserve requirements from NREL’s Resource Planning Model (RPM) [29], as presented in Tab. 4. To assess the sensitivity to the assumptions regarding the requirements for operating reserves, two additional cases are considered. A first case does integrate all technical constraints of individual generators but does not impose operating reserve requirements (referred to as the ’No op res’ 6
Table 4: Overview of the reserve requirements adopted in different planning models Model RPM [29]
ReEDS [20]
Activation time
Sizing
Sub 5 minutes, 100% spin 10 minutes, 50% spin 1 hour, 100% spin
1% of demand Maximum of 6% of demand and the largest contingency 10% of wind generation + 7.5% of solar generation
Sub-minute, 100% spin 10 minutes, 50% spin roughly an hour, 17% spin
1.5% of demand
1 minute
To cover 99% of net load variations on 1-min time frame 1% of demand
NETPLAN [21] Sub 5 minutes, 100% spin 10 minutes 10 minutes 30 minutes
6% of demand Maximum difference in generation output between 2 consecutive hours in the last 15 days
Largest contingency Based on net load variability on 10-min time frame -
5 minutes Palmintier and 10 minutes, 50% Webster [5] spin
1% of demand + 0.385% of installed wind capacity Maximum of two largest generators and 3.3% of demand + 7.95% of installed wind capacity + 13.9% of instantaneous wind generation
De Jonghe et al. 60 [30] minutes
6.5% of installed wind capacity + 12.5% of instantaneous wind generation
case). A second additional case does impose reserve requirements, but does take into account that scheduled curtailment reduces the exposure to deal with IRES forecast errors, and hence the reserve requirements (referred to as the ’Curt’ case). An overview of the considered cases are presented in Tab. 5. Table 5: Overview of the considered cases regarding the requirements for operating reserves Case
Description
REF No op res Curt
Reserve requirements adopted from NREL’s RPM No operating reserve requirements considered Reserve requirements adopted from NREL’s RPM, but it is considered that scheduled curtailment reduces the need for operating reserves and IRES can provide upward reserves.
4
Case study description
All simulations in this paper are performed for a system inspired by the German electricity system for the year 2050. Furthermore, the optimization model is used in a greenfield mode, meaning that the focus is on a single year and no existing capacity is considered. The capacity factor time series for onshore and offshore wind generation and solar photovoltaic (PV) generation are taken from the EMHIRES data sets [31] for Germany, as provided 7
Probability
exposure to wind scheduled forecast curtailerrors ment genw (1 − α)W Wind generation
W
Figure 1: Illustration of the reduction of the need for operating reserves to deal with forecast errors if there is scheduled curtailment. A probability distribution of wind generation is presented. The forecasted wind generation when there would be no curtailment is indicated by W . The wind generation that can be guaranteed with a reasonable certainty is indicated by (1 − α)W , where α represents the uncertain fraction of the forecasted wind generation. The scheduled wind generation is indicated by genw . by the Strategic Energy Technologies Information System (SETIS) of the European Commission. These time series are scaled according to the endogenously determined capacity of wind and PV. The electricity demand time series for Germany are taken from the transparency platform provided by the European Network of Transmission System Operators for Electricity (ENTSO-E) [32]. To manage the computational complexity in the reference case, the year is represented by 8 representative weeks, which are selected using our model developed in [14]. Data regarding investment costs, fixed and variable operations and maintenance costs, as well as life times and lead times are taken from [33, 24]. Additionally, fuel prices are taken from the new policies scenario from the International Energy Agencys World Energy Outlook 2015 [34]. An exception is made for the fuel related costs for nuclear plants which are adopted from [24]. A discount rate of 5% is used to annualize the investment costs.
5
Results and discussion
First, Section 5.1 focuses on the impact of neglecting technical constraints for varying assumptions regarding the flexibility of thermal power plants and the availability of other flexibility providers. Subsequently, Section 5.2 analyzes the sensitivity of the model results to the assumptions taken regarding the characterization of the operating reserve requirements when integrating detailed technical constraints.
5.1
Cycling capabilities and other flexibility providers
Fig. 2a displays the projected total annual system costs for all scenarios and flexibility cases in the simulations where all technical constraints are included (REF) and the simulations where all technical constraints are omitted (MO). Additionally, Fig. 2b presents the capacity mix resulting from the different scenarios and flexibility cases, both with and without including technical constraints. By comparing the solution of the REF model to that of the MO model for the different cases, it can be observed from Fig. 2 that for the majority of scenarios and flexibility cases considered the integration of detailed technical constraints has only a limited impact on both the projections of the total system cost and the capacity mix. In addition, it can be observed that the impact of neglecting technical constraints tends to be higher whenever the share of IRES generation and less flexible baseload generation is higher. From Fig. 2, it can clearly be observed that, aside from the specific scenario, the impact of neglecting technical constraints is highly sensitive to the assumptions taken regarding the flexibility of thermal generators and the availability of other sources of flexibility. This can also be observed very clearly from Tab. 6, which summarizes, for each flexibility case, the average underestimation of the projected total annual system cost and the range of the underestimation 8
r
sto
r
Sto
No
r
sto
r Sto
REF Inflex REF Flex MO
No
REF Inflex REF Flex MO
S
REF Inflex REF Flex MO
s
REF Inflex REF Flex MO
No
REF Inflex REF Flex MO
40.0
S
MO
tor
tor
REF Inflex REF Flex MO
60.0
s No
REF Inflex REF Flex MO
80.0
REF Inflex REF Flex MO
Total annual system cost [BEUR/a]
REF tor
tor
20.0 0.0
uc rN
The
r The
nuc
No
o SN
nuc
IRE
IRE
uc SN
Scenarios
(a)
r
Sto
MO
REF Inflex REF Flex
MO
REF Inflex REF Flex
MO
MO
100.0
REF Inflex REF Flex
200.0
No
r r sto Sto
No
r r sto Sto
REF Inflex REF Flex
r
MO
sto
MO
No
REF Inflex REF Flex
r
Sto
REF Inflex REF Flex
r
MO
sto
PV WIND ONSHORE WIND OFFSHORE
MO
No
REF Inflex REF Flex
Capacity [GW]
300.0
CCGT OCGT PHS BAT
REF Inflex REF Flex
NUC CCGT CCS COAL SC CCS COAL SC
0.0
uc rN The
o rN The
nuc
uc on SN IRE Scenarios
uc SN IRE
(b) Figure 2: Impact of neglecting technical plant and system-level constraints on (a) the projected and effective total system cost and (b) the capacity mix for the different scenarios and flexibility cases considered. of the total system cost resulting from not considering technical constraints for the different scenarios. Although for the majority of cases and scenarios the impact of neglecting technical constraints is rather limited, a few exceptions exist for which neglecting technical constraints has a significant impact on the projections of the total system cost and/or has a significant impact on the capacity mix deemed optimal by the model. 5.1.1
Exception 1: flexibility scarcity
As can be observed from Tab. 6 and Fig. 2, a first exception is if thermal power plants are limitedly flexible and no other sources of flexibility are available (Inflex No stor case). In this case, neglecting technical constraints turns out to have a significant impact for all scenarios and particularly for the scenarios with high penetrations of IRES. Under these assumptions regarding the available flexibility, the value of flexibility can become extremely high. This is also apparent from the high difference in the projected total annual system cost between the In9
Table 6: Impact of not incorporating technical constraints for the different flexibility cases Case Average underestimation of the total system cost over the four considered scenarios [%] Minimim and maximum underestimation of the total system cost within the four considered scenarios [%]
Inflex No stor
Flex No stor
Inflex Stor
Flex Stor
18.6
4.1
4.1
2.0
5.0-34.0
1.2-6.7
3.4-5.1
1.1-3.0
Table 7: Total annual system cost [BEUR/a] for different assumptions regarding the availability and cost of storage technology-types Scenario
Inflex No stor
Inflex Stor
Inflex Storx3
IRES No nuc IRES Nuc
33.4 43.3
26.9 29.7
28.6 31.9
flex No stor case and the three other flexibility cases in scenarios IRES No nuc and IRES Nuc (see Fig. 2a). These high differences indicate that every source of flexibility, be it more flexible thermal generators, storage technology-types or other sources of flexibility, has the potential to reduce costs significantly (even if this source of flexibility would be highly expensive). For this reason, assuming that no source of flexibility will be found can be considered highly unrealistic. A resulting pitfall when integrating technical constraints in planning models is that overly conservative assumptions are taken regarding the available flexibility. Under such assumptions, the challenge of integrating large shares of IRES can be strongly overestimated, which in turn can lead to a strong overestimation of the projection of the total system cost and a technology bias. To illustrate this point, Tab. 7 presents the annual total system costs in the Inflex No stor and the Inflex Stor case for both scenarios with a high IRES penentration. Additionally, this table presents the annual total system costs one would obtain whenever the solution of the Inflex Stor case is taken, but the investment cost of all storage technology-types are multiplied by a factor of 3 (indicated by Inflex Storx3 ). This table shows that, even when storage technology-types would be three times as expensive as considered here, the total system costs would be significantly lower than those projected by a model in which storage technologytypes would not be considered at all. Similarly, in such a flexibility-constrained system, it can be expected that thermal power plants would be designed to offer more flexibility [35], or simply operated more flexibly at the expense of higher wear and tear costs. We therefore strongly recommend to account for other sources of flexibility and/or the ability to increase the flexibility of thermal power plants when incorporating detailed technical constraints in planning models. Integrating technical constraints with overly conservative assumptions regarding the available flexibility can introduce errors, both in terms of projecting the total system cost and deriving the optimal capacity mix, which can be significantly higher than the errors one would obtain if technical constraints would be completely omitted. 5.1.2
Exception 2: dedicated flexibility providers
A second exception relates to investments in storage technology-types, and especially batteries. From Fig. 2b, it can be observed that if technical constraints are omitted, no or few investments in batteries and PHS can be observed, whereas significant investments in storage technologytypes can be observed if technical constraints are incorporated. However, the investments in storage technologies (and batteries in particular) are highly sensitive to the assumptions taken regarding the flexibility of thermal power plants. This due to the fact that storage technologytypes are in direct competition with thermal generators and other flexibility providers for the provision of flexibility. Although not explicitly considered here, it can be expected that these 10
findings can be translated directly to other dedicated flexibility providers (e.g., active demand response).
5.2
Characterizing operating reserve requirements
This section analyzes the sensitivity to the assumptions made regarding the characterization of operating reserves when detailed technical constraints are considered. Tab. 8 shows the relative decrease in the projected total system cost if operating reserves requirements are omitted for the different scenarios and flexibility cases (REF versus No op res case). Table 8: Relative decrease of the total system cost when omitting operating reserve requirements [%] Scenario\flexibility case
Inflex No stor
Inflex Stor
Flex No stor
Flex Stor
3.6 2.5 24.4 29.0
0.6 1.0 1.3 1.0
0.9 0.3 4.2 2.8
0.5 0.4 1.0 0.7
Ther Nuc Ther No nuc IRES No nuc IRES Nuc
From this table, it can be observed that integrating operating reserve requirements in planning models can have a very high impact on the cost projections. However, the impact of considering operating reserve requirements is strongly dependent on the assumptions taken regarding the available flexibility. If thermal generators are assumed to be rather flexible and particularly if storage-technology types can provide reserves, the impact of reserves on the projections of the total system costs becomes almost negligible. This highlights the importance of not only considering different sources of flexibility, but also their ability to contribute to meeting operating reserve requirements. If there are high shares of IRES and thermal generators are assumed to be the only flexibility providers, ensuring sufficient operating reserves can become very expensive. This is due to multiple reasons. First, an increasing penetration of IRES increases the need for reserves to deal with possible deviations from the forecasted conditions. Second, with an increasing instantaneous generation of IRES, the number of thermal generators which need to be online to generate electricity is reduced. Therefore, a higher volume of reserves needs to be provided by fewer units. Ar moments of high IRES generation, a minimum number of spinning units needs to remain online in order to provide the reserve requirements. Since these units are bound to generate above a minimum threshold, i.e., the minimum operating point, curtailment of IRES generation will be required even though the entire demand is not served by IRES. At higher penetration levels of IRES, this need to curtail will occur more frequently. If thermal generators are assumed to be more flexible, ensuring sufficient operating reserves becomes less costly. In this regard, partularly the minimum operating point and the ramping capabilities play a key role. As the ramping capabilities increase and the minimum operating point decreases, thermal units can provide more reserves per unit of electricity generated. As a result, in periods of high wind and/or solar generation, IRES can provide a much higher fraction of the demand which strongly reduces the fuel costs. Storage systems have the inherent advantage that they can provide upward reserves without having to be generating electricity. Batteries are sufficiently fast to provide reserves without having to be charging or discharging. Whereas this is not necessarily the case for PHS systems, PHS systems can still provide upward reserves by charging (i.e., pumping). For instance, when charging at rated capacity, upward reserves can be provided by reducing the charging power. Particularly in systems with a high penetration of IRES, this offers the advantage that curtailment of IRES and the need to generate electricity using thermal generators can (to some extent) be avoided. Tab. 9 shows the investments in storage technologies for the different scenarios, flexibility cases and operating reserves cases. A first observation that can be made by comparing the investments in storage technologies in the REF case to those in the No op res case is that, 11
despite the limited impact of operating reserve requirements on the total system cost when storage systems can provide reserves, these operating reserve requirements form a determining factor for investments in storage technologies in general and batteries in particular. Table 9: Investments in storage technologies [GW] for the different scenarios, flexbility and operating reserve cases Scenario\ Case
REF
Inflex Stor Curt No op res
REF
Flex Stor Curt
No op res
Ther Nuc
PHS BAT
0.5 7.2
0.3 7.2
1.1 1.0
0 2.8
0 3
0 0.2
Ther No nuc
PHS BAT
0 8.1
0 7.7
0 0.8
0 1.2
0 1.1
0 0.3
IRES No nuc
PHS BAT
13.8 5.6
13.8 5.5
11.2 0.9
12.1 1.9
11.6 0.4
7.5 0.8
IRES Nuc
PHS BAT
20.4 4.8
20.2 4.3
18.4 0.7
17.8 1.1
17.4 0.8
13.7 0.5
A second observation that can be made is that the investments in storage technologies (and batteries in particular) are highly sensitive to both the assumptions taken regarding the flexibility of thermal generators (Inflex Stor versus Flex Stor case), and the assumptions taken for characterizing operating reserve requirements (REF versus Curt and No op res case). It must be noted that in this work we have only investigated the impact of certain assumptions. Other decisions which likely have a high impact on investments in storage technologies are the amount of reserves required, the required activation time, the ability of fast-starting units to provide non-spinning reserves and the ability of other sources of flexibility, such as active demand response, to contribute to the provision of reserves. We conclude by stating that, although operating reserve requirements are a determining factor for investments in storage technologies, caution is needed when implementing reserve requirements in planning models. First, if different sources of flexibility are not considered for the provision of reserves, the impact of reserves can be severely overestimated. Second, planning models typically make assumptions regarding both the sizing of reserves, the required activation time of different types of reserves as well as the extent to which different technologytypes can contribute to the provision of reserves. These assumptions can have a significant impact on the results.
6
Summary and conclusions
In the context of an increasing penetration of intermittent renewable energy sources, concerns regarding flexibility have driven modelers to integrate detailed technical constraints in generation expansion planning problems and energy-system optimization models. A number of authors have also assessed the impact of neglecting such detailed technical constraints in planning models. The overall consensus in the current literature seems to be that these technical constraints have a significant impact on the results of planning models, and should therefore not be neglected. However, the literature dealing with assessing the impact of neglecting technical constraints has certain limitations. First, in the current literature assessing the impact of neglecting flexibility constraints, thermal generators are often considered to be the only source of flexibility. Second, the sensitivity of the results to certain modeling assumptions, such as the cycling capabilities of thermal generators and the characterization of operating reserve requirements, have not been investigated. This paper contributes to the existing literature by assessing the relevance of integrating detailed technical constraints in planning models for a variety of assumptions regarding the available flexibility and the need for operating reserve requirements. A first main conclusion from the presented analysis is that the impact of neglecting technical constraints in planning models has a very limited impact on both the projections of the total 12
system cost, and the investments in thermal and renewable capacity ´ıf other technologies than thermal generators can contribute to the provision of flexibility. A second main conclusion is that integrating detailed technical constraints of power plants in combination with operating reserve requirements does have a significant impact on the investments in storage technologies (which can be generalized to other dedicated flexibility providers). If the planning model is used to investigate the role of storage technologies or other dedicated flexibility providers, considering these detailed technical constraints is thus essential. However, caution is needed as the investments in storage technologies in general, and batteries in particular, were shown to be highly sensitive to the assumptions taken regarding the cycling capabilities of thermal generators and the characterization of the operating reserve requirements. Finally, the presented analysis exposed a potential pitfall when integrating detailed technical constraints in planning models. Namely, if overly conservative assumptions are taken regarding the flexibility of thermal generators and no other flexibility providers are being considered (or they cannot contribute to the provision of operating reserves), the integration of detailed technical constraints can lead to a strong overestimation of the challenges related to integrating large shares of IRES. This can in turn lead to a high overestimation of the projected system costs and a strong technology bias. These errors can be signifcantly higher than the errors introduced by not integrating technical constraints in the planning model. To circumvent this potential pitfall, we recommend that different sources of flexibility and/or the ability to increase the flexibility of thermal generators should be explicitly considered when integrating technical constraints in planning models.
References [1] K. Poncelet, E. Delarue, D. Six, J. Duerinck, and W. D’haeseleer, “Impact of the level of temporal and operational detail in energy-system planning models,” Applied Energy, vol. 162, pp. 631–643, Jan. 2016. [2] S. Collins, J. P. Deane, K. Poncelet, E. Panos, R. C. Pietzcker, E. Delarue, and ´ Gallach´ B. O oir, “Integrating short term variations of the power system into integrated energy system models: A methodological review,” Renewable and Sustainable Energy Reviews, vol. 76, no. Supplement C, pp. 839 – 856, 2017. [Online]. Available: http://www.sciencedirect.com/science/article/pii/S1364032117304264 [3] A. Shortt, J. Kiviluoma, and M. O’Malley, “Accommodating variability in generation planning,” IEEE Transactions on Power Systems, vol. 28, no. 1, pp. 158–169, Feb 2013. [4] B. Palmintier and M. Webster, “Impact of unit commitment constraints on generation expansion planning with renewables,” in 2011 IEEE Power and Energy Society General Meeting, July 2011, pp. 1–7. [5] B. S. Palmintier and M. D. Webster, “Impact of operational flexibility on electricity generation planning with renewable and carbon targets,” IEEE Transactions on Sustainable Energy, vol. 7, no. 2, pp. 672–684, April 2016. ´ Gallach´oir, F. Rogan, M. Bazilian, [6] M. Welsch, P. Deane, M. Howells, B. O. and H.-H. Rogner, “Incorporating flexibility requirements into long-term energy system models a case study on high levels of renewable electricity penetration in ireland,” Applied Energy, vol. 135, pp. 600 – 615, 2014. [Online]. Available: http://www.sciencedirect.com/science/article/pii/S0306261914008836 [7] D. S. Kirschen, J. Ma, V. Silva, and R. Belhomme, “Optimizing the flexibility of a portfolio of generating plants to deal with wind generation,” in 2011 IEEE PES General Meeting, Jul. 2011, pp. 1–7. [8] J. Ma, V. Silva, R. Belhomme, D. S. Kirschen, and L. F. Ochoa, “Evaluating and planning flexibility in sustainable power systems,” IEEE Transactions on Sustainable Energy, vol. 4, no. 1, pp. 200–209, Jan 2013. 13
[9] A. van Stiphout, K. D. Vos, and G. Deconinck, “The impact of operating reserves on investment planning of renewable power systems,” IEEE Transactions on Power Systems, vol. 32, no. 1, pp. 378–388, Jan 2017. [10] A. Belderbos and E. Delarue, “Accounting for flexibility in power system planning with renewables,” International Journal of Electrical Power & Energy Systems, vol. 71, no. 0, pp. 33 – 41, 2015. [Online]. Available: http://www.sciencedirect.com/science/article/pii/ S0142061515001167 [11] F. J. de Sisternes, “Investment model for renewable electricity systems (IMRES): an electricity generation capacity expansion formulation with unit commitment constraints,” 2013, MIT Center for Energy and Environmental Policy Research Working Paper 2013-06. [Online]. Available: http: //home.eng.iastate.edu/∼jdm/wesep594/Investment%20Model%20for%20Renewable% 20Electricity%20Systems%20(IMRES) An%20Electricity%20Generation%20Capacity% 20Expansion%20Formulation%20with%20Unit%20Commitment%20Constraints’.pdf [12] F. J. de Sisternes and M. D. Webster, “Optimal Selection of Sample Weeks for Approximating the Net Load in Generation Planning Problems,” Jan. 2013. [Online]. Available: http: //citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.362.8358&rep=rep1&type=pdf [13] P. Nahmmacher, E. Schmid, L. Hirth, and B. Knopf, “Carpe Diem: A Novel Approach to Select Representative Days for Long-Term Power System Models with High Shares of Renewable Energy Sources.” [Online]. Available: http://dx.doi.org/10.2139/ssrn.2537072 [14] K. Poncelet, H. H¨ oschle, E. Delarue, A. Virag, and W. Dhaeseleer, “Selecting representative days for capturing the implications of integrating intermittent renewables in generation expansion planning problems,” IEEE Transactions on Power Systems, vol. 32, no. 3, pp. 1936–1948, May 2017. [15] A. Flores-Quiroz, R. Palma-Behnke, G. Zakeri, and R. Moreno, “A column generation approach for solving generation expansion planning problems with high renewable energy penetration,” Electric Power Systems Research, vol. 136, pp. 232 – 241, 2016. [Online]. Available: http://www.sciencedirect.com/science/article/pii/S0378779616300177 [16] B. S. Palmintier and M. D. Webster, “Heterogeneous unit clustering for efficient operational flexibility modeling,” IEEE Transactions on Power Systems, vol. 29, no. 3, pp. 1089–1098, May 2014. [17] B. Palmintier, “Flexibility in generation planning: Identifying key operating constraints,” in Power Systems Computation Conference (PSCC), 2014. IEEE, 2014, pp. 1–7. [18] L. Zhang, T. Capuder, and P. Mancarella, “Unified unit commitment formulation and fast multi-service lp model for flexibility evaluation in sustainable power systems,” in 2016 IEEE Power and Energy Society General Meeting (PESGM), July 2016, pp. 1–1. [19] K. Poncelet, “Long-term energy-system optimization models: Capturing the challenges of integrating intermittent renewable energy sources and assessing the suitability for descriptive scenario analyses,” Ph.D. dissertation, KU Leuven, 2018. [Online]. Available: https://lirias.kuleuven.be/bitstream/123456789/604991/1/thesis final.pdf [20] W. Short, P. Sullivan, T. Mai, M. Mowers, C. Uriarte, N. Blair, D. Heimiller, and A. Martinez, “Regional Energy Deployment System (ReEDS),” National Renewable Energy Laboratory (NREL), Tech. Rep., 2011, nREL/TP-6A20-46534. [Online]. Available: http://www.nrel.gov/docs/fy12osti/46534.pdf [21] V. Krishnan, T. Das, E. Ibanez, C. A. Lopez, and J. D. McCalley, “Modeling operational effects of wind generation within national long-term infrastructure planning software,” IEEE Transactions on Power Systems, vol. 28, no. 2, pp. 1308–1317, May 2013.
14
[22] P. Nahmmacher, E. Schmid, and B. Knopf, “Documentation of limes-eu - a long-term electricity system model for europe,” Potsdam Institute of Climate Impact Research (PIK), https://www.pik-potsdam.de/members/paulnah/limes-eu-documentation-2014.pdf. ´ Gallach´oir, “Soft-linking of a power [23] J. Deane, A. Chiodi, M. Gargiulo, and B. P. O. systems model to an energy systems model,” Energy, vol. 42, no. 1, pp. 303 – 312, 2012, 8th World Energy System Conference, WESC 2010. [Online]. Available: http://www.sciencedirect.com/science/article/pii/S0360544212002551 [24] A. Schr¨ oder, F. Kunz, J. Meiss, R. Mendelevitch, and C. von Hischhausen, “Data documentation- current and prospective costs of electricity generation until 2050,” dIW Working Paper, https://www.diw.de/documents/publikationen/73/diw 01.c.424566.de/ diw datadoc 2013-068.pdf. [25] K. V. den Bergh and E. Delarue, “Cycling of conventional power plants: Technical limits and actual costs,” Energy Conversion and Management, vol. 97, no. Supplement C, pp. 70 – 77, 2015. [Online]. Available: http://www.sciencedirect.com/science/article/pii/ S0196890415002368 [26] N. Kumar, P. Besuner, S. Lefton, A. D., and D. Hilleman, “Power plant cycling costs,” NREL, Tech. Rep., 2012. [Online]. Available: https://www.nrel.gov/docs/fy12osti/55433. pdf [27] K. Poncelet, A. van Stiphout, J. Meus, E. Delarue, and W. D’haeseleer, “LUSYM Invest: a generation expansion planning model with a high level of temporal and technical detail,” KU Leuven, TME Working Paper WP EN2018-07, https://www.mech.kuleuven.be/en/ tme/research/energy environment/Pdf/WP LUSYM INVEST.pdf. [28] J. Meus, K. Poncelet, and E. Delarue, “Applicability of a clustered unit commitment model in power system modeling,” IEEE Transactions on Power Systems, vol. PP, no. 99, pp. 1–1, 2017. [29] T. Mai, C. Barrows, A. Lopez, E. Hale, M. Dyson, and K. Eurek, “Implications of Model Structure and Detail for Utility Planning: Scenario Case Studies Using the Resource Planning Model,” National Renewable Energy Laboratory (NREL), Tech. Rep., 2015, nREL/TP-6A20-63972. [Online]. Available: https: //www.nrel.gov/docs/fy15osti/63972.pdf [30] C. D. Jonghe, E. Delarue, R. Belmans, and W. Dhaeseleer, “Determining optimal electricity technology mix with high level of wind power penetration,” Applied Energy, vol. 88, no. 6, pp. 2231 – 2238, 2011. [Online]. Available: http://www.sciencedirect.com/science/article/pii/S0306261910005660 [31] Strategic Energy Technologies Information System, European Commission, “EMHIRES data sets,” 2017, available at https://setis.ec.europa.eu/EMHIRES-datasets. [32] European Network of Transmission System Operators for Electricity (ENTSO-E), “ENTSO-E Transparency Platform,” 2017. [Online]. Available: https://transparency. entsoe.eu/ [33] S. Simoes et al., “The JRC-EU-TIMES model - Assessing the long-term role of the SET Plan Energy technologies,” JRC’s Institute for Energy and Transport, Tech. Rep., 2013. [34] International Energy Agency (IEA), World Energy Outlook 2015, 2015. [35] H. Holttinen, A. Tuohy, M. Milligan, E. Lannoye, V. Silva, S. M¨ uller, and L. S¨o”, “The flexibility workout: Managing variable resources and assessing the need for power system modification,” IEEE Power and Energy Magazine, vol. 11, no. 6, pp. 53–62, Nov 2013.
15
