Generation Adequacy in the Nordic and Baltic Region: Case Studies from 2020 to 2050: Flex4RES Project Report

(1)

Project Report

Generation Adequacy in the Nordic and Baltic Region: Case Studies from 2020 to

2050

Author:

Alessandro Crosara

Supervisors:

Lennart S¨oder Egill T´omasson

School of Electrical Engineering and Computer Science (EECS) Division of Electric Power and Energy Systems (EPE) Integration of Renewable Energy Sources (IRES) Group

May 2019

(2)

Abstract

Generation adequacy is a concern in today’s electricity market where intermittent renewable energy sources are rapidly becoming a greater share of the generation mix. This study focuses on the Nordic and Baltic power system that is comprised of the system areas of the Nord Pool spot market. Sequential Monte Carlo Simulation is applied to assess the generation adequacy of this multi-area system for several future case studies, based on scenarios defined within the Nordic Flex4RES project. The report gives insights into the characteristics of these adequacy problems that the system could face in a more sustainable future, quantifies their magnitude and presents their characteristics. Finally, a solution based on the demand flexibility of residential electric heating is discussed, as a way to counter capacity deficit problems.

(3)

Acknowledgements

First and foremost, I would like to express my gratitude to my two supervisors, Lennart S¨oder and Egill T´omasson, for the opportunity of keeping working in this topic, and for their constant support and helpful insights during all of these months.

I would also like to express my appreciation for Anders Nilsberth, Erik Hellström and Erik Böhlmark from Svenska Kraftnät, the Swedish Transmission System Operator (TSO), for their input when developing the generation adequacy analysis tool as well as when gathering data about the North-European power system.

Many thanks to the members of the Flex4RES project too, especially Matti Juhani Koivisto, Philipp Andreas Gunkel and Hardi Koduvere, for their support with data collection.

I would also like to thank Lars Herre and Pontus Dahlstr¨om for the data regarding the flexibility of residential electric heating in Sweden.

Moreover, I would like to thank all of my colleagues at KTH for the nice time spent in these months.

Last but not least, I sincerely thank my family and my friends, who always support me throughout every challenge.

ii

(4)

List of Figures

2.1 The Nordic and Baltic power system in 2020, and its interconnections (adapted from [1]). 4

3.1 Module-based generation adequacy analysis tool. . . . . 6

4.1 The four Flex4RES scenarios [2]. . . . . 10

4.2 Generation capacity mix for the different case studies (data source: Flex4RES. Data retrieved as of March 2019). . . . . 13

5.1 LOLE trends in areas SE3 and EE. . . . . 23

5.2 EENS trends in areas SE3 and EE. . . . . 23

5.3 PNS histogram for SE3 in year 2050, scenario 2. . . . . 24

5.4 PNS patterns for SE3 in year 2050, scenario 2, for an example period of 150 years. Outages with a duration in the range 0 ≤ h ≤ 8 are drawn in blue, while those in the range 8 < h ≤ ∞ are drawn in orange. . . . . 24

5.5 Relation between ENS and peak PNS per load shedding event for SE3 in year 2050, scenario 2. A total number of 24830 events occurred in the simulation of 14000 Monte Carlo years. . . . . 24

5.6 Relation between ENS and peak PNS per load shedding event for SE3 in year 2050, scenario 2. The probability of some peak PNS levels is drawn, with respect to the total number of load shedding events occurred in the simulation. Only events with an ENS ≤ 10000 MWh and a peak PNS ≤ 2000 MW are displayed in the figure (that is, only 83% of events are shown). . . . . 25

5.7 Relation between ENS and peak PNS per load shedding event for SE3 in year 2050, scenario 2. The probability of some ENS levels is drawn, with respect to the total number of load shedding events occurred in the simulation. Only events with an ENS ≤ 10000 MWh and a peak PNS ≤ 2000 MW are displayed in the figure (that is, only 83% of events are shown). 25 6.1 The impact of flexible residential electric heating: PNS histogram for SE3 in year 2050, scenario 2. . . . . 28

6.2 The impact of flexible residential electric heating: PNS patterns for SE3 in year 2050, scenario 2, for an example period of 150 years. Outages with a duration in the range 0 ≤ h ≤ 8 are drawn in blue, while those in the range 8 < h ≤ ∞ are drawn in orange. . . 28

6.3 The impact of flexible residential electric heating: relation between ENS and peak PNS per load shedding event for SE3 in year 2050, scenario 2. A total number of 8310 events occurred in the simulation of 14000 Monte Carlo years.. . . . 28

6.4 The impact of flexible residential electric heating: relation between ENS and peak PNS per load shedding event for SE3 in year 2050, scenario 2. The probability of some peak PNS levels is drawn, with respect to the total number of load shedding events occurred in the simulation. Only events with an ENS ≤ 10000 MWh and a peak PNS ≤ 2000 MW are displayed in the figure (that is, only 61% of events are shown). . . . . 29

6.5 The impact of flexible residential electric heating: relation between ENS and peak PNS per load shedding event for SE3 in year 2050, scenario 2. The probability of some ENS levels is drawn, with respect to the total number of load shedding events occurred in the simulation. Only events with an ENS ≤ 10000 MWh and a peak PNS ≤ 2000 MW are displayed in the figure (that is, only 61% of events are shown). . . . . 29

iv

(6)

List of Tables

3.1 FORs, MTTRs and MTTFs used in this project. . . . . 7 4.1 Scenario comparison in terms of the total installed generation capacity, total VRE installed

capacity, total interconnection NTC and total peak load. (Data source: Flex4RES. Data retrieved as of March 2019). Notice: the total peak load comprises also the contribution given by electric boilers, heat pumps, uncontrolled PHEV and BEV charging (see section 4.1). . . . . 11 4.2 Per-area peak loads for each case study (data source: Flex4RES. Data retrieved as of

March 2019). . . . . 13 4.3 Capacity factors and full load hours for onshore wind power (data source: DTU Wind

Energy. Data retrieved as of November 2018). . . . . 14 4.4 Capacity factors and full load hours for offshore wind power (data source: DTU Wind

Energy. Data retrieved as of November 2018). . . . . 14 4.5 Per-area onshore wind power installed capacities for each case study (data source: Flex4RES.

Data retrieved as of March 2019).. . . . 15 4.6 Per-area offshore wind power installed capacities for each case study (data source: Flex4RES.

Data retrieved as of March 2019).. . . . 15 4.7 Per-area hydro power installed capacities for each case study. The Flex4RES data (re-

trieved as of March 2019) are here already scaled.. . . . 16 4.8 Per-area nuclear power installed capacities for each case study (data source: Flex4RES.

Data retrieved as of March 2019).. . . . 16 4.9 Per-area natural gas power installed capacities for each case study (data source: Flex4RES.

Data retrieved as of March 2019).. . . . 17 4.10 Per-area coal power installed capacities for each case study (data source: Flex4RES. Data

retrieved as of March 2019). . . . . 17 4.11 AC interconnections’ maximum NTC for each case study (data source: Flex4RES. Data

retrieved as of March 2019). . . . . 19 4.12 DC interconnections’ maximum NTC for each case study (data source: Flex4RES. Data

retrieved as of March 2019). . . . . 20 5.1 Results: year 2040, scenarios 2 and 4, respectively. . . . . 22 5.2 Results: year 2050, scenarios 2 and 4, respectively. . . . . 22

v

(7)

Abbreviations

BAU Business As Usual BEV Battery Electric Vehicles CV Coefficient of Variation

EENS Expected Energy Not Supplied ENS Energy Not Supplied

EPNS Expected Power Not Supplied FOR Forced Outage Rate

LOLD Loss Of Load Duration LOLE Loss Of Load Expectation LOLF Loss Of Load Frequency LOLO Loss Of Load Occasion MTTF Mean Time To Failure MTTR Mean Time To Repair NTC Net Transfer Capacity

PHEV Plug-in Hybrid Electric Vehicles PNS Power Not Supplied

VRE Variable Renewable Energy

vi

(8)

Chapter 1

Introduction

This report is an extended version of [3].

The aim of this study is to estimate the risk of capacity deficit in the multi-area power system that is comprised of the Nordic and Baltic electric power systems of the Nord Pool Spot market. The report presents results of sequential Monte Carlo simulations carried out for a base scenario (defined for the year 2020) as well as for future scenarios for the years 2030, 2040 and 2050. The main reliability metrics used to quantify the risk of capacity deficit are the Loss Of Load Expectation (LOLE) and the energy related metric of the Expected Energy Not Supplied (EENS). In addition, the Loss Of Load Duration (LOLD) and the Loss Of Load Frequency (LOLF) indices are reported.

The different scenarios studied are based on scenarios defined within the Nordic Energy Research project Flex4RES [2] (results retrieved as of March 2019). In the Flex4RES project, the assumptions from the Nordic Energy Technology Perspectives 2016 report [4] are used as a starting point. The project was initiated to identify and assess regulatory and technical pathways towards coherent Nordic energy systems in 2050 based on strong interactions between different energy markets that ensure resilience, sustainability and efficiency. However, the Flex4RES project, which uses Balmorel as a model, is such that load shedding situations are not allowed in the simulations, and capacity deficit problems are solved by making investments in the power system, such as in new power plants and in new transmission lines.

On the contrary, quantifying the generation adequacy of the power system is among the purposes of this study.

34 year long meteorological data sets are used to construct time series for wind and solar power generation and gradient boosting machine learning algorithms are implemented to predict future electricity demand.

As a result, the report quantifies the capacity deficit of the power system by assuming non-flexible demand. However, in its final part, the report explores the potential of temporary disconnection of residential electric heating, in order to mitigate generation adequacy issues.

1

(9)

Chapter 1. Introduction 2

The report is structured as follows. Chapter 2 introduces the system that is modelled in the study.

Chapter3describes the sequential Monte Carlo reliability evaluation model used, as well as the reliability indices that quantify the reliability of the multi-area power system. Chapter 4 presents the scenarios and case studies which were analyzed. Chapter5 provides the results from the case studies, followed by a discussion in Chapter6, where the potential of flexible residential electric heating is evaluated. Lastly, chapter7ends the report with the conclusions and some suggestions for future work.

(10)

Chapter 2

System Description

The electric power system is modeled as a multi-area power system, with area division as in the 15-area Nord Pool Spot market [1]: SE1–SE4 for Sweden, FI for Finland, NO1–NO5 for Norway, DK1–DK2 for Denmark, and EE, LV, and LT for the Baltic countries of Estonia, Latvia, and Lithuania, respectively.

The different system areas are interconnected through transmission lines whose transmission limits are captured by their Net Transfer Capacities (NTCs)¹. Moreover, some of the areas are connected with foreign countries (meaning non-Nordic and non-Baltic countries). Figure2.1shows the Nordic and Baltic power system, where the interconnections (both AC and DC links, between system areas) as of year 2020 are drawn as arrows. In general, an arrow stands for one or multiple lines.

The areas are characterized with the following random variables:

• Electricity demand.

• Generation from Variable Renewable Energy (VRE) sources, i.e. onshore wind, offshore wind, and solar PV.

• An available power generation capacity from all the other conventional sources (that is, except for wind and solar power), which will be denoted from now on as available other generation capacity.

• A Net Transfer Capacity (NTC) for each interconnection between system areas.

• A Net Transfer Capacity (NTC) for each import from a foreign area (not belonging to the system).

The above variables are modeled in different ways, depending on their nature (a more detailed description is given in Chapter 4). First, the demand and the VRE generation, which have the property of being highly variable in time, are given as inputs in the format of time series. Second, both interconnections

1The definition of NTC, according to Nord Pool [5], is the following: “NTC is the maximum exchange program between two areas compatible with security standards applicable in both areas and taking into account the technical uncertainties on future network conditions”.

3

(11)

Chapter 2. System description 4

Figure 2.1: The Nordic and Baltic power system in 2020, and its interconnections (adapted from [1]).

and imports are modeled as individual or multiple transmission lines between (or to) system areas. Each line is either available up to its full NTC or not available at all due to an outage. Notice that transmission lines are treated as unidirectional, so for each real transmission line linking two system areas, there exist two lines in the model², each of them having a possibly different NTC depending on the direction of the power flow. Whenever one of these two lines is unavailable, the other line is considered unavailable too: in reality, if a line is subject to a fault, power cannot flow in any direction. A transmission line’s availability is based on an exponential distribution with a certain Mean Time To Failure (MTTF) as well as a certain Mean Time To Repair (MTTR). Based on these parameters, an outage time series is generated for each transmission line. This applies to both AC and DC lines, where the DC lines are in general much less reliable. Third, the other generation (which comprises all the power generation sources apart from wind and solar power) is modeled in a similar way as transmission lines. Generation units belonging to this group are available most of the time and, when available, their generation capacity typically equals their installed capacity. Each generation unit is then assigned a MTTF and a MTTR, and based on them an outage time series is generated for each generation unit.

2Instead, since power exports are not considered, for import links the number of lines modeled is the same as in reality.

(12)

Chapter 3

Reliability Evaluation Model

In this project, the adequacy of the power system is estimated by means of a sequential Monte Carlo simulation. The basic steps of a Monte Carlo simulation are the following:

1. Sample the system for a certain number of hours, i.e. randomly generate a batch of inputs.

2. Compute the system outputs for each sample of the generated batch.

3. Store the result sums, update the estimates of the outputs of interest, and test the pre-defined stopping rule. If the latter is satisfied, stop the simulation, else repeat from step1.

The distinguishing trait of a sequential Monte Carlo simulation is that system samples are generated in batches which are related in time. Load loss events are encountered sequentially and therefore not only the specific random system states but also the transition process between system states can be simulated [6]. That is, the effects of the load loss event duration, severity, and frequency can be considered in the simulation. Figure 3.1 shows a diagram of the module based generation adequacy analysis tool developed for the study. Its different parts are described in the rest of this chapter.

3.1 Sample Year Generator

Generating a system sample means to randomly assign a value to each system input variable for a single hour. That is, demand, available VRE generation and available other-generation for each area, and interconnection or import NTC for each transmission line. Samples are generated in 34-year-long batches which are related in time. This means that at each iteration, a 34-year-long time series is created for each system variable.

5

(13)

Chapter 3. Reliability evaluation model 6

34-year-long time series for the demand, onshore and offshore wind power, as well as solar power are given as input to the simulation. This means that in each simulation batch of a given case study, such 34- year-long time series remain the same. How they are obtained is case study dependent, and is explained in Chapter4.

Conversely, 34-year-long stochastic outages time series are generated for interconnection and import lines and for generation units at each iteration of the simulation. Generally speaking, such time series differ at each iteration of the simulation. Outages time series are obtained assuming that each transmission line or generation unit’s availability follows an exponential distribution with given MTTF and MTTR, following the state duration sampling approach described in [6]. The FORs, MTTRs and MTTFs used in this project are shown in Table3.1.

For the generation units, hydro power aside (see section 4.3), the FORs used are computed from more detailed data provided Svenska Kraftn¨at, the Swedish TSO. For coal and nuclear the MTTR values are the weighted average of the values for such power plants in the IEEE-RTS 79 [7]. For the other generation units, the MTTR is set equal to 51 h, which is the weighted average of all the power sources in the IEEE-RTS 79. The MTTFs are computed according to FORs and MTTRs.

In addition, the FORs used for both AC and DC lines, and the MTTR used for AC lines, are suggested as average values by Svenska Kraftn¨at. The MTTF for the AC lines are computed accordingly. It is also assumed that the MTTF for AC lines equals the MTTF for DC lines. Lastly, the MTTR for DC lines is computed based on the FOR and MTTF assumed for such a technology.

3.2 Monte Carlo Optimization Module

For a given system sample, the available generation capacity can be computed for each area as the sum of the available VRE generation, the available other generation capacity, and the available imports to

Sample Year Generator Generates 34 sample years

for demand and VRE time series as well as stochastic outage time series

MC Optimization Module Solves an optimization problem minimizing loss of

load hourly via sequential Monte Carlo Simulation

Power System data Generators and interconnections, 34-year-long meteorological data set

Aggregator Aggregates results and

updates estimates of reliability indices

50 million winter hours

evaluated?

Return Results LOLE, EENS, LOLD, LOLF, etc.

no yes

Figure 3.1: Module-based generation adequacy analysis tool.

(14)

that area (if assumed possible in a given case study). Furthermore, the available transmission capacity between the different areas can be computed based on which lines are in operation. An optimization problem is then constructed and solved to determine whether load shedding is necessary in the system for the given hour. The process is repeated for each sample belonging to the batch generated in the previous step. The optimization problem used to assess a system sample i is given in (3.1),

min

fit,dia,pia

n_A

X

a=1

x^L_ia−

n_A

X

a=1

dia=

n_A

X

a=1

(x^L_ia− dia) (3.1a)

s.t. pia+ X

t∈Ψ⁻_a

fit= dia+ X

t∈Ψ⁺_a

fit ∀a (3.1b)

0 ≤ x^L_ia− d_ia≤ M_ia· (1 − u_ia) ∀a (3.1c)

0 ≤ fit≤ x^T_it ∀t (3.1d)

0 ≤ pia≤ p_ia ∀a, (3.1e)

where x^L_ia is the demand in area a, d_ia is the served load in area a, p_ia is the generation in area a, f_it is the power flow in interconnection t, p_ia is the available generation capacity in area a, x^T_it is the NTC of interconnection t, and Mia= x^L_ia− p_ia indicates the difference between the demand and the available generation capacity in each area a. uia is a binary parameter which indicates whether each area a is self-sufficient (uia = 1), i.e. can generate at least as much power as its demand, or not (uia= 0). Ψ⁻_a denotes the set of interconnections entering area a, while Ψ⁺_a indicates the set of interconnections exiting area a.

The optimization problem minimizes the difference between the total demand and the total served load (3.1a). Denoting n_A as the number of system areas, this optimization problem has n_A load balance constraints (3.1b), which ensure that the sum of the generated power and all the power flows entering an area must equal the sum of the served load and all the power flows exiting that area. There are nA inequality constraints (3.1c) too, which force each area to prioritize its own power supply before transmitting power to any neighbouring area. They also ensure a certain level of served demand in

Table 3.1: FORs, MTTRs and MTTFs used in this project.

Technology FOR MTTR [h] MTTF [h]

Generation units

Hydro 0 / /

Nuclear 0.05 150 2850

Coal 0.0875 47 490

Gas 0.065 51 734

Other 0.08 51 587

Transmission lines

AC lines 0.0005 5 9995

DC lines 0.035 363 9995

(15)

each area: if in a given area the available generation capacity is not sufficient to cover the demand, the load served must anyway be greater than or equal to the available generation capacity in that area. In addition, there are variable limits for the power flow in each transmission line (3.1d), and the power generation in each area (3.1e).

3.3 Aggregator

Solving the optimization problem returns the served load in each area, the power flow on each transmission link, and the power generation in each area. Based on the served load, two additional output variables are computed for each hour, namely the Power Not Supplied (PNS) and the Loss Of Load Occasion (LOLO).

The PNS is the power that cannot be delivered due to capacity limitations (generation or transmission) in the system [6]. For a given system sample i and for each area a, the PNS can be defined as:

P N S_ia= x^L_ia− d_ia (3.2)

The LOLO is a binary indicator which is equal to one if load shedding occurs for a specific hour and zero otherwise [8]. For a given system sample i and for each area a, the LOLO can be defined as:

LOLOia=







1 if x^L_ia− dia= P N Sia> 0 0 otherwise

(3.3)

Several reliability indices are used to quantify the reliability of each power system area, namely the Loss Of Load Expectation (LOLE), the Expected Energy Not Supplied (EENS), the Loss Of Load Duration (LOLD) as well as the Loss Of Load Frequency (LOLF) [6,9,10]. The LOLE [h/year] is given by the expected value of the LOLO. The Expected Power Not Supplied, EPNS [MW], is the expected value of the PNS and it is computed in order to obtain the EENS. The EENS [MWh/year] is given by the product of the EPNS with the number of hours contained in a year, or can alternatively be computed as the expectation of the Energy Not Supplied (ENS). Furthermore, the LOLD [h/occurrence] and the LOLF [occurrences/year] represent the expected duration and frequency of outages suffered by each system area.

At each simulation iteration, the estimates of the reliability indicators are updated, and so are their corresponding coefficients of variation (CV), which are a measure of the accuracy of the estimates.

According to [6], the EENS has the lowest rate of convergence. Being the EENS estimate computed in this work based on the EPNS estimate, as mentioned in the previous paragraph, the CV values presented in Chapter5refer directly to the PNS.

(16)

3.4 Stopping Rule

A number of stopping rules can be defined in a Monte Carlo simulation to decide if, after having collected a certain number of samples, another batch of samples should be generated or if instead the simulation should end. In this report, a fixed-number-of-samples rule is used, such that the simulation will stop only after 50 million hours have been simulated.

As explained in Chapter4, only winter months are simulated, which means that 50 million hours comprise only winter hours. As a result, the simulation will stop when approximately 14000 Monte Carlo years have been simulated.

(17)

Chapter 4

Scenarios and Case Studies

In the Flex4RES project, four different scenarios are defined and investigated. Scenario 1 is a business- as-usual (BAU) case. Scenario 2 considers a larger transmission capacity between the system areas, if compared to the BAU scenario. Scenario 3 investigates, if compared to scenario 1, the removal of regulatory barriers and the introduction of incentives to support a greater flexibility. Lastly, scenario 4 is a combination of scenarios 2 and 3: if compared to the BAU case, the transmission capacity is increased and regulatory barriers are removed, so to obtain maximum flexibility. Figure4.1summarizes the differences between the four scenarios.

Figure 4.1: The four Flex4RES scenarios [2].

In this paper, simulations for 7 different case studies are carried out, mostly based on data from specific years and scenarios defined in the Flex4RES project. Scenario 1 is used only for the year 2020 and serves as a reference case. Scenario 2 is analyzed for the years 2030, 2040 and 2050. Finally, scenario 3 is neglected and scenario 4 is investigated for the years 2030, 2040 and 2050. A comparison of some relevant quantities in the different case studies is given in Table4.1. In all of the case studies, imports

10

(18)

Chapter 4. Scenarios and Case Studies 11

were assumed not possible, following a conservative approach. Therefore, load shedding is interpreted as a circumstance where the isolated Nordic and Baltic system is unable to fulfill the demand by itself.

Table 4.1: Scenario comparison in terms of the total installed generation capacity, total VRE installed capacity, total interconnection NTC and total peak load. (Data source: Flex4RES. Data retrieved as of March 2019). Notice: the total peak load comprises also the contribution given by electric boilers,

heat pumps, uncontrolled PHEV and BEV charging (see section4.1).

GEN. CAP.

[GW]

VRE CAP.

[GW]

NTC [GW]

Peak Load [GW]

2020 Sc.1 121.8 24.3 133.0 67.2

2030 Sc.2 142.6 57.2 159.0 73.5

Sc.4 143.5 58.6 158.2 75.0

2040 Sc.2 150.8 73.1 167.1 78.8

Sc.4 150.8 73.6 166.7 79.2

2050 Sc.2 155.5 84.7 174.4 81.1

Sc.4 157.2 85.0 175.0 81.4

In the following sections, the data used in each case study are presented. Note that for each year, only the hourly value corresponding to winter months (January, February, March, November and December) are used¹. A total of 50 million winter hours are simulated in each case study, corresponding to approximately 14 000 Monte Carlo years.

The simulations are run in Julia, version 0.6.4. The solver used for the optimization problem is Gurobi, version 7.5.2.

4.1 Load

Hourly load time series are obtained by scaling preliminary time series, provided by Dr. Jon Olauson, KTH, according to peak loads computed for each area and each case study from Flex4RES data.

The preliminary load time series are generated based on a load model developed for each system area.

The goal was to create long time series (1980–2017) for the load, based on the historical load in some recent years, following a method similar to [12]. That is, the simulated load in e.g. 1980 should be interpreted as the load that would have been observed with the same meteorological conditions as in 1980, but with the consumption patterns of the recent years. In the model, gradient boosting machine learning algorithms were implemented [13]. The following predictors were used:

• Temperature and wind speed from the MERRA2 data set (empirical orthogonal functions were used to reduce the dimensionality [14]).

1The reason why only winter months are simulated is that they are the most critical for the Nordic countries due to the highest peaks in the electricity demand [11].

(19)

• Lagged temperature time series.

• Month of the year, hour of the day and weekday.

• Holidays (according to the Swedish calendar).

• An economic indicator for Sweden.

• A linear trend (only used for NO3 and NO5).

Three years (either 2013–2015 or 2014–2016, depending on the system area) of historical observations were used to train the model, and one year (either 2016 or 2017) was used for testing. Looking at the test data, the performance of the model was satisfactory. The correlation between simulated and observed loads varied between 0.94 and 0.99 for the 15 system areas considered (with 0.97 as an average). The mean absolute errors relative to the mean load for each area varied between 2.6% and 6.2% (with 3.8%

as an average). Out of the 38 years (1980–2017) of generated load time series, only 34 years (1982–2015) are selected when defining the preliminary time series, so to make them synchronous with the VRE time series.

Finally, the peak load values are computed from Flex4RES data and account also for electric boilers, heat pumps, uncontrolled Battery Electric Vehicles (BEV) charging and uncontrolled Plug-in Hybrid Electric Vehicles (PHEV) charging. It should be noted that by considering the charging as uncontrolled – rather than smart – a conservative approach is followed in the study. BEV and PHEV, which could be a resource for the power system if used smartly (e.g. avoiding charging in the periods with demand peaks) are instead represented as mere load components in this model. The per-area peak load values obtained in this way are shown in Table4.2. They are used to scale the preliminary time series, so that in each area the maximum load experienced coincides with the computed peak load for that area.

4.2 Variable Renewable Energy Sources

Hourly VRE (solar PV power and wind power) time series in p.u. are provided for all the case studies by DTU Wind Energy, and are obtained from the VRE time series modelling tool CorRES [15]. They are based on 34 meteorological years (1982-2015) and on assumptions about wind power technological evolution, including higher hub heights for future installations, lower specific power and offshore wind farms further from the shore. Such assumptions are quantitatively summarized for onshore wind power and offshore wind power, respectively, in Tables4.3 and4.4. In such tables, FIn and FIs stand for the north of Finland and the south of Finland. Notice that it is assumed that the same values stated for the year 2014 applies to the year 2020 too. Moreover, no modelling of technological change for solar PV power is considered.

(20)

Table 4.2: Per-area peak loads for each case study (data source: Flex4RES. Data retrieved as of March 2019).

Area Peak load [MW]

2020 Sc. 1

2030 Sc. 2

2030 Sc. 4

2040 Sc. 2

2040 Sc. 4

2050 Sc. 2

2050 Sc. 4

SE1 1517 1656 1670 1811 1811 1903 1901

SE2 3059 3201 3272 3445 3463 3491 3488

SE3 14555 15567 15993 16358 16787 17050 17056

SE4 4286 4571 4599 4798 4782 4951 4993

FI 12640 13584 13715 14332 14310 14903 14906

NO1 6732 7397 7541 7723 7724 7889 8023

NO2 5469 6030 6048 6382 6384 6617 6632

NO3 3901 4272 4322 4514 4516 4655 4659

NO4 2739 3038 3055 3233 3233 3357 3358

NO5 2351 2629 2647 2815 2815 2924 2925

DK1 3349 3661 4039 4038 4038 4023 4100

DK2 2310 2742 2941 3196 3196 3265 3282

EE 1392 1583 1601 1830 1830 1926 1927

LV 1130 1404 1404 1736 1736 1718 1718

LT 1742 2146 2146 2590 2590 2473 2473

The generated time series are obtained for each case study by multiplying the p.u. time series with the onshore wind, offshore wind and solar installed capacities in each area. The per-area installed capacities for onshore and offshore wind power (the most significant VRE sources) are shown in Tables4.5and4.6, respectively, for each case study.

The share of VRE sources out of the total installed capacity in each case study can be observed, together with the other generation, in Figure4.2, where the generation capacity mix for the different case studies is shown.

2020 Scenario 1 2030 Scenario 2 2030 Scenario 4

2040 Scenario 2 2040 Scenario 4 2050 Scenario 2 2050 Scenario 4

Coal & Gas Nuclear Wind & solar Hydro

Remaining technologies

10.2%

19.9%

9.8% 45.4%

14.7%

3.2%

40.1%

8.2%

39.2%

9.2%

3.2%

40.8%

8.2%

39.0%

8.8%

0.6%

48.5%

5.4%

37.1%

8.4%

0.6%

48.8%

5.4%

37.1%

8.1%

1.0%

54.4%

1.8%

36.0%

6.8% 1.7%

54.1%

1.8%

35.6%

6.8%

Figure 4.2: Generation capacity mix for the different case studies (data source: Flex4RES. Data retrieved as of March 2019).

(21)

Table 4.3: Capacity factors and full load hours for onshore wind power (data source: DTU Wind Energy. Data retrieved as of November 2018).

Area Capacity factor Full load hours

2014 2030 2040 2050 2014 2030 2040 2050

SE1 0.25 0.31 0.36 0.40 2169 2747 3117 3478

SE2 0.22 0.29 0.33 0.38 1957 2525 2908 3296

SE3 0.26 0.33 0.37 0.41 2300 2870 3245 3621

SE4 0.33 0.40 0.45 0.49 2918 3522 3899 4256

FIn 0.26 0.31 0.35 0.39 2277 2751 3073 3393

FIs 0.25 0.32 0.36 0.41 2151 2777 3173 3583

NO1 0.31 0.38 0.42 0.46 2731 3341 3689 4018

NO2 0.28 0.36 0.41 0.45 2421 3126 3570 3973

NO3 0.29 0.33 0.36 0.38 2532 2898 3132 3367

NO4 0.34 0.39 0.42 0.45 2963 3389 3652 3906

NO5 0.21 0.25 0.28 0.30 1881 2216 2436 2661

DK1 0.24 0.31 0.36 0.41 2145 2732 3158 3576

DK2 0.24 0.30 0.36 0.41 2068 2667 3141 3589

EE 0.26 0.30 0.33 0.36 2289 2659 2899 3140

LV 0.26 0.32 0.36 0.40 2296 2822 3165 3514

LT 0.24 0.28 0.30 0.32 2095 2415 2624 2834

Table 4.4: Capacity factors and full load hours for offshore wind power (data source: DTU Wind Energy. Data retrieved as of November 2018).

Area/Country Capacity factor Full load hours

2014 2030 2040 2050 2014 2030 2040 2050

Sweden 0.34 0.45 0.46 0.49 2968 3941 4064 4322

FI 0.30 0.39 0.42 0.45 2649 3407 3678 3961

Norway 0.40 0.43 0.46 0.50 3478 3809 4037 4353

DK1 0.43 0.49 0.51 0.54 3789 4280 4428 4770

DK2 0.42 0.49 0.53 0.55 3682 4331 4611 4841

EE 0.29 0.35 0.42 0.45 2535 3044 3652 3920

LV 0.31 0.36 0.39 0.42 2681 3148 3434 3695

LT 0.36 0.42 0.45 0.47 3131 3638 3915 4145

4.3 Other Generation

As mentioned in Chapter2, the conventional generation is modeled by means of individual units, each of them being assigned an installed capacity, a Forced Outage Rate (FOR), a MTTF and a MTTR depending on its technology type. An outage time series is generated for all of the units, covering each simulated hour.

The generation capacity mix for each case study is shown in Figure 4.2 (where also VRE sources are considered). The generation units data set is based on the one used in [8] for 2020. For each case study, fictitious units are added (or, depending on the case, some units are removed) to match the Flex4RES case studies’ installed capacity for each technology type in each area.

(22)

Table 4.5: Per-area onshore wind power installed capacities for each case study (data source:

Flex4RES. Data retrieved as of March 2019).

Area Onshore wind power installed capacity [MW]

2020 Sc. 1

2030 Sc. 2

2030 Sc. 4

2040 Sc. 2

2040 Sc. 4

2050 Sc. 2

2050 Sc. 4

SE1 1989 2607 3458 3134 3134 2145 2145

SE2 3786 3743 4074 1779 2163 4196 4196

SE3 3019 7518 7518 6934 6934 9887 9887

SE4 1685 6087 6087 6087 6087 6087 6087

FI 3206 7102 7102 6496 6496 9860 10001

NO1 262 701 701 701 701 701 701

NO2 1102 2793 2793 2793 2793 2793 2793

NO3 1014 2314 2314 3826 3826 3608 3608

NO4 755 3466 3466 4857 4857 4857 4857

NO5 0 0 0 0 0 0 0

DK1 3272 4810 4810 4468 4468 5694 5893

DK2 654 1107 1107 1063 1063 1013 1013

EE 320 2366 2828 5087 4975 5992 6008

LV 60 6672 6280 6638 6840 7782 7782

LT 509 1669 1726 6397 6339 7122 7122

Table 4.6: Per-area offshore wind power installed capacities for each case study (data source:

Flex4RES. Data retrieved as of March 2019).

Area Offshore wind power installed capacity [MW]

2020 Sc. 1

2030 Sc. 2

2030 Sc. 4

2040 Sc. 2

2040 Sc. 4

2050 Sc. 2

2050 Sc. 4

SE1 0 0 0 0 0 0 0

SE2 0 0 0 0 0 0 0

SE3 0 0 0 0 0 0 0

SE4 0 0 0 500 500 500 500

FI 0 0 0 0 0 0 0

NO1 0 0 0 0 0 0 0

NO2 0 0 0 0 0 0 0

NO3 0 0 0 0 0 0 0

NO4 0 0 0 0 0 0 0

NO5 0 0 0 0 0 500 500

DK1 1257 1994 1994 10382 10425 9862 9821

DK2 444 1379 1379 1140 1140 1431 1431

EE 0 0 0 0 0 0 0

LV 0 0 0 0 0 0 0

LT 0 0 0 0 0 0 0

An exception are hydro power units, which are modelled in a very simplified way and assumed always available. Their available capacity is however reduced in the simulation because, historically, the hydro generation never reaches its full potential. This is the result of several factors such as maintenance, lower head in reservoirs, river system restrictions, environmental regulations, capacity set aside for frequency regulation reserves, etc. As an example, the maximum historical hydro power generation in Sweden is around 13.4 GW, even though the installed capacity of Swedish hydro power is approximately 16.3 GW.

Thus, the ratio 13.4/16.3 is used to scale the installed hydro power generation capacity in all of the

(23)

system areas. The resulting hydro power installed capacity for each area and each case study is shown in Table4.7.

Lastly, the nuclear, natural gas and coal installed capacities for each system area and for each case study are shown in Tables4.8,4.9and4.10, respectively.

Table 4.7: Per-area hydro power installed capacities for each case study. The Flex4RES data (retrieved as of March 2019) are here already scaled.

Area Hydro power installed capacity [MW]

2020 Sc. 1

2030 Sc. 2

2030 Sc. 4

2040 Sc. 2

2040 Sc. 4

2050 Sc. 2

2050 Sc. 4

SE1 4373 4385 4385 4385 4385 4385 4385

SE2 6677 6703 6703 6703 6703 6703 6703

SE3 2161 2170 2170 2170 2170 2170 2170

SE4 284 284 284 284 284 284 284

FI 2777 2844 2844 2844 2844 2844 2844

NO1 3005 3074 3074 3074 3074 3074 3074

NO2 9112 9251 9251 9251 9251 9251 9251

NO3 3750 3862 3862 3862 3862 3862 3862

NO4 4268 4376 4376 4376 4376 4376 4376

NO5 6486 6525 6525 6525 6525 6525 6525

DK1 6 0 0 0 0 0 0

DK2 0 0 0 0 0 0 0

EE 3 3 3 0 0 0 0

LV 2051 2051 2051 2051 2051 2051 2051

LT 1063 1063 1063 1063 1063 1063 1063

Table 4.8: Per-area nuclear power installed capacities for each case study (data source: Flex4RES.

Data retrieved as of March 2019).

Area Nuclear power installed capacity [MW]

2020 Sc. 1

2030 Sc. 2

2030 Sc. 4

2040 Sc. 2

2040 Sc. 4

2050 Sc. 2

2050 Sc. 4

SE1 0 0 0 0 0 0 0

SE2 0 0 0 0 0 0 0

SE3 7569 6691 6691 5291 5291 0 0

SE4 0 0 0 0 0 0 0

FI 4369 5062 5062 2800 2800 2800 2800

NO1 0 0 0 0 0 0 0

NO2 0 0 0 0 0 0 0

NO3 0 0 0 0 0 0 0

NO4 0 0 0 0 0 0 0

NO5 0 0 0 0 0 0 0

DK1 0 0 0 0 0 0 0

DK2 0 0 0 0 0 0 0

EE 0 0 0 0 0 0 0

LV 0 0 0 0 0 0 0

LT 0 0 0 0 0 0 0

Generation Adequacy in the Nordic and Baltic Region: Case Studies from 2020 to 2050: Flex4RES Project Report

Project Report