Determining an Optimal Level of Power System Investments Under Large Scale Penetration of Solar Power in Saudi Arabia

(1)

IN

DEGREE PROJECT ELECTRICAL ENGINEERING, SECOND CYCLE, 30 CREDITS

,

STOCKHOLM SWEDEN 2019

Determining an Optimal Level of

Power System Investments Under

Large Scale Penetration of Solar

Power in Saudi Arabia

HATEM ALATAWI

KTH ROYAL INSTITUTE OF TECHNOLOGY

(2)

(3)

KTH

M

ASTER

T

HESIS

Determining an Optimal Level of Power System

Investments Under Large Scale Penetration of

Solar Power in Saudi Arabia

Author: Hatem ALATAWI Supervisor: Bertrand RIOUX Lennart SÖDER Egill TÓMASSON Examiner: Mikael AMELIN

A thesis submitted in fulfillment of the requirements for the degree of Master’s degree in Electric Power Engineering

in the

School of Electrical Engineering and Computer Science (EECS) Department of Electric Power and Energy Systems (EPE)

(4)

(5)

iii

Abstract

The Saudi Arabia government set ambitious plans to adopt renewable energy sources in the energy mix as part of its strategy to diversify its economy away from oil. Ac-cording to the renewable energy project development office (REPDO), the total RES capacity installed would amount to 27.3 GW by 2023 and 58.7 GW by 2030. Due to the geographic location of Saudi Arabia, solar energy is a promising renewable energy source. However, there are many challenges to achieving a future where so-lar generation represents a significant portion of the Saudi generation mix. These concerns relate to the characteristics of solar PV (e.g., Variability, Aerosols, intermit-tency). As a result, measures should in place to take full advantage of the ambitious plans.

By modeling the generation of PV using real-time measurements, it is possible to quantify the potential energy produced (e.g., Power DC output from the PV panel). Also, through optimization techniques, it is possible to optimize future investments (e.g., capacity and transmission line investments) to minimize the costs while ensur-ing a reliable power system. In this thesis, the model accounts for the variability in the hourly solar production for an entire year, by investing in the required capacity to meet the hourly demand and the necessary PV-operational reserves for multiple interconnected regions. Also, the model optimizes the investments over a given fu-ture growth in demand (electricity consumption) while accounting for the current generation mix, fuel prices, and PV deployment in each region.

The model could be used to investigate multiple policies and their outcomes (e.g., fuel prices, PV regional deployment, PV capacity). In this thesis, two cases have been simulated based on the policymaker’s plans for adopting PV. The first case ex-amines 9.5 GW of installed PV capacity by 2023, while the second looks into 40 GW of installed PV capacity by 2030. The outcome is quantified by measuring the system costs and CO2emissions.

(6)

(7)

v

Sammanfattning

Saudiarabiens regering har fastställt ambitiösa planer för att integrera förnybara en-ergikällor i energimixen som en del av sin strategi för att diversifiera sin ekonomi bort från olja. Enligt ”Renewable energy project development office” (REPDO) kom-mer den totala kapaciteten för förnybar elproduktion uppgå till 27,3 GW år 2023 och 58,7 GW år 2030. På grund av Saudiarabiens geografiska läge är solenergi en lovande förnybar energikälla. Men det finns många utmaningar att uppnå en framtid där solenergi generationen representerar en betydande del av Saudiarabiens produk-tionsmix. En utmaning gäller solkraftens egenskaper (t.ex. variabilitet, aerosoler, in-termittens). Som ett resultat bör åtgärder vidtas för att dra full nytta av de ambitiösa planerna. Genom att modellera solkraftsproduktionen med hjälp av realtidsmät-ningar är det möjligt att kvantifiera den producerade potentiella energin (t.ex. Aktiv DC-effekt från panelerna). Genom optimeringstekniker är det också möjligt att opti-mera framtida investeringar (t.ex. gällande produktionskapacitet och transmission-sledningar) för att minimera kostnaderna samtidigt som man säkerställer ett tillför-litligt kraftsystem. I den här rapporten beaktar modellen variationen i solkrafts-produktionen per timme under ett helt år genom att investera i nödvändig kapacitet för att möta konsumtionen per timme och de nödvändiga drifts-reserverna för flera sammankopplade regioner. Modellen optimerar också investeringarna under en an-tagen framtida tillväxt i efterfrågan (elförbrukning) och beaktar samtidigt den nu-varande produktionsmixen, bränslepriser och solkraftens utveckling i varje region. Modellen kan användas för att undersöka flera möjliga framtida alternativ och deras resultat (t.ex. bränslepriser, solkraftens fördelning över landet, mängden solkraft to-talt). I denna rapport har två fall simulerats utifrån olika antagna planer för att öka mängden solkraft. Det första fallet undersöker 9,5 GW installerad PV-kapacitet år 2023, medan den andra undersöker 40 GW installerad PV-kapacitet år 2030. Resul-tatet kvantifieras genom att mäta systemets kostnader och koldioxidutsläpp.

(8)

(9)

vii

Acknowledgements

I would like to thank Bertrand Rioux, who is my friend and supervisor at KAPSARC for his continuous support and motivation.

I would like to thank Egill Tómasson my supervisor at KTH for his support and guidance throughout my thesis. It was a pleasure learning from you.

I would like to thank my examiner Lennart Söder for his guidance throughout my time at KTH. I still remember the first lecture at KTH where he shared his home electricity usage information from the installed solar panels. I was fascinated, and it set up my interests in the field for my studies at KTH.

(10)

(11)

ix

List of Figures

2.1 Saudi Electricity Grid (SEC). . . 6

2.2 Yearly fuel consumption of power and seawater desalination by type (“The Value of Integrating Saudi Arabia Into the Global Gas Market” 2019). . . 8

2.3 KAPSARC solar farm (Darwish, Abdulrahim, and Sharif, 2014). . . 8

2.4 Saudi ARAMCO solar farm (Aramco: Oil Giant Feels the Heat – Technol-ogy and Operations Management). . . 9

2.5 KAUST solar farm (KAUST boasts Saudi’s largest solar installation) . . . . 9

3.1 DNI, GHI and DHI for 24 hours measurements in Riyadh. . . 12

3.2 Graphical representation of the short-term forecast for solar Ibanez et al., 2012. . . 15

3.3 Power and clear-sky power output and ramps. . . 16

4.1 Delta PV Results. . . 20

4.4 Histogram of PV Operation Reserves. . . 25

4.5 Riyadh: Hourly Share of Capacity VS. Unavailability. . . 26

4.6 Riyadh: Hourly Share of Capacity VS. Unavailability. . . 27

6.1 Weekly Average Demand for All Regions in KSA. . . 40

6.2 ECRA Peck Demand Forecast . . . 40

6.3 Actual Peck Demand Forecast (ECRA). . . 41

6.4 PV Output & Reserve Requirements . . . 42

6.5 PV Capacity Location 2023 (Saudi Arabia 2030 Renewable Energy Targets). 43 6.6 PV generation displacement 2023. . . 46

(16)

xiv

6.7 PV generation displacement 2030. . . 47 6.8 Comparing scenario 1 and scenario 2. . . 48 6.9 Comparing scenario 3 and scenario 4. . . 48 6.10 Comparing Central Region Marginal Cost for scenario 1 and scenario 2. 49 6.11 Comparing Central Region Marginal Cost for scenario 3 and scenario 4. 50 6.12 Comparing Eastern Region Marginal Cost for scenario 3 and scenario 4. 50

(17)

xv

List of Tables

2.1 Electricity tariff in 2015 . . . 7

2.2 Electricity tariff in 2017 . . . 7

3.1 Variable and units. . . 13

3.2 ∆T assumptions for different Modules. . . 14

4.1 Meteorological station location and parameters. . . 18

4.2 Solar radiation monitoring station location and parameters. . . 18

4.3 Seasonal Capacity Factor Results. . . 23

4.4 Seasonal Energy Yield Results (All values are in kWh/kW p). . . 24

5.1 Summary of Model Input Data. . . 31

5.2 Optimization sets. . . 32

5.3 Optimization Parameters. . . 33

5.4 Optimization Variables. . . 33

6.1 Generation capacity by region, fuel and technology. . . 38

6.2 Technology Parameters . . . 39

6.3 Regional Choice of City Solar Profile. . . 42

6.4 Approximated PV Capacity allocation in 2023. . . 43

6.5 Assumptions on PV Capacity allocation in 2030. . . 44

6.6 Ramping costs for power plant technologies (Bergh and Delarue, 2015). 45 6.7 System Cost Under S1 & S2. . . 45

6.8 System Cost Under S3 & S4. . . 46

6.9 MMBTU to lb CO2(U.S. Energy Information Administration - EIA - In-dependent Statistics and Analysis). . . 47

(18)

(19)

xvii

List of Abbreviations

REPDO Renewable Energy Project Development Office

PV Photo Voltaic

NREL National Renewable Energy Laboratory

POA Plane Of Array

LP Linear Programming

GAMS General Algebraic Modeling System

SCECOs Saudi Consolidated Electricity Companies

SEC Saudi Electricity Company

KAPSARC King Abdullah Petroleum Studies And Research Center

ARAMCO ARabian AMerican Oil Company

KAUST King Abdullah University Of Science And Technology

DNI Direct Normal Irradiance

DHI Direct Horizontal Irradiance

GHI Global Horizontal Irradiance

SAPM Sandia PV Array Performance Model

LCOE Levelized Cost Of Electricity

NREL National Renewable Energy Laboratory

SPI Solar Power Index

NCDC National Climatic Data Center

NASA National Aeronautics And Space Administration

KACST King Abdulaziz City for Science and Technology

MILP Mixed Integer Linear Programming

(20)

(21)

1

Chapter 1

Introduction

1.1 Background

The Saudi Arabian government set ambitious plans to adopt renewable energy sources in the energy mix. According to the renewable energy project development office (REPDO), the total RES capacity would amount to 27.3 GW by 2023 and 58.7 GW by 2030 (Saudi Arabia 2030 Renewable Energy Targets). The goal of adopting renewable energy sources is part of the Saudi Kingdom’s plans to diversify its economy away from oil. Thus, the adoption of renewable sources is of most valuable. Jobs will be created for a nation where youth are 20-30 years old are 70% of society (Alkhalisi, 2017), and it will reduce the footprint of CO2emissions in the power system. Lastly, it will free up diesel and other fossil fuels, which in return would provide a net profit.

Due to the geographic location of Saudi Arabia, solar energy is a promising re-newable energy source. However, there are many challenges to achieving a future where solar generation represents a significant portion of the Saudi generation mix. Since deserts cover most of the land, the variability and uncertainty in the solar generation for large solar farms will be a great challenge. Which may increase the intermittency cost (Elshurafa and Matar, 2017). Thus, quantifying the additional costs and benefits will, in return, provide a clearer picture to policymakers on the implications of large Photovoltaics (PV) investments.

(22)

2 Chapter 1. Introduction

1.2 Objective

The proposed thesis aims to, first, quantify the variability and uncertainty in solar generation. Second, to study PV implications on power system reliability, by find-ing out the optimal level of investment which minimizes the cost of operation and investment while ensuring reliability. The thesis will look into multiple solutions to address reliability such as investments in transmission lines between areas, in-creasing the generation reserves, and installing new capacity. The study will also improve the representation of renewable investment in future models by quantify-ing the variability and uncertainty in solar production for multiple cities in Saudi Arabia.

1.3 Approach for the studies

1.3.1 PV model

The goal of the PV model is to generate DC power output for different cities in Saudi Arabia, based upon the site meteorological and solar measurements. The model will receive cite location parameters corresponding to each city (e.g., latitude, longitude, tilt angle). The PV model follows NREL’s PVWatts work, which takes into perspec-tive the thermal losses, and plane-of-array (POA) irradiance. These parameters ac-count for the sky and ground diffuse irradiance attributes. The model is built using the open-sourced pvlib in python.

1.3.2 Power System Optimization

For this thesis, the model should account for the variability and uncertainty of hourly solar production for an entire year, by investing in the required capacity to meet the hourly demand and the required PV-operational reserves for multiple intercon-nected regions in a future scenario (e.g., 2023). Also, the model should optimize the investments over a given future growth in demand (electricity consumption). While accounting for the current generation mix, fuel prices, and PV deployment in each region. Since the model should account for the constraints on the available data and solution time, an LP model would be favorable. Thus, the choice was an optimal

(23)

1.4. Thesis Structure 3

economic dispatch model with investment decisions. The model would equalize generation and demand in each hour at each region, taking into account the vari-ability and uncertainty of solar production. The model is built using GAMS.

1.4 Thesis Structure

This thesis is structured as follows. Chapter 2 provides a brief overview of the Saudi power system, going over the history, current state, and renewable plans. In chapter 3, we describe the PV model and its characteristics. In chapter 4, we present and discuss the results from the PV model. In chapter 5, we go over the optimization function for the power system model. In chapter 6, we present and discuss the re-sults from the Saudi power system model, going over the investments made, and the implications on the power system cost. Chapter 7 summarizes the thesis.

(24)

(25)

5

Chapter 2

Saudi Arabia Electricity Market

Review

2.1 History of the Saudi Electricity Sector

The Saudi electricity industry went through many transformations since the estab-lishment of The Kingdom of Saudi Arabia in 1932 and the discovery of oil in 1938. First, before 1961, electricity was provided on a commercial basis. Power was sup-plied to cities and small villages through private investments. As a result, the elec-tricity prices varied for the area and the provider.

The first entity to regulate the electricity in Saudi Arabia was the Electricity Af-fairs Administration established in 1961 by the Ministry of Commerce aimed at set-ting laws and regulations to govern the operations and ownership. Later in 1974, the Department of Electricity Services was formed under the Ministry of Industry and Electricity Services. The department of electricity services passed legislation which set a uniform electricity tariff. The tariff was heavily subsidized and was set below the cost of production. Hence, double-digit growth rates in the Saudi economy were recorded in the following years as a result of the legislation.

In 1976, as part of the kingdom’s five years investment plan, the Electricity Cor-poration was formed amid at coordinating and implementing the electricity invest-ments. The electricity cooperation merged electricity providers into four regional companies named the Saudi Consolidated Electricity Companies (SCECOs). They were located in the south, west, center, and east regions of Saudi Arabia. See figure 2.1.

(26)

6 Chapter 2. Saudi Arabia Electricity Market Review

FIGURE2.1: Saudi Electricity Grid (SEC).

As a result, the Saudi government achieved high electrification rates of 90.9% by 2000. However, the Saudi grid was poorly interconnected as a result of the four regions’ independent company setup. Thus, in 2000, the Saudi government made a decision to merge the four regional companies into one vertically integrated single joint stock named the Saudi Electricity Company (SEC).

The private electricity sector role has been growing as a result of introducing the Power Purchase Agreements (PPA). These agreements are between the Independent Power Producers (IPP’s) and SEC. There are also generation rental agreements pro-vided by SEC to small energy companies, where the enterprises provide electricity to villages and towns.

2.2 Current state

During the period leading to 2015 the electrification rate reached 100% (Trading Eco-nomics). Nowadays there are 9,049,712 customers as of 2015. The residential electric-ity tariffs are calculated based on three categories, see table 2.1.

(27)

2.3. Renewable Energy Sources Investments 7

Consumption Category (kWh) Price ($/kWh)

1-2000 0.013

2000-4000 0.027

4000-6000 0.053

above 6000 0.080

TABLE2.1: Electricity tariff in 2015

In 2016, energy price reforms were applied, resulting in a change in the electricity tariff, see table 2.2. The increase in electricity tariff was aimed at the inefficient usage of electricity, which was a result of the low electricity tariff.

Consumption Category (kWh) Price ($/kWh)

1-6000 0.048

above 6000 0.080

TABLE2.2: Electricity tariff in 2017

Natural gas is used as the primary power generation fuel for electricity in Saudi Arabia. The Kingdom set a plan to double its natural gas production to 238 bcm by 2030. The increase in demand for gas by power generation has been outstripping the supply in the last couple of years; as a result, Saudi Arabia is burning oil for elec-tricity. The price reforms mentioned above were aimed at rationalizing the energy consumption and freeing-up oil barrels for export (“The Value of Integrating Saudi Arabia Into the Global Gas Market” 2019).

In 2017, the Kingdom consumed an average of almost 900 thousand barrels per day Kbbl/d to meet power and water demand. They account for 46% of Saudi Ara-bia’s fuel mix. The liquids are broken down into diesel, heavy fuel oil (HFO), and crude oil, see figure 2.2.

2.3 Renewable Energy Sources Investments

The Saudi Arabian government set ambitious plans to adopt renewable energy sources in the energy mix. According to the renewable energy project development office (REPDO), the total RES capacity will amount to 27.3 GW by 2023 and 58.7 GW by 2030 (Saudi Arabia 2030 Renewable Energy Targets). This section mentions, established and future renewable energy projects.

(28)

8 Chapter 2. Saudi Arabia Electricity Market Review

FIGURE 2.2: Yearly fuel consumption of power and seawater de-salination by type (“The Value of Integrating Saudi Arabia Into the

Global Gas Market” 2019).

2.3.1 Established Renewable Energy Projects

King Abdullah Petroleum Studies and Research Center (KAPSARC)

The first solar project to be built in Saudi Arabia on a mega level was KAPSARC Solar Park (figure 2.3), located in Riyadh, the capital of Saudi Arabia. The park capacity is 5 MW, and it is estimated to offset 4900 tons of CO2 annually (“Study of NEOM city renewable energy mix and balance problem”).

FIGURE 2.3: KAPSARC solar farm (Darwish, Abdulrahim, and Sharif, 2014).

(29)

2.3. Renewable Energy Sources Investments 9

Saudi Arabian American Oil Company (Saudi ARAMCO)

The Saudi ARAMCO solar farm has a capacity of 10 MW; it is installed on a parking lot, see figure 2.4. The farm is located in the eastern region, near the city of Dahran.

FIGURE2.4: Saudi ARAMCO solar farm (Aramco: Oil Giant Feels the Heat – Technology and Operations Management).

King Abdullah University of Science and Technology (KAUST)

KAUST solar farm has a capacity of 2 MW; it is installed on a building rooftop, figure 2.5. The farm is located in the western region, near the city of Thuwal.

FIGURE2.5: KAUST solar farm (KAUST boasts Saudi’s largest solar in-stallation)

(30)

(31)

11

Chapter 3

Photovoltaic DC model

This chapter aims to describe the modeling approach for the PV DC electricity gen-eration. The goal of the model is to generate PV DC output using multiple mea-surement inputs from various cities in Saudi Arabia. Furthermore, by the end of the chapter, the forecast error of PV is calculated from the real-time measurements and optimal modeled values (clear-sky model).

3.1 Input Data

In the following section, the input data to the model will be described. The section will highlight modeled and measured data.

3.1.1 Meteorological data

The model will receive hourly wind and temperature measurements.

3.1.2 Solar data

The model will receive hourly Direct Normal Irradiance (DNI), Diffuse Horizon-tal Irradiance (DHI), Global HorizonHorizon-tal Irradiance (GHI). The clear sky DC output is modeled using optimal solar synthetic variables, which represents the optimal condition for a PV panel. The DC output is modeled using site measurements to simulate an output of a panel under real-time conditions (e.g., Aerosols). Figure 3.1 shows an example of the measurements for Riyadh city during a 24 hour period.

(32)

12 Chapter 3. Photovoltaic DC model

FIGURE 3.1: DNI, GHI and DHI for 24 hours measurements in Riyadh.

3.2 PV DC output

3.2.1 Model Choice

For the purpose of the study, the PV model should account for multiple site mea-surements, which affect the generation of PV arrays. The variables include aerosols and thermal losses, which are results of mainly dust, cloud, and heat. Many models simulate PV output, such as the Sandia PV Array Performance Model (SAPM) by the Sandia National Laboratories. The SAPM model takes into account PV manu-facturing parameters and chemistry materials and many more detailed inputs that are used in calculating the Levelized Cost Of Electricity (LCOE). However, this is beyond the scope of the thesis. The aim of the model is to capture the variability and uncertainty in solar generation due to mainly two effects: heat and aerosols. Hence, inverter losses, AC\DC cables losses, and soiling losses are neglected in the thesis. The National Renewable Energy Laboratory (NREL) PVwatts thermal model satisfies the goals laid before, which takes into account the thermal losses, and plane-of-array (POA) irradiance. Hence, it will account for the sky and ground diffuse ir-radiance attributes. The model is built using the open-sourced pvlib in Python (pvlib python 2018).

(33)

3.2. PV DC output 13

3.2.2 PVWatts DC Module Model

The PVWatts DC model will be described below to understand how the calculations are made, what parameters are used and how they affect the output. The input parameters will be highlighted and the layout of the model will be analyzed. It should be noted that the PVWatts model is well documented and is widely used. Thus, for a more detailed analysis, please refer to (“PVWatts Version 5 Manual”).

The DC output of the model is calculated through equation 3.1, where the heat transfer is a function of the reference temperature and the plane of array irradiance. The panel efficiency is assumed to decline at a linear rate with respect to tempera-ture, affected by the temperature coefficient γ. Table 3.1 summarizes the values and units for equation 3.1.

Pdc

=

IPOA

I0

P_dc0

(

1

+

γ

(

Tcell

−

Tre f

))

(3.1)

Variable Description Units

Pdc DC power W

IPOA Plane of array irradiance W/m2 I0 Reference irradiance 1000 W/m2 Pdc0 Nameplate DC rating W

γ Temperature coefficient 1/C

Ta Ambient Air Temperature C Tcell Cell Temperature C Tre f Reference Temperature 25 C Tm Back-surface Temperature C

W Wind Speed m/s

TABLE3.1: Variable and units.

The model takes into account the back-surface temperature through equation 3.2, where the back-surface temperature is a function of the ambient air tempera-ture, wind speed and the empirically determined coefficients (a, b) based the model assumption (see table 3.2).

Tm

=

IPOAea+b W

+

Ta (3.2)

(34)

they are related to each other through equation 3.3. The equation assumes a one-dimensional thermal heat conduction, which allows the use of a pre-calculated tem-perature difference between the back-surface temtem-perature and cell temtem-perature for a given material. The value∆T depends on the model assumption. Table 3.2 shows the associated models and there values for∆T.

Tcell

=

Tm

+

IPOA

I0 ∆T (3.3)

Module Type a b ∆T

Glass/cell/glass (Open rack) -3.47 -0.0594 3 Glass/cell/glass (Close roof mount) -2.98 -0.0471 1 Glass/cell/polymer sheet (Insulated back) -2.81 -0.0455 0

TABLE3.2:∆T assumptions for different Modules.

Using the Perez algorithm. The plane-of-array component beam Ib, sky diffuse Id,s, and ground diffuse Id,gare calculated from the Direct Normal Irradiance (DNI), Diffuse Horizontal Irradiance (DHI), and the Global Horizontal Irradiance (GHI) (Perez et al., 1990). Equation 3.4 shows the summation of the latter components which results in the total plane of array IPOAincident on the module.

IPOA

=

Ib

+

Id,s

+

Id,g (3.4)

3.3 PV Operational Reserves

The PV operational reserves are defined as the idle spinning generation units which react within minutes to the variability in renewable generation. In this section, we use the forecast error to calculate the variability in PV production.

The forecast error is calculated by leveraging the PV model with clear sky input values. This section describes the process in more detail. The methodology of the approach is based on NREL’S work (Ibanez et al., 2012).

(35)

3.3. PV Operational Reserves 15

3.3.1 Operational Reserves Calculation

First, the solar power index (SPI) needs to be calculated in order to perform the short-term forecast. The solar power index is defined as Prealthe actual solar power, divided by Pcs, clear sky power. This results in a value between 1 and zero, as shown in equation 3.5

SPI

=

PReal

PCS

(3.5)

With the above SPI, the short-term forecast PFwill be defined for a time step∆t, which is the power output at time t plus the predictable clear-sky power output. as shown in equation 3.6, presented graphically in figure 3.2.

PF

(

t

)

(3.7)

FIGURE3.3: Power and clear-sky power output and ramps.

Figure 3.3 highlights the clear sky ramp and its relation with power output, pos-itive in the morning, negative at the evening. The figure shows the magnitude of the ramp requirements on a normalized value.

(37)

17

Chapter 4

Resource Assessment: Quantifying

Solar Variability

This chapter aims to address the variability of solar generation in multiple cities within Saudi Arabia by analyzing the outputs derived from the model described in chapter 3. The data used to leverage the PV model will be presented in this chapter, followed by the results and discussion.

4.1 Data

This section will describe the data used to calibrate the PV model. The data is cate-gorized as meteorological and solar irradiance.

4.1.1 Meteorological Data

Meteorological data are obtained from the National Climatic Data Center (NCDC), through an open-source online platform (National Centers for Environmental Informa-tion). The data are hourly metric recordings of more than 48 stations, located in multiple cities within Saudi Arabia. They include wind, visibility, temperature, dust observation, among others. The data mainly spans the years 1980 till 2019. For the thesis, 10 locations have been chosen based on solar data availability (e.g., a station has been selected in a city where the solar station is located). Table 4.1 shows the stations chosen with the respective longitude and latitude and metric used.

(38)

18 Chapter 4. Resource Assessment: Quantifying Solar Variability

Location Latitude Longitude Elevation (m) Metric Riyadh 24.96 46.70 624.53 Temp,Wind Abha 18.24 42.66 2090.31 Temp,Wind Al-ahsa 25.29 49.49 179.22 Temp,Wind Gizan 16.90 42.59 6.09 Temp,Wind Qassim 26.30 43.77 648.00 Temp,Wind Jeddah 21.68 39.16 14.63 Temp,Wind Al-qaisumah 28.34 46.13 357.83 Temp,Wind Sharurah 17.47 47.12 720.24 Temp,Wind Al-Jouf 29.79 40.10 689.15 Temp,Wind Wadi Al-Dawaser 20.50 45.20 628.49 Temp,Wind

TABLE4.1: Meteorological station location and parameters.

4.1.2 Solar Data

The solar data used are from the National Aeronautics and Space Administration (NASA) remote sensing validation data. The joint project was done by King Abdu-laziz City for Science and Technology (KACST) and NREL. They operated ten solar radiation monitoring networks in multiple cities across Saudi Arabia from 1998 to 2003. Table 4.2 shows the location of each monitoring station, alongside the respec-tive longitude and latitude and metric used.

Location Latitude (◦) Longitude (◦) Elevation (m) Metric

Riyadh 24.91 46.41 650 DNI,DHI,GHI,temp Abha 18.23 42.66 2039 DNI,DHI,GHI,temp Al-ahsa 25.3 49.48 178 DNI,DHI,GHI,temp Gizan 16.9 42.58 7 DNI,DHI,GHI,temp Qassim 26.31 43.77 647 DNI,DHI,GHI,temp Jeddah 21.68 39.15 4 DNI,DHI,GHI,temp Al-qaisumah 28.32 46.13 358 DNI,DHI,GHI,temp Sharurah 17.47 47.11 725 DNI,DHI,GHI,temp Al-Jouf 29.79 40.1 669 DNI,DHI,GHI,temp Wadi Al-Dawaser 20.44 44.68 701 DNI,DHI,GHI,temp

TABLE4.2: Solar radiation monitoring station location and parame-ters.

4.2 Model Configuration

First, the data above is used to generate the results; the following assumptions have been made:

(39)

4.3. Results 19

• The PV tilt angle is assumed to be equal to the location latitude.

• The temperature and wind speed of similar cities from two different data sets are used.

• The model does not capture any dust deposition left on the PV panel.

4.3 Results

The following section highlights the results of the PV model. First, the section exam-ines the results from the delta PV index by analyzing the impact of cloud cover and aerosols on the seasonal-region variability. Then an analysis is provided on quan-tifying the potential of solar energy in each city using two metrics, capacity factor, and yield energy. Followed by evaluating the operation reserves using forecast er-ror. Lastly, we examine the unavailability of PV generation due to meteorological factors and the potential contribution of PV to the firm capacity.

4.3.1 Results and Analysis: Delta PV

The delta PV is the difference between the optimal solar generation given a clear sky and the modeled results using observed irradiance measurements. The index measures the loss of power generation as a result of the non-clear sky condition (e.g., dust, clouds, aerosols).

The results highlight five years worth of hourly data broken down into seasons to examine the effect of seasonality on solar generation. The results are presented as histograms spanning five bins, each bin representing increments of 20%. The Y-axis shows the number of hours for the given duration.

The results of the model which corresponds to the city of Riyadh in the central region, figure 4.1a, show that during the summer season Riyadh generates PV near the optimal benchmark when compared to other seasons. This indicates that during the summer, there are fewer aerosols in the air. When looking at the second bin (20%-40%), it is observable that PV delta is higher in fall, spring, and winter compared to summer. Hence, aerosols are occurring more frequently in these seasons.

(40)

The city of Qassim, located in the central region, figure 4.1b, shows similar char-acteristics to Riyadh, summer is the optimal season for PV generation. However, the drop in delta PV in the bin (20%-40%) for all seasons is less compared to Riyadh.

(A) Riyadh Delta PV Histogram. (B) Qassim Delta PV Histogram.

(C) WadiAldawaser Delta PV Histogram. (D) Alqaisumah Delta PV Histogram. FIGURE4.1: Delta PV Results.

The city of Wadi Aldawaser, located in the central region, figure 4.1c, shows out-standing results compared to other cities in Saudi Arabia. The results show that most of the hours occupy the first bin compared to the other bins; this means that the location observes the least aerosols compared to other cities.

The city of Alqaisumah, located in the eastern region, figure 4.1d, shows simi-lar characteristics to cities in the central region, fewer aerosols during the summer compared to other seasons. However, there more observed aerosols in total. This is indicated by the increase in hours located in the bin (40%-60%). Thus resulting in a more significant drop from the optimal clear sky benchmark.

The city of Alahsa, located in the eastern region, figure 4.2a, show similar char-acteristics to Alqaisumah in the eastern region, fewer aerosols during the summer compared to other seasons, and similar aerosols increase in the third bin (40%-60%)

(41)

4.3. Results 21

resulting in a higher drop from the optimal clear sky benchmark.

(A) Alahsa Delta PV Histogram. (B) Gizan Delta PV Histogram.

(C) Sharurah Delta PV Histogram. (D) Abha Delta PV Histogram. FIGURE4.2: Delta PV Results.

The city of Gizan, located in the southern region, figure 4.2b, shows different characteristics compared to other cities, fewer aerosols during spring when compar-ing it to other seasons. However, there are more hours with higher aerosols com-pared to cities. This is observed through the increase in the second bin (20%-40%).

The city of Sharurah, located in the southern region, figure 4.2c, show similar characteristics to other cities in Saudi Arabia, fewer aerosols during the summer compared to other seasons, and similar aerosols increase in the third bin (40%-60%) resulting in a higher drop from the optimal clear sky benchmark. There is also a slight increase in the number of hours in the third bin (40%-60%) during the fall season.

The city of Abha, located in the southern region, figure 4.2d, shows similar char-acteristics compared to the city of Gizan, fewer aerosols during spring when com-paring it to other seasons. However, there are more hours with higher aerosols compared to Gizan. This is observed through the increase in the third, fourth bin

(42)

(40%-60%) and (60%-80%) respectively.

(A) Aljouf Delta PV Histogram. (B) Jeddah Delta PV Histogram. FIGURE4.3: Delta PV Results.

The city of Aljouf, located in the northern region, figure 4.3a, show similar char-acteristics to other cities in Saudi Arabia, fewer aerosols during the summer com-pared to other seasons, and similar aerosols increase in the third bin (40%-60%) re-sulting in a higher drop from the optimal clear sky benchmark across all seasons.

The city of Jeddah, located in the western region, figure 4.3b, shows different characteristics compared to other cities in Saudi Arabia, fewer aerosols during spring when comparing it to other seasons. However, there are more hours with higher aerosols compared to any other city. This is observed through the increase in the third, fourth bin (40%-60%) and (60%-80%) respectively. The results show that dur-ing winter there are more hours with higher aerosols compared to other seasons, followed by the fall season.

4.3.2 Results and Analysis: Capacity Factor & Yield Energy

The following section highlights the results of the PV model regarding electricity generation. The aim is to evaluate the seasonality of PV generation, specifically to look at the impact of temperature on PV generation. Since different metrics are used in power planning for PV generation in Sadui Arabia, the section presents the results using the energy yield and capacity factor indices.

(43)

4.3. Results 23

Capacity Factor

The capacity factor is the ratio of energy generated divided by the hourly capacity generation. Usually it is measured over a single year, this is shown through equation 4.1. The unit for the later value is in percentages.

CF

=

Electricity Generated in a Year MWh

(

365 days

) × (

24 hours_days

) × (

peak MW

)

(4.1)

The percentage drop between the capacity factor measured at the clear sky and the actual measurements will be referred to as the delta capacity factor∆CF, shown in equation 4.2.The unit for the later value is in %.

∆CF

=

CFCS

−

CFReal

CFCS (4.2)

The results are presented in table 4.3. In general, the values are very similar with regard to the yearly CF. The city of Aljouf, located in northern Saudi Arabia, has the highest capacity factor value. Wadi Aldawaser, and Riyadh follow it. In Abha, Aljouf and Wadi Aldawaser the electricity generated in winter is higher than summer. This may be a result of thermal losses due to heat. The delta capacity factor indicates the percentage drop (losses) from the optimal clear sky as a result of aerosols and cloud coverage. The most significant drop is in Gizan, followed by Alqaisumah and Abha. The least losses are in Wadi aldawaser and Aljouf.

Location CF WI (%) CF SP (%) CF SU(%) CF FA(%) CF Year(%) ∆CF(%)

Abha 6.12 5.99 5.18 5.92 23.21 16.03 Alahsa 5.06 5.98 5.72 5.26 22.02 15.39 Alqaisumah 5.17 5.80 5.35 4.80 21.12 17.42 Gizan 5.18 5.73 5.03 5.16 21.09 18.99 Jeddah 5.51 6.75 5.74 5.09 23.09 12.97 Qassim 5.44 6.01 5.73 5.31 22.49 14.61 Riyadh 5.75 6.11 6.04 5.76 23.67 11.05 Sharurah 5.79 5.89 5.50 5.51 22.69 13.37 WadiAldawaser 6.17 5.84 5.69 5.97 23.68 9.30 Aljouf 5.89 6.50 6.16 5.35 23.91 10.89

(44)

Energy Yield

The energy yield is the ratio of annual energy production divided by the installed ca-pacity, this is shown through equation 4.3. The unit for the later value is in kWh/kW p.

EY

=

annual energy production kWh

Peak Capacity kW p (4.3)

The percentage drop between the energy yield measured at the clear sky and the actual measurements will be referred to as the delta energy yield ∆EY, shown in equation 4.4.

∆EY

=

EYCS

−

EYReal

EYCS

(4.4)

The results are similar to the capacity factor results above. The purpose is to present different metrics so that the audience would choose the preferable index to observe the PV energy generation.

Location EY WI EY SP EY SU EY FA EY Year ∆EY Abha 535.78 524.78 453.87 518.42 2032.85 16.03 Alahsa 442.88 523.69 501.37 461.20 1929.14 15.39 Alqaisumah 452.67 508.52 468.92 420.30 1850.40 17.42 Gizan 453.70 501.59 440.31 451.92 1847.52 18.99 Jeddah 482.40 591.41 502.96 445.54 2022.31 12.97 Qassim 476.54 526.17 502.10 465.15 1969.97 14.61 Riyadh 503.96 535.64 529.37 504.41 2073.38 11.05 Sharurah 507.39 515.60 481.81 482.75 1987.55 13.37 WadiAldawaser 540.74 512.02 498.44 522.81 2074.01 9.30 Aljouf 515.70 569.58 540.04 468.87 2094.19 10.89

TABLE 4.4: Seasonal Energy Yield Results (All values are in kWh/kW p).

4.3.3 Results and Analysis: PV Operation Reserves

The PV operational reserves as discussed earlier in section 3.3 is measured through the forecast error. This section will provide a brief analysis on the PV operational reserves for the city of Riyadh.

A more thorough analysis will be presented in the power system results section, where the cost of the reserves is taking into account.

(45)

4.3. Results 25

FIGURE4.4: Histogram of PV Operation Reserves.

The results in this section will focus on the capacity needed for the operational reserves as well as the duration (hours expected for a PV operation reserve).

Figure 4.4 , presents the five-year histogram distribution of the hourly opera-tional reserves as a percentage of the installed PV capacity. For example, there are 46 hours where the required operational reserves are between (35%-40%) of the in-stalled PV capacity.

4.3.4 Results and Analysis: PV Generation Availability

The following section addresses the availability of PV generation, taking the heat and aerosols factors into account. The aim is to measure the unavailability assuming different shares of firm capacity level commitments from PV.

There are two methodologies used to calculate the results. In the first analysis, we measure the share of capacity from the average generation for a span of 5 years (e.g., Finding the average generation at 08:00 am for a span of 5 years, then the share of capacity will be calculated based on the latter average.) For the second analysis, we measure the share of capacity from the maximum generation for five years (e.g., Finding the maximum generation at 08:00 am for five years, then the share of capacity will be calculated based on the latter value.)

(46)

The Average Metric

Using the average generation of an hour as an index to calculate the share of capacity and unavailability in Riyadh, results in figure 4.5. Unavailability means, in this case, the percentage of hours that are below the capacity share (e.g., for a 10% share of ca-pacity in hour 05:00 pm, there is a probability of 23.6% where the generation might be below the 10% share of capacity). There is high uncertainty with morning and evening hours (e.g., 6 am, 7 am, 4 pm, and 5 pm) whereas we increase the capacity factor, the number of hours below the index capacity increases proportionally. Also, we would observe that for morning and evening 30% of the time on average the gen-eration would be below the desired 50% capacity factor of the average gengen-eration.

FIGURE4.5: Riyadh: Hourly Share of Capacity VS. Unavailability.

The Maximum Metric

Using the maximum generation of an hour as an index to calculate the share of ca-pacity and unavailability in Riyadh, results in figure 4.6. The results are similar to the average metric case, where there is high uncertainty with morning and evening hours. However, the unavailability is percentage is higher since the share is mea-sured for the peak generation of the hour for five years.

(47)

4.3. Results 27

(48)

(49)

29

Chapter 5

Modeling The Saudi Power System

This chapter aims to describe the modeling approach for the Saudi Power System. The goal of the model is to minimize the investments given policy decisions towards PV capacity deployment. The model takes into account parameters and constraints for each region, such as generation capacity, fuel, load, and transmission lines.

5.1 Model Choice

For the study, the model does account for the variability and uncertainty of hourly solar production for an entire year, by investing in the required capacity to meet the hourly demand and the required PV-operational reserves for multiple intercon-nected regions in a future scenario (e.g., 2023). Also, the model optimizes the in-vestments over a given future growth in demand. While accounting for the current generation mix, fuel prices, and PV deployment in each region.

There are multiple model methods to approach the question. One favored model would be a security-constrained unit commitment model. However, this choice re-quires a plant by plant modeling and sub-regional representation of demand and generation. Furthermore, the solving time of MILP compared to other approaches is much longer.

Since the model should account for the constraints on the available data and solution time, an LP model is favorable. Thus, the choice was an optimal economic dispatch model with investment decisions. The model matches supply and demand in each hour in each region, taking into account the variability and uncertainty of solar production.

(50)

30 Chapter 5. Modeling The Saudi Power System

• Start-up and shut-down costs.

• Minimum up and downtime of power plants. • Frequency reserves.

• Minimum uptime.

The model follows the work done by (Söder and Amelin, 2011); it is built using the General Algebraic Modeling System (GAMS), which uses the algebraic modeling language (AML) (Hawthorn, Weber, and Scholten, 2001).

5.2 Input data

In the following subsections, the input data to the model is described. The section highlights modeled and measured data. Such as hourly PV generation, PV invest-ment capacity, forecast error of solar generation, demand, the capacity of thermal units, heat-rates, and prices. Table 5.1 at the end of the section summarizes the input data.

5.2.1 Load data

The model receives the hourly load in MWh for each region. Hence, 8760 hours for each region.

5.2.2 Thermal units data

The model accounts for four types of thermal units: Gas turbines, combined cycle, steam turbines, and nuclear. The model receives the following input data:

• Thermal capacity of each plant for each region in MW.

• Fixed operation and maintenance cost of each plant type in $/MWh. • Heat rate of each plant type in MMBtu/MWh.

• Fuel price of each plant type in $/MMBtu. • Annual Fixed costs of each plant type in $/MW.

(51)

5.3. Sets, Parameters & Variables 31

• Annualized Investment cost of each plant type in $/MW.

The input data would determine the cost of generation, merit order, and invest-ment decisions in type of technologies.

5.2.3 PV data

The model receives an hourly normalized PV generation in MWh with a capacity of 1 MW. The capacity of PV in each region could be scaled to match policy investment decisions. Also, another input parameter would be the PV normalized generation forecast error in MWh from the hourly solar generation. Both of the two previous values are being fed from the model in chapter 3.

5.2.4 Summary

Data Description Units

Demand Hourly load in each region MWh

Thermal Units

Thermal capacity by plant type and region MW Variable cost for each plant type $/MWh Heat rate of each plant type MMBtu/MWh Fuel price of each plant type $/MMBtu Annual Fixed costs of each plant type $/MW Annualized Investment cost of each plant type $/MW PV

Desired PV capacity in each region MW Nomalized PV generation P.u. PV operation reserves in each hour for each region MWh

TABLE5.1: Summary of Model Input Data.

5.3 Sets, Parameters & Variables

This section presents the sets, parameters, and variables used in the optimization model. They include modeled and measured values.

5.3.1 Sets

(52)

Sets Description

t Time index in hours 1,...,8760. n, m Regional area 1,...,4.

g Power plants by type 1,...,4.

N Set of areas n.

G Set of thermal power plants g.

G_n Set of thermal power plants g located in region n.

P Set of interconnections (n,m).

P_n←m Set of regions m capable of exporting to regions n.

Pn→m Set of regions m capable of importing from regions n.

TABLE5.2: Optimization sets.

5.3.2 Parameters

In this section, the parameters of the optimization model will be described. The following parameters are input values to the optimization problem; can be found in section 5.2. On the thermal cost side of the parameters: FCg represents the fixed operating costs for plant type g in MW, where VCg represents the Variable cost for plant type g in MWh. In addition to the fixed and the operation costs, heat rate Hg and fuel price Pgare considered as well. The values should be reported in MMBtu in order to compare the fuels and their impact on the system cost.

In the demand side, Dn,t represents the hourly consumption of each region n in MWh. For the PV parameters, PVn,twill be the hourly normalized generation in each region n per MWh. The normalized PV operational reserves in MWh for each hour and region is represented by OEn,t, PVE is the cost of the PV operational reserves is $/MWh.

In the investment decision side: ICGg is the capacity cost of plant type g in $/MW, annualized. Also, ICLP will be the annualized capacity cost of increasing

the transmission line capacity by 1 MW. For the PV investment, IPVnis the amount of capacity for solar investment in each region, n.

Finally, Gn,g,t and Tn,m represents the current capacities for the generation and transmission line respectively. Table 5.3 summarizes the parameters.

(53)

5.3. Sets, Parameters & Variables 33

Parameters Description Units

ICLp Investment capacity cost of transmission line p . $/MW ICGg Investment capacity cost of plant type g. $/MW VCg Fixed operation and maintenance cost for plant type g. $/MWh

Hg Heat rate for plant type g. MMBtu/MWh Pg Fuel price for plant type g. $/MMBtu OEn,t PV operation reserves region n in hour t. MWh Gn,g Maximum capacity for plants type g in area n. MW Dn,t Demand for region n in hour t. MWh IPVn PV capacity in each region n. MW PVn,t Normalized PV generation for region n in hour t. P.u. Tn,m Maximum capacity for transmission line p. MW FCg Annual Fixed costs of each plant type g. $/MW PVE Cost of PV Operation reserves. $/MWh

TABLE5.3: Optimization Parameters.

5.3.3 Variables

In the following section, the optimization variables will be introduced. Gn,g,t rep-resents the generation for each plant in each region for each hour. Tn,m will be the power transferred from area n to area m. The unserved load will be represented by Un,t.

On the investment side of the optimization: LIP is the new capacity investment

in transmission, where GIn,gwill be the new capacity investment in the generation. Table 5.4 summarizes the variables.

Variables Description Units Gn,g,t Generation in time t, area n, for plant g MW Tn,m Power transferred from area n to m MW Un,t Unserved load in area n, hour t MW LIP capacity investment in transmission MW

GIn,g capacity investment in generation MW

(54)

5.4 Optimization Function

In the following section, the optimization problem will be analyzed. The goal of the optimization model is to equalize generation and demand for a future scenario, given the growth in electricity demand. The model would choose to either build generation capacity or invest in transmission lines. The model will take into account the PV generation along with the PV operational reserves. The model will also ac-count for the fixed cost, variable costs, and fuel prices.

Hg

×

Pg

) ×

Gn,g,t

] +

Un,t

+

OEn,t

×

PVE (5.1)

The optimization function minimizes the total system costs based on the projected growth in electricity demand and PV deployment. The optimization function is con-strained by a set of constraints, described in the following section.

5.5 Constraints

The constraints of the optimization model will be described in the following section.

5.5.1 Load Balance

The hourly load balance constraint matches the generation, demand, and transferred power from other regions. It ensures that thermal units and PV generation will serve the load. This is shown in equation 5.2.

∀

t,

∀

n

∈ N

:

∑

g∈G Gn,g,t

+

PVn,t

+

∑

m∈Pn←m Tn,m

=

Dn

−

Un,t

+

∑

m∈Pn→m Tn,m (5.2)

(55)

5.5. Constraints 35

5.5.2 Generation Capacity Constraint

Generation is limited by the maximum current capacity of plant type in each region, with the additional investment decided by the model. This is shown in equation 5.3.

∀

g

+

GIn,g (5.4)

5.5.4 Transmission Capacity Constraint

For the transmission capacity constraint, the transmission between regions should be limited by the capacity of the line. In addition to investments made by the model in additional line capacities. This is shown in equation 5.5.

(56)

(57)

37

Chapter 6

Case Study: Variability and

Uncertainty of Solar Generation,

The Impact on the Saudi Power

System

6.1 Scope

The previous two chapters, 5 and 3, describe the models applied in this chapter. First, the data described in chapter 4 is passed to the PV generation model. Then, statistical analysis will be performed to quantify and measure the impact of dust and other matters in different Saudi cities. Secondly, the power system model will be used to study the impact of policy decisions on PV deployment in certain regions. The outcome is quantified by measuring the total system costs, CO2emissions, and fossil fuel savings.

The case study aims to address the following: • Optimizing Investments.

• CO2Emissions. • Fuel Consumption.

(58)

38 Chapter 6. Case Study: Variability and Uncertainty of Solar Generation, The Impact on the Saudi Power System

6.2 Hypotheses

6.2.1 General hypotheses

The objective of the case study is to find the optimal investments to ensure reliability and stability, under many policy decisions such as PV capacity, PV location, and fuel prices.

The scope of the study is limited to the Saudi power system, neglecting any trans-mission lines connecting the eastern province with other GCC countries. Hence the case study will consider Saudi Arabia as an isolated system with four regions.

6.2.2 Generation Input Parameters

Capacity of Each Region with Fuel Type

The generation capacity in each region is obtained from the Electricity and Cogen-eration Regulatory Authority (ECRA). Table 6.1 summarizes the genCogen-eration capacity by region, fuel and technology in MW.

Region Fuel\Tech. CC (MW) GT (MW) ST (MW) Central CRUDE 1651.7 7888.08 460 GAS 1196.6 4973.12 HFO 600 353 Diesal 1154.5 W est CRUDE 1096.3 3488 1572 GAS 456 229.5 HFO 800 18486.1 Diesal 2710.9 East CRUDE 680 GAS 8421.3 8750.8 13445 HFO 620.8 Diesal 1158.6 South CRUDE 1106.12 1020 GAS HFO Diesal 2829.74

TABLE6.1: Generation capacity by region, fuel and technology. Below are assumptions made regarding the current generation capacity:

(59)

6.2. Hypotheses 39

• Primary fuel was considered only in table 5.3, since the secondary fuel for non-gas fuels (e.g., HFO, Crude) are of the same kind.

• Diesel generators (0.89% of the generation mix) were left out since there was not enough data on their location.

Variable and fixed cost, heat rate

Assumptions regarding variable and fixed cost as well as heat rate are summarized in Table refTechnologyParameters. The data on generation variables are from the KAPSARC analysis of the Saudi Power system (Rioux, Pierru, and Alkathiri, 2017). The data also includes the investment cost and the lifetime years of each plant type when the model decides to build more capacity.

Ramp & PV Operation Reserves Cost

Assumptions are made regarding ramp cost and PV operation reserves. First, the ramp cost is calculated through fining the change in a generation through delta time over the entire year. Table 6.6 shows the cost of the delta by technology in MW. Regarding the operation reserves cost, the assumption is 20% of fuel from original consumption will be used to operate the spinning reserves.

I.C. (M$/MW) L.T. (Y) F.C (k$/MW) V.C ($/MWh) H.R. MMBtu/MWh gas oil

GT 1.016 25 10.7 1.68 11.3 13.55

CC 1.102 30 19.9 1.24 7.655 9.676

ST 1.68 35 38.7 1.22 10.37 10.2

TABLE6.2: Technology Parameters

Regarding fuel prices, they will be highlighted in detail in the scenario section. Concerning the investments, the discount rate is set to 6%, which is in alignment with other papers that addressed investments in the Saudi power system (Matar et al., 2015; Rioux, Pierru, and Alkathiri, 2017; Matar et al., 2017).

(60)

6.2.3 Load Input Parameters

The load input will be the hourly consumption for each region in Saudi Arabia dur-ing the year 2015. The data are obtained from KAPSARC analysis. Figure 6.1 shows the weekly average demand for all regions in Saudi Arabia.

FIGURE6.1: Weekly Average Demand for All Regions in KSA.

The hourly consumption data obtained represents the year 2015, and since the goal would be to forecast a future scenario (e.g., 2023; 2030), some assumptions are required. ECRA has published a forecast in 2012 which predicts the peak consump-tion for the next ten years 6.2 (ECRA).

FIGURE6.2: ECRA Peck Demand Forecast

(61)

6.2. Hypotheses 41

However, in the past few years, the Saudi government applied many price reforms ranging from fuel to electricity tariffs. That, in return, affected the growth in GDP and electricity consumption. Figure 6.3 shows the actual peak demand for the past couple of years.

FIGURE6.3: Actual Peck Demand Forecast (ECRA).

The reform impact can be observed in figure 6.3. In 2015 the price reforms were introduced, which affected the peak consumption considerably, which in return may alter ECRA’s forecast. Thus in this case-study, the slope of the actual peak demand will be used to forecast a future peak. Then, the step difference will be applied to the hourly electricity load in 2015.

6.2.4 PV Input Parameters

The PV input parameters are derived from the first model in chapter 3. The two input parameters are:

• The normalized PV output for each region is Saudi Arabia. • PV operational reserves requirements normalized.

Figure 6.4 shows an example of the hourly input parameters for a couple of days in Riyadh. Furthermore, assumptions have been made when deciding which solar city profile will represent its corresponding region. Table 6.3 shows the selection of cities and region allocation.

(62)

FIGURE6.4: PV Output & Reserve Requirements

Region South West Center East City Abha Jeddah Riyadh Alhasa

TABLE6.3: Regional Choice of City Solar Profile.

6.3 Methodology

The system cost will be computed for two different years in the future, 2023 and 2030. The selection of the year depends upon the goals of renewable capacity and location given in the National Transformation Program (NTP) and Saudi Vision 2030. For each year, there will be four scenarios described below. Hence, a total of eight years are studied.

6.3.1 National Transformation Program 2023

One of the goals for the national transformation program is to install 9.5 GW of solar capacity by 2023. Figure 6.5 highlights the location of the installed PV capacity.

In order to feed the data to the model, approximations were made concerning PV capacity location, table 6.4.

(63)

6.3. Methodology 43

FIGURE6.5: PV Capacity Location 2023 (Saudi Arabia 2030 Renewable

Energy Targets).

Region South West Center East PV Capacity (GW) 0.2 6 3 0.3

TABLE6.4: Approximated PV Capacity allocation in 2023.

• Reference 2023 Scenario (S1): No PV capacity investments. Fuel prices are set to 3$/MMBtu, representing the subsidized price.

• PV 2023 Scenario (S2): 9.5 GW of PV capacity investments. Fuel prices are set to 3$/MMBtu, representing the subsidized price.

6.3.2 Saudi Vision 2030

One of the goals for Saudi Vision 2030 is to install 40 GW of solar capacity by 2023. Table 6.5, shows the assumptions on PV capacity location, based on REPDO an-nouncements (Saudi Arabia 2030 Renewable Energy Targets).

(64)

Region South West Center East PV Capacity (GW) 5 8 11 16

TABLE6.5: Assumptions on PV Capacity allocation in 2030.

The higher allocation of PV capacity in the center and east is due to the fuel supply. Where the east and center are heavily dependent on gas, thus offsetting generation units with solar will help in exporting more natural gas. Below is the description of each scenario under the Saudi Vision 2030 case.

• Reference 2030 Scenario (S3): No PV capacity investments. Fuel prices are set to 3$/MMBtu, representing the subsidized price.

• PV 2030 Scenario (S4): 40 GW of PV capacity investments. Fuel prices are set to 3$/MMBtu, representing the subsidized price.

6.4 Results & Discussion

The following section highlights the results of the power system model. First, the section highlights the system costs on all scenarios by comparing the cases of in-vestments vs. BAU. Then we look into the effects on the system CO2 emissions, followed by an analysis of the effect of PV investments on the marginal cost. Lastly, we examine the net generation offset due to PV deployment.

6.4.1 Power System Costs

The power system costs are broken down into multiple segments: fixed cost, invest-ment cost, PV operation reserves, operation cost, and ramp cost. The values will be calculated by taking the costs and dividing it by the yearly electricity consumption, resulting in a c$/kWh value.

The ramping costs have been added to the power system costs by examining the change in the dispatchable generation. The ramping occurs regardless of renewable energy sources; thus, the added cost is important when assessing PV penetration (Denholm and Margolis, 2007). The costs of ramping are derived from (Bergh and Delarue, 2015) , as shown in Table 6.6.

(65)

6.4. Results & Discussion 45

Power plant technology Ramping cost ($/∆MW) Oil- or gas-fired Steam turbines 1.61

Open-cycle gas turbines 0.92 Combined-cycle gas turbines 0.58

TABLE6.6: Ramping costs for power plant technologies (Bergh and Delarue, 2015).

NTP 2023 results and analysis

This section examines the results under the NTP 2030 plan. Due to excess capacity, and the price reforms (e.g., increase in electricity tariff), we observed no additional investments in thermal units capacity. The value of maintaining PV operation re-serves amount to 0.0055 c$/kWh. Furthermore, scenario 1 has an increase of 5% in the operation cost segment when compared to scenario two due to the penetration of PV. Ramping cost is higher in Scenario 2 when compared to scenario one by 11.6% because of the variability in PV. In total, there is a reduction of 30% in total system costs in favor of scenario 2. Table 6.7 summarizes the results.

S1 (c$/kWh) S2 (c$/kWh) Fixed Cost 0.4888 0.4888 Investment Cost 0 0 PV Operation Reserves 0 0.0055 Operation Cost 3.0625 2.9082 Ramp Cost 0.0043 0.0048 Total 4.8660 3.4073

TABLE6.7: System Cost Under S1 & S2.

SV 2030 results and analysis

This section examines the results under the Saudi Vision 2030 plan. As a result of the increase in electricity demand, we observe capacity investments in the combined cycle for scenario 3 and 4 to match demand. The additional capacity in scenario 3 amounts to 5 GW, wherein scenario four it is 4.3 GW. Furthermore, the value of maintaining PV operation reserves amount to 0.0177 c$/kWh. Scenario 4 has an increase of 18% in the operation cost segment when compared to scenario three, due to the penetration of PV. Ramping cost is higher in Scenario 3 when compared to

(66)

scenario four by 70% because of the variability in PV. In total, there is a reduction of 36% in total system costs in favor of scenario 3. Table 6.8 summarizes the results.

S3 (c$/kWh) S4 (c$/kWh) Fixed Cost 0.4301 0.4271 Investment Cost 0.0810 0.0690 PV Operation Reserves 0 0.0177 Operation Cost 3.0280 2.5643 Ramp Cost 0.0046 0.0077 Total 4.8660 3.0858

TABLE6.8: System Cost Under S3 & S4.

6.4.2 Net generation offset

This section highlights the net generation difference between the scenarios. In other words, the generation being displaced by PV, broken down into technologies.

NTP 2023 results and analysis

The net generation offset as a result of the PV generation is compared to a business as usual scenario in figure 6.6. Under the assumption that steam turbines run on oil, and the south region power plants are only consuming oil. The power system will save 27.3 million barrels of oil yearly using the thermal conversion table 6.2.

Determining an Optimal Level of Power System Investments Under Large Scale Penetration of Solar Power in Saudi Arabia

Determining an Optimal Level of

Power System Investments Under

Large Scale Penetration of Solar

Power in Saudi Arabia

HATEM ALATAWI

KTH

M

T

Determining an Optimal Level of Power System

Investments Under Large Scale Penetration of

Solar Power in Saudi Arabia

Abstract

Sammanfattning

Acknowledgements

Contents

List of Figures

List of Tables

List of Abbreviations

Chapter 1

Introduction

1.1

Background

1.2

Objective

1.3

Approach for the studies

1.4

Thesis Structure

Chapter 2

Saudi Arabia Electricity Market

Review

2.1

History of the Saudi Electricity Sector

2.2

Current state

2.3

Renewable Energy Sources Investments

Chapter 3

Photovoltaic DC model

3.1

Input Data

3.2

PV DC output

=

(

+

(

−

))

=

+

=

+

=

+

+

3.3

PV Operational Reserves

=

(

+

) =

(

) +

(

) × [

(

+

) −

(

)]

=

(

) −

(

) +

(

) ×

(