STOCKHOLM SWEDEN 2019,
Hydropower Modelling of Continental Europe Using EMPS
REBECA BRENES BRENES
KTH ROYAL INSTITUTE OF TECHNOLOGY
SCHOOL OF ELECTRICAL ENGINEERING AND COMPUTER SCIENCE
Using EMPS
Authors
Rebeca Brenes Brenes <rebecabb@kth.se>
Electric Power Engineering
KTH Royal Institute of Technology
Place for Project
Stockholm, Sweden
Examiner
Mikael Amelin
KTH Royal Institute of Technology
Supervisors
Lennart Söder
KTH Royal Institute of Technology
Evelin Blom
KTH Royal Institute of Technology
Jussi Mäkelä Fortum
Flexible hydropower plays a vital role when integrating large shares of variable renewable generation in the power system. This project proposes a method for creating a stochastic model for hydropower in Continental Europe using EMPS, a power market simulator software that specialises on hydrothermal power systems.
The model is intended to be used in a power market analysis context, to assess the hydropower behaviour in Continental Europe in the medium and long term.
The first step to create the model was to gather data from multiple sources regarding hydropower stations and reservoirs located in Continental Europe. The gathered data was built and unified into a geographical information system to provide visual analysis and facilitate performing of different operations and algorithms. The countries under study were split into different areas depending on the water basins and other location factors. Representative inflow series were created for each area, and these were fed directly into EMPS. The stations and reservoirs were aggregated per area and the inputs required by EMPS were calculated using a model based on historical data and an optimisation process.
The model provides results for 45 different weather scenarios. The hydropower generation and reservoir content for each country were validated using historical data. In general, the results obtained provide a satisfactory representation of the hydropower in the region under study. Fine tuning is still required to improve the results, but this project successfully provides a solid method to create the core of the stochastic hydropower model of Continental Europe in EMPS.
Keywords
Hydropower, Continental Europe, stochastic model, geographic information systems
ii
Flexibel vattenkraft är grundläggande för reglering i elkraftsystem med en stor mängd variabel förnybar elproduktion. Detta projekt föreslår en metod för att skapa en stokastisk modell av vattenkraftproduktion i Kontinentaleuropa med EMPS, en mjukvara som simulerar elmarknaden och optimerar produktion i system med både vattenkraft och kraftvärme. Modellen är avsedd för att användas till elmarknadsanalys, för att utvärdera vattenkraftens beteende i Kontinentaleuropa på medellång och lång sikt.
Första stegen för att skapa modellen var att samla information från olika källor om vattenkraftverken och magasinen i Kontinentaleuropa. Den insamlade informationen byggdes in i ett gemensamt geografiskt informationssystem. Detta för att underlätta visualisering och utföra operationer med datan. Länderna som studerades delades in i olika områden beroende på vattendrag och andra geografiska faktorer, inklusive avrinningsområden. Representativa inflöden skapades för varje område, och användes som indata till EMPS. Kraftverken och magasinen aggregerades utefter område. Indata för EMPS beräknades med en modell baserad på historiska data och en optimeringsprocess.
Modellen ger resultat för 45 olika väderscenarion. Vattenkraftsproduktionen och nivån i magasinen för varje land validerades med historisk data. Generellt är de erhållna resultaten en bra representation av vattenkraft i alla områden som studerades. Modellen behöver finjusteras för att förbättra resultaten, men detta projekt ger en solid metod som fungerar som kärnan till den stokastika vattenkraftmodellen av Kontinentaleuropa i EMPS.
Nyckelord
Vattenkraft, Kontinentaleuropa, stokastisk modell, geografiskt informationssystem
iii
EMPS EFI’s Multi-area Power-market Simulator GIS Geographic Information System
ENTSO-E European Network of Transmission System Operators for Electricity IHA International Hydropower Association
QGIS Quantum GIS
LP Linear Problem
NVE The Norwegian Water Resources and Energy Directorate RES Renewable Energy Sources
SDP Stochastic Dynamic Programming SFOE Swiss Federal Office of Energy
SMHI Swedish Meteorological and Hydrological Institute
iv
P hydropower [M W ]
η efficiency
ρ density [kg
m3
] Q water discharge [
m3 s
]
g gravity [m
s2
]
h head or height of fall [m]
Qmax maximal discharge[
m3 s
]
e energy equivalent[kW h
m3
] Jt cost function at time t
xrest reservoir level at the beginning of week t
Lkt cost dependent on operation in week t, scenario k xhydt hydropower production in week t
S (xresT ) value of final reservoir at the end of week T
Ciτ cost of supplementary generation option i in week τ xSiτ supplementary generation option i in week τ
Diτ cost of demand reduction option i in week τ xrediτ demand reduction option i in week τ
Isup number of supplementary generation options Ired number of demand reduction contracts ximpit import option i in week t
xexpit export option i in week t
xiniit reference demand option i in week t ytinf inflow in week t
xspillt spillage in week t κ0 calculated water value
pi probability of a certain inflow i κi water value with a certain inflow i Iu unregulated inflow [GW h
week
]
Gu generation due to non regulated inflow to the power systems [GW h
week
] Gmd generation due to minimum discharge and/or bypass constraints [GW h
week
] Gas generation necessary to avoid spillage [GW h
week
]
v
Ep energy used for pumping to avoid spillage week Ir regulated inflow [GW h
week
]
Sg sum production (Including time-of-use purchase contracts)[GW h
week
] Vres increase in reservoir volume (or decrease in reservoir volume -) [GW h
week
] cthi,j,l thermal cost in area i in load period l [NOK]
ctrj,l transmission cost of line j in load period l [NOK]
prati,j,l demand rationing segment j in area i in load period l [GWh]
strti,j,l relative start-up cost in area i in load period l [NOK]
xdm,i reservoir segment m for area i at the end of week t [GWh]
νj rationing cost of segment j L number of load periods NA number of areas
Nm number of discrete reservoir segments Nrat number of rationing levels
Nt(i) number of thermal unit in area i Ntr number of transmission lines
SCi,j start-up cost of unit j in area i [NOK]
Wm,i marginal end of week water value for reservoir level m in area i xi,t+1 reservoir level in area i at the end of week t [GWh]
qi,t total discharge in area i in week t [GWh]
si,t hydro spillage in area i in week t [GWh]
Xi,t initial reservoir level for area i at the beginning of week t [GWh]
Il,i,t regulated inflow to area i in week t αj,l loss factor of line j in load period l
Di,l inelastic demand in area i in load period l
phli,l hydro production in area i in load period l [GWh]
pti,j,l thermal production of unit j in area i in load period l [GWh]
Mt1,i transmission lines importing to area i Mt2,i transmission lines exporting to area i
tr1,j,l transmission on line j in direction k in load period l [GWh], k = 1, 2 α degree of regulation
vi
Rmax maximum reservoir capacity Ia annual inflow to the reservoir
D damping
R actual reservoir
Rtarget weekly target for the reservoir level
Resum desired energy in the total system at the aggregated level CR capacity ratio
Psb total capacity per sub-basin [M W ] Psa total capacity per specific area [M W ] Isa−n inflow series of specific area n [m3/s]
CRm capacity ratio of sub-basin m Isb−m inflow series of sub-basin m [
m3 s
]
VA reservoir volume of area [M m3]
Vn reservoir volume of individual reservoir n [M m3]
En yearly energy produced from hydropower in area region n [GW h]
en energy equivalent in area region n [GW h
M m3
] In yearly natural inflow in area region n [M m3]
Ei−2017 generation of year i scaled to installed capacity of 2017 [GW h]
Ei generation of year i [GW h]
P2017 installed capacity in year 2017 [M W ] Pi installed capacity in year i [M W ]
En yearly energy produced from hydropower in area region n [GW h]
ET total yearly energy produced from hydropower at a country level [GW h]
Pn installed capacity in area n [M W ]
PT total installed capacity at a country level [M W ]
nsamples number of samples, in this case depending on number of years available X training data, in this case the yearly natural inflows
ω optimisation variable, in this case energy equivalent y target values, in this case the yearly energy generation α constant for LASSO formulation, default value 1.0
Q discharge [
m3 s
]
vii
Iy−reg yearly mean regulated inflow [M m ]
Ec−stg yearly generation at country level from hydro storage units [GW h]
Pa−stg installed capacity of storage units in area n [M W ]
Pc−stg installed capacity of storage units at a country level [M W ] en energy equivalent in area n[kW h
m3
] Iy−ureg yearly mean unregulated inflow [M m3]
Ec−ror yearly generation at country level from run-of-river units [GW h]
Pa−ror installed capacity of run-of-river units in area n [M W ]
Pc−ror installed capacity of run-of-river units at a country level [M W ] ηp pump efficiency
Phydraulic hydropower or power delivered to the fluid [M W ] Pbrake pump power or drive power to the pump [M W ] Qp pump flow capacity [ms3]
VP volume for pump reservoirs, upper and lower [M m3] Esp energy stored for pumping [GW h]
viii
1 Introduction 2
1.1 Background . . . 2
1.2 Problem . . . 2
1.3 Purpose . . . 3
1.4 Goal . . . 3
1.5 Methodology . . . 4
1.6 Delimitations . . . 4
1.7 Outline . . . 5
2 Theory 6 2.1 Hydrology . . . 6
2.2 Hydropower . . . 7
2.3 Literature Review . . . 15
3 EMPS 19 3.1 Model Concept . . . 20
3.2 Hydropower Modules . . . 22
3.3 Water Value Method . . . 29
3.4 Strategy Phase . . . 33
3.5 Simulation Phase . . . 37
3.6 Norway Example . . . 43
4 Data Processing with GIS 46 4.1 Hydropower Plants in Continental Europe . . . 46
4.2 Main Reservoirs in Continental Europe . . . 49
4.3 Pumped Storage Database . . . 50
4.4 Main Rivers in Continental Europe . . . 52
4.5 Areas Definition Criteria . . . 54
4.6 Inflow Series Creation . . . 59
5 Hydropower Model 65 5.1 Model Concept . . . 66
5.2 EMPS Inputs . . . 68
ix
6 Results and Validation 86 6.1 General Results . . . 87 6.2 Switzerland . . . 88 6.3 Validation . . . 91
7 Discussion 100
7.1 Recommendations and Future Work . . . 102
8 Conclusions 103
x
2.1 Visual representation of a catchment area and sub-basin. . . 7
2.2 Common layout of a hydropower station. . . 8
2.3 Installed capacity per country, data from Platts database. . . 12
2.4 Yearly hydropower generation per country, data from IHA and ENTSO-E. . . 13
2.5 Reservoir volumetric capacity per country, data from AQUASTAT. . 14
2.6 Water reservoirs and hydro storage plants maximum capacity in GWh between 2015 and 2018, data from ENTSO-E. . . 14
3.1 Example of EMPS model concept, nine areas with respective transmission capacity between areas. . . 20
3.2 Modelling Concept of EMPS. . . 21
3.3 Standard hydropower module in EMPS. . . 22
3.4 Example of reservoir curve and power-discharge curve of a hydropower module. . . 24
3.5 Different configurations for hydrological coupling between reservoirs in EMPS. . . 26
3.6 EMPS dialog window showing main parameters of a hydropower module. . . 28
3.7 Planning period divided into weeks, different scenarios. . . 31
3.8 Illustration for principle of water value calculation. . . 32
3.9 Iteration process for the water value calculation. . . 33
3.10 Aggregated system model in the strategy phase. . . 34
3.11 Example of single reservoir model aggregation of an area model. . . . 35
3.12 Simulation process of the EMPS model at a weekly level. . . 43
3.13 Grytten hydropower system schematic in Norwegian area Norgemidt. 45 4.1 Comparison of installed capacity according to different sources. . . 47
4.2 Hydropower plants in selected countries in Europe sorted by installed capacity in MW, Platts DB. . . 48
4.3 Hydropower plants in selected countries in Europe sorted by installed capacity in MW, Global DB. . . 49
4.4 Hydropower reservoirs in selected countries in Europe sorted by capacity in Mm3. . . 50
xi
4.7 Main catchment areas in Continental Europe . . . 52
4.8 Average inflow per sub-basins in Continental Europe (1981-2017). . . 53
4.9 France Area with geographical division for module definition. . . 55
4.10 Switzerland Area with hydrology division. . . 56
4.11 Switzerland Area with hydrological division and elevation map. . . . 56
4.12 Switzerland Area with hydrological division and average inflow. . . . 57
4.13 Module division for EMPS modelling. . . 58
4.14 Inflow creation representation in an specific area . . . 61
4.15 Daily inflow series for Austria . . . 62
4.16 Daily inflow series for Italy . . . 63
4.17 Daily inflow series for Switzerland . . . 63
4.18 Daily inflow series for Germany . . . 64
4.19 Daily inflow series for France . . . 64
5.1 Country layout of model concept. . . 66
5.2 Model concept of hydropower modules for Continental Europe. . . 67
5.3 Reservoir capacity per country from AQUASTAT database. . . 69
5.4 Energy equivalents in Continental Europe map. . . 73
5.5 Adjustment of yearly mean regulated and regulated inflow proportion in Austria. . . 82
5.6 Adjustment of run-of-river and storage capacity proportion in France. 84 5.7 Adjustment of yearly generation per specific area. . . 85
6.1 Energy inflows in Switzerland. . . 89
6.2 Hydropower production in Switzerland. . . 89
6.3 Reservoir content in Switzerland. . . 90
6.4 Hydropower production validation of Switzerland. . . 92
6.5 Reservoir content validation of Switzerland. . . 92
6.6 Hydropower production validation of Austria. . . 93
6.7 Reservoir content validation of Austria. . . 94
6.8 Hydropower production validation of France. . . 95
6.9 Reservoir content validation of France. . . 96
6.10 Hydropower production validation of Germany. . . 97
xii
6.11 Hydropower production validation of Italy. . . 98
6.12 Reservoir content validation of Italy. . . 99
List of Tables
2.1 Hydropower plants classification by installed capacity. . . 96.1 General simulation results for year 2020, average values for the 45 scenarios. . . 87
A.1 Country division in specific areas . . . 112
A.1 Country division in specific areas . . . 113
A.1 Country division in specific areas . . . 114
1 Introduction
1.1 Background
Hydropower has historically been one of the main power generation sources, currently it is the largest source of renewable electricity in Europe [1]. Wind and solar power generation are expected to rapidly grow in the upcoming years as a direct consequence of the decarbonization of the power sector in Europe [2]. This has simultaneously amplified the attention provided to hydropower with storage, as it is one of the most flexible types of generation; facilitating the integration of larger variable renewable sources into the power systems [1, 2, 3].
Medium and long term forecasting of hydropower is vital to assess the water resources and ensure that the hydropower operation goes accordingly to keep the reservoirs in acceptable levels. However, this is not a simple task, as it strongly depends on the inflows; which are affected by uncertain weather conditions such as precipitation and the snow-melting process. For this reason it is necessary to consider the stochastic nature of the problem, to properly assess the hydropower future generation. [2, 4]
Power companies use different tools to obtain realistic forecasts of hydropower generation for the planning process and improve the performance of the trading in electricity markets. A popular tool used in the Nordic countries is the EMPS (EFI’s Multi-area Power-market Simulator), developed by SINTEF; a Norwegian research organisation. The speciality of EMPS is optimisation and simulation of power systems with significant amounts of hydropower [5].
1.2 Problem
Stochastic modelling is fundamental to describe the uncertainty of the hydropower behaviour and obtain realistic results in the long term market forecasts, this can be achieved using powerful market simulators like EMPS [6]. Fortum has a functioning model of the Nordic market in EMPS. The next challenge is to create a stochastic model of hydropower in Continental Europe in EMPS, using limited data.
1.3 Purpose
The purpose of this thesis work is to define a method to model the hydropower of Continental Europe in EMPS, and validate the results obtained from the model by assessing the behaviour of flexible hydropower in the studied region. The model should consider the stochastic nature of hydrological values related to hydropower operation, such as representative inflow series for each region. The model should also respect the modelling by area concept of EMPS, which means that the geographical location of the hydropower plants in Continental Europe must be taken into consideration to some degree within the model.
1.4 Goal
In total eight goals are defined ahead as part of the framework of this thesis.
1. Understand the main characteristics of hydropower in Continental Europe and detect the key elements that the model must represent.
2. Study the EMPS model to understand how the tool functions in general and explain how hydropower is modelled.
3. Learn how to create new hydropower modules in EMPS and also how to edit existing hydropower modules.
4. Gather required input data for the Continental Europe model, such as information of hydropower plants, inflow series of the studied areas and historical data of the hydropower and reservoir contents.
5. Propose a method for defining new EMPS area modules and create representative inflow series for each specific area using geo-located information.
6. Create a model that represents the hydropower generation in Continental Europe.
7. Propose a method to implement the created model in EMPS, considering the input constraints of EMPS.
8. Run simulations with the existing EMPS model integrated with the new EMPS
modules of Continental Europe and validate the created model by comparing the simulation results with historical data.
1.5 Methodology
The first phase of the project consisted of defining the project goals and doing a general literature review. A state of the art of hydropower models of Continental Europe and hydropower equivalents was done. The second phase involved familiarising with EMPS, and understanding its structure, as well as how to create new hydropower modules; all this using official documentation from SINTEF. Next, a geographical information system (GIS) was built based on all the available information on hydropower installations and inflows in Continental Europe.
Different methods were tested to create the model inputs for EMPS. These methods involved processing algorithms, combining programming in Python and GIS, as well as optimisation processes. Finally, the results were analysed and validated with historical data.
1.6 Delimitations
This study focuses on the hydropower modelling of Continental Europe at a higher level, meaning that individual and specific characteristics are not modelled in detail;
instead, the hydropower behaviour is modelled per areas. The model is limited by the available information of each hydropower plant, which vary from case to case accordingly to the openness and availability of data in each country.
The countries included in EMPS are: Albania, Austria, Bosnia and Herzegovina, Bulgaria, Croatia, France, Germany, Hungary, Italy, Latvia, Macedonia, Montenegro, Portugal, Romania, Serbia, Slovakia, Slovenia, Spain and Switzerland.
However, given the magnitude of the model, this report focuses on explaining the general method and provide more detailed explanation for chosen example countries, which are Austria, France, Germany, Italy, and Switzerland; but the method is still valid for the Balkans and the Iberian Peninsula as well.
The model is strictly implemented in EMPS, which directly delimits the model by
the EMPS functionality and restrictions. Extra processing steps are required to provide the data in the format required by EMPS, and calibration steps are also needed to fine tune the results of the model. Finally, the explanation of the general EMPS tool is limited by the information and guides provided by SINTEF; as there are some elements of the EMPS model that are out of reach of users.
1.7 Outline
An overview of each chapter in this report is presented:
• Chapter 2 provides general theoretical concepts related to hydrology and hydropower, an overview of the hydropower situation in Continental Europe as well as a literature review of previous studies on hydropower equivalents.
• Chapter 3 includes the most important elements of the workflow of EMPS, in particular it explains how the hydropower is modelled. An overview of the strategy and simulation phase of EMPS is included, as well as the general theory behind it such as the area modelling and water value method.
• Chapter 4 presents the dataset available for the study. It also includes how this data is processed and built into a common system to facilitate further operations. Some strategical decisions for creating the model are also introduced in this chapter.
• Chapter 5 contains the core of the developed model, it explains how the information of Chapter 4 is used to obtain inputs required by EMPS. This chapter also includes a section on how to calibrate the model.
• Chapter 6 presents the model results for some example countries. The results are analysed and validated with historical data.
• Finally, Chapter 7 summarises the main conclusions of this study and provides recommendations for future studies and opportunities for improvement.
2 Theory
In this chapter fundamental concepts about hydrology and hydropower are introduced, as well as literature review of the previous work done on related topics such as hydropower stochastic modelling for long term planning, hydropower equivalents and in general hydropower models in Continental Europe.
2.1 Hydrology
Hydrology refers to the study of water over and below the surface, particularly the movement through different pathways on the continental plates, such as rivers, streams, lakes and aquifer [7]. A significant part of the uncertainty of hydropower models comes from the inflows [2], which is directly linked to the hydrological systems.
Hydrological cycle: Water is in constant circulation as a part of the natural hydrological process, also known as water cycle; water changes states and is transported through the atmosphere, hydrosphere and lithosphere [9, 8]. The main driving force behind this process is the solar energy and gravity [9], and is composed by different elements such as water evaporation, transport of water in the atmosphere, cloud formation and dynamics, liquid and solid precipitation, soil moisture and groundwater [10]. This process is vital for hydropower as it is the responsible of river and reservoir inflows. The hydrological cycle can be quantified with inputs and outputs, this is often referred to the water balance.
Inflow: In general, bodies of water like rivers, lakes and reservoirs have a source of water known as the inflow (also referred as streamflow) [14]. The inflow is often referred in terms of the volume of incoming water in a time unit [m3/s] [14]. Inflows can have different sources like precipitation, water streams joining a river or rivers flowing into lakes and reservoirs; and even discharge from upstream hydropower or pumping units [2]. In rivers the water flows mainly due the effect of gravity but in the case of lakes and reservoirs it is attributed to pressure and attraction forces [13].
It is important to highlight that depending on the location, the precipitation may come as snow in the winter.
Catchment area: The body water is delimited by the catchment area (also known as drainage basin or river basin), it contains upstream areas that connect in certain points downstream; all these areas contribute to the river and form the total catchment area [15]. Figure 2.1 presents an example of this concept where the catchment area is marked in yellow, it delimits the impact area of the water body, containing upstream and downstream regions of the river system. The smaller areas marked in lighter blue correspond to the sub-basins, which can be defined as smaller sections of the catchment area.
Figure 2.1: Visual representation of a catchment area and sub-basin.
2.2 Hydropower
The operation principle of hydropower consists of harnessing the energy from running water to generate electricity. The basic physical layout of a hydropower plants is composed by a dam that holds the water back in a reservoir, a powerhouse and a penstock connecting both of them. The water upstream in the reservoir (head water level) has a higher level than the water downstream (tail water level), creating a height difference also known as head. At the bottom of the damn there is the water intake and a control gate, which allows to control the water flow into the penstock.
The running water in the penstock meets a turbine which produces a mechanical movement in the shaft. The turbine is coupled to an electrical generator, which is connected to the power grid. [16, 17]
Figure 2.2: Common layout of a hydropower station [17].
Equation (1) presents the hydropower equation, which allows to calculate how much power can be obtained from the water given an overall efficiency, certain water height fall and inflow conditions [18].
P = η· ˙m · g · h = η · ρ · Q · g · h (1) P power [W]
˙
m mass flow of water falling [kg/s]
η overall efficiency of power station ρ water density [kg/m3]
Q flow rate of water [m3/s]
g gravity [m/s2]
h head or height of fall [m]
The key aspect for hydropower production is to utilise the difference in potential energy between the different water levels. Depending on the terrain characteristics and elevation difference, it might be possible to obtain a larger height fall by finding a strategical point to put the intake to obtain the largest fall. In other cases, where the head is low, the penstock instead may be shorter. [16]
There are three main ways to classify hydropower plants [17]:
• By installed capacity: classification presented in Table 2.1. It is important to mention that the specific ranks may vary between sources, but generally small hydropower refers to plants with less than 10 MW.
Table 2.1: Hydropower plants classification by installed capacity.
Type Installed Capacity
Pico 0 - 10 kW
Micro 10 - 100 kW
Mini 100 kW - 1 MW
Small 1 - 10 MW
Medium 10 - 100 MW Large Over 100 MW
• By head height: low (less than 30 m), medium (30 - 300 m) and high head (more than 300 m) [20].
• By operative mode: run-of-river, storage and pumped storage.
Run-of-river: These type of hydropower plants utilise the water flowing through a river using a canal to guide the water directly towards the turbine [21]. Run-of-river plants do not have large water storage systems, meaning that they must harness the water as it naturally flows [20]. During wet periods water may be spilled and during the dry season they usually have a low generating capacity; sometimes they have a pondage that allows to store water during short periods [20]. Run-of-river plants offer some flexibility for daily fluctuations, but they are usually designed to provide a continuous amount of electricity determined by the design inflow [21].
Storage power plants: Also known as reservoir plants. Oppositely to run-of- river plants, these have a large storage system where a dam holds the water in a reservoir [21]. Reservoirs allow to store the water, in some cases for a couple of days or weeks, but usually between wet and dry seasons [20]. Depending on the reservoir sizes it could even be inter-annual storage [6] and can operate independently for many weeks [21], this means that they can use the excess water from previous years during dry years, or save water for future usage during wet years. Storage plants can be used as base load due to their firm capacity or as peak load plant given their ability to quickly shut down and start up [20, 21].
Pumped Storage: Usually pumped storage plants are storage power plants that have a pump installed, to be able to pump back the water into the upper reservoir.
The pumping is done during moments with lower electricity prices, to later utilise the water when the prices are higher; usually periods with high energy demand [17, 21]. Typically, the pumping occurs during nighttime [17].
2.2.1 Hydropower Modelling and Scheduling
In this context, hydropower modelling refers to models used for hydropower scheduling; which is related to the production optimisation that minimises the cost of generation while optimally using the hydro resources [6]. Hydropower modelling depends on the application that the model is used for. This can refer to long term scheduling, seasonal scheduling, and at a more operational level it could be a model used for the short term optimisation. The type of model for each of these purposes will differ in the detail level, particularly in parameters taken into consideration and the results obtained from the model. [6]
Long term scheduling: Usually long term in this context refers to 1 to 5 years ahead. In this case stochastic models are used to obtain robust models, and to simulate different scenarios of the value of water, as the optimisation seeks to utilise the resources in the best way according to the future prices [6]. The stochastic data is usually statistical data describing meteorological variables such as temperature, wind and precipitation (or inflows in case of EMPS, which will be later explained in Chapter 3), as well as demand. The long term models are particularly useful to understand the market horizon, which is why they are larger, containing complex physical systems but at the same time the level of detail of each hydropower plant is more general. Often in the long term models power plants are aggregated in order to keep the solution time feasible, otherwise it could take extremely long time to converge to a solution. [6]
Seasonal scheduling: The seasonal scheduling is done in a time range of 3 to 18 months ahead. This is an intermediate step between long and short term scheduling.
It has the same physical description of the system as the long term model, but the
mathematical approach is different; as greater details are required to have more exact water values in the individual reservoirs. Seasonal scheduling can be done with stochastic or multi-scenario models [6], in the Nordics, ProdRisk1 is commonly used for this purpose.
Short term optimisation: As the name suggests, the time frame is shorter, up to 1 to 2 weeks ahead. The short term optimisation is used as reference for the actual operation of the hydropower plants. Therefore, they require exact models, with high level of detail to represent all the important variables and different conditions. The short term optimisation is complemented by simulations that fine tune the operation plan of the power plants [6].
The hydropower model proposed in this project is intended to be used in long and medium term applications.
2.2.2 Hydropower in Continental Europe
Hydropower is considered a mature technology, as it has been used in Europe for over a century to produce electricity [22]. In total, there are 249 GW installed in Europe, and it is expected to grow even more given the renewable energy goals of the European Union [1]. The country with the most installed hydropower capacity in the region is Norway, also being the top producer. A list of the top six countries and their installed capacity [1] is presented ahead.
1. Norway 31 837 MW 2. Turkey 26 681 MW 3. France 25 517 MW 4. Italy 21 884 MW 5. Spain 20 344 MW 6. Switzerland 16 657 MW
1ProdRisk is a program developed by SINTEF for hydropower optimisation and simulation, which uses stochastic dual dynamic programming [23].
In the beginning of the 20th century many hydropower plants were built. As they have a long lifetime, of around 80 years [25], there are many that are still operating;
others have gone through modernisation procedures [22]. The current hydropower potential in Europe is based mainly on modernisation [22], and project development in eastern Europe, specifically the Western Balkan region [1]. Most of the biggest hydropower projects were built in the 1960s and 1970s, therefore they will soon require an upgrade [1, 22]. Repowering projects could offer gains of 5-10%, but at the same time it requires extra technical and legal challenges [22].
This study is delimited to analyse the countries mentioned in Section 1.6, to which are referred as Continental Europe in this study. Figure 2.3 presents a visual representation of the countries included. It is composed by both the traditionally larger producers of hydropower in western Europe and the growing region of the Balkans.
Figure 2.3: Installed capacity per country, data from Platts database [26].
To assess hydropower in Continental Europe there are four parameters that should be analysed. These are the installed capacity, the energy produced from hydropower,
the reservoirs capacity and the hydropower potential. Figure 2.3 also presents the installed capacity per country, and Figure 2.4 the yearly hydropower generation.
Comparing both pictures, France is leading both in terms of installed capacity and generation; Italy is the second place of installed capacity but third in generation.
Another example is Austria, which is the second that generates most hydropower electricity, but is placed as fifth with most capacity.
Figure 2.4: Yearly hydropower generation per country, data from IHA [1] and ENTSO-E [24].
Regarding reservoir capacity, the databases with concrete data about reservoirs are scarcer. Therefore, to study the reservoirs in different countries two parameters are studied: the volumetric capacity and the energy stored in reservoirs, both per country of study. Figure 2.5 presents the volumetric capacity, as observed Spain is directly outstanding in this category. To complement the analysis, Figure 2.6 presents the maximum energy stored in the reservoirs. The Iberian Peninsula is leading in terms of volumetric reservoir capacity, and in actual energy stored, besides Spain, the Alpine region is among the top.
Pumped storage plants are also becoming very relevant for the security of supply in Europe, specially for the grid reserve in Southern Germany. Southern Germany, Austria and Switzerland have a combined installed capacity of approximately 5900 MW by 2015 [30], and constantly increasing. For instance, in 2017, over 1150 MW of pumped storage capacity was added in Europe [1]. In general, the
pumped storage installations are gaining strength, with new projects in areas outside the Alpine region. For example in Portugal, it is strategically implemented to complement the wind and solar power in the region [1].
Figure 2.5: Reservoir volumetric capacity per country, data from AQUASTAT [27].
Figure 2.6: Water reservoirs and hydro storage plants maximum capacity in GWh between 2015 and 2018, data from ENTSO-E [28].
Technical studies point that the western Balkan region (which refers to Albania, Bosnia and Herzegovina, Kosovo, Macedonia, Montenegro and Serbia) has an estimated potential of 80000 GWh, particularly in the mountainous area of Montenegro and Albania [29]. There are many ambitious projects in the area, however, due to environmental concerns many of them may not be implemented [29].
The case of Nordic Hydropower and wind generation in Denmark is an example of how hydropower storage installations can be optimised to interact with the variable RES in Continental Europe, with the support of pumped storage to supply peak demand and avoid curtailments [30]. Hydropower is an enabler of other technologies, with the combination of flexibility, firm capacity and electricity storage [1, 30].
2.3 Literature Review
A representative hydropower model is vital when studying systems with a large share of hydropower. However, modelling multi-river systems, considering the stochasticity of the problem, is often a challenging problem as it requires high computational time. Another challenge when modelling hydropower is the lack of information. Most of the public power plants databases include only the installed capacity per power plant, but other required variables are not included and often are only known by the power plant owner and operator. For this reason, modelling the hydropower in several countries of Europe, becomes a complex task.
There are plenty of studies regarding hydropower modelling and equivalents, as this has been a relevant topic since the seventies. However, there are only a few studies that propose models that could realistically be applied at European level. This section will briefly present the hydropower models for Europe found, as well as a short summary of the existing aggregation methods and hydropower equivalents, as this is essential to reduce the complexity of the problem.
2.3.1 Continental Europe Modelling
This study is focused on stochastic hydropower modelling in Continental Europe for mid and long term analysis using EMPS, which is a wide yet specific topic. To the author’s knowledge, there is no publicly available study with the same or similar objective.
Härtel and Korpås [31] propose an aggregation method for modelling hydropower in Europe, with the objective to study the behaviour of hydropower in a highly decarbonised system. A bottom-up approach is followed by identifying the most relevant hydropower plants and reservoirs across Europe. A simple natural inflow model is implemented and an algorithm to interconnect the elements is used.
A typical one-dam representation is carried out, which is obtained by summing the water stored at each station, while taking into consideration the downstream generating capacity by studying the water pathways with the interconnection.
Another equivalent using clustering is also proposed. The study reveals that the equivalents provide marginally higher levels of production. The authors point out that this method may be helpful in the stochastic case, however the database is not public, and there is no description of the interconnection algorithm; which are both necessary to replicate and implement into an stochastic version.
Brovold, Skar and Fosso describe a method in [32] to create a hydropower planning formulation in a long-term for the European expansion planning context. In this study the hydropower of the EMPIRE model2 is improved by adding the water values concept to implement the water-saving feature to the existing model. The dispatch model of the EMPIRE model consists of energy balance constraints for every nodes as well as technical and transmission constraints, and the hydropower is simply handled by summing the generating and storage capacity at each node of the simplified network [33]. The unregulated and regulated hydropower are modelled separately in EMPIRE [32] and the generation is limited by the annual energy production, this limit is used to handle the water uncertainty [33]. EMPIRE is more focused on the investments part of the model, whereas EMPS implements a more sophisticated operational model [33].
2European Model for Power system Investment with (high shares of) Renewable Energy, developed by NTNU, Norway.
2.3.2 Hydropower Equivalents
Brandão validates in [34] a simplified equivalent reservoir representation of a multi- reservoir hydroelectric system with the purpose to obtain the optimal operation, using previous simplification aggregation techniques introduced by Arvanitidis and Rosing in [35]. Brandão carries out a comparison of simulation results of a real system with detailed information, against the results of the equivalent model of Arvanitidis and Rosing; OPOM and EROM 3 respectively. The method presented in [35] consists of a single measure potential energy concept to obtain a one-reservoir representation of the hydroelectric systems. This implies that each reservoir is represented using the potential energy, which can be obtained by multiplying the reservoir stored volume by the productivity (or energy equivalent, as referred to in previous sections), taking into consideration its own productivity and the one of the reservoirs downstream [34]. In this model, a natural potential energy inflow is calculated using the water balance equation and the energy equivalent.
Results from the case study of Brandão in the São Francisco River, in Brazil, highlight that the results of the equivalent model are satisfactory with less than 6% underestimation. There are clear computational advantages of the equivalent model, such as less time to solve the problem up to ten times faster, but for larger systems the behaviour of different water basins (or catchment areas) can be a concern affecting the results. According to the author, one of the biggest disadvantages of EROM is that it cannot produce good results to analyse power generation in dry, median and wet periods. Another issue with this model is that it uses a constant energy equivalent, which actually varies accordingly to the water head. [34]
Arvanitidis and Rosing [35] is also the base for other hydropower equivalent studies, for example the learning disaggregation technique developed by Saad and others [36].
In this paper, the single measure potential energy concept is used for the aggregation of the system, which is then disaggregated by training a neural network to determine the storage levels of each reservoir merged into the one-reservoir equivalent. The results highlight that the neural network provides satisfactory results, close to the optimal results. However, this study was developed on a single river system, on La
3Model names: Operational Optimization Model (OPOM) and Equivalent Reservoir Optimization Model (EROM)
Grande River in Canada.
Similarly, this method has been applied in different case studies, for instance in another aggregation-disaggregation approach proposed by Valdés and others in [37]. The concept of formulating a single equivalent and determining the optimal operation for it, and then disaggregating into the operation of the multi-reservoir system, was implemented in the lower Caroni system, a hydroelectric system in Venezuela; which is also one of the largest in Latin America. Once again, the main conclusions drawn point a satisfactory result and lower computational load, however with the compromise of obtaining instead a suboptimal solution. [37].
Söder and Rendelius [38] introduce a new concept of creating a two-station reservoir equivalent, instead of the traditional one-reservoir model. In this study, one of the hydropower plants is set as one the equivalents and then the rest of the power plants are aggregated into an equivalent. The idea is to run an optimisation to find the parameters that minimise the difference in the production plans obtained from the equivalent in comparison with the original system. The results highlight that the computational time is significantly reduce in comparison with the original detailed system and the results obtained are better than those obtained with the one-reservoir model.
Based on the previous study of two-station equivalent, Shayesteh, Amelin and Söder [39] present an upgraded method that formulates the equivalent problem as a bilevel optimisation problem to find the station parameters that minimise the difference within the original system and the equivalent model. Additionally, this study proposes different configurations to interconnect the station equivalents, such as cascade and v-shape configurations with three or up to four stations. This method was implemented in a case study of a Swedish hydropower system of 37 hydropower plants; results indicate that a v-shape of four stations provided the most accurate results for the given case study. In general, all the configurations significantly improve the absolute errors in comparison with the single station configuration.
The concept of multiple-station equivalents with different configurations is further developed by Risberg and Söder in [40], one of the key findings is that more detailed production equivalent curves could potentially provide more accurate results while keeping the computational time lower than the original model.
3 EMPS
The focus of this chapter is the EMPS model, specially the hydropower modelling.
It is worth noting that the literature regarding EMPS is very limited, and most of the internal procedures and calculations are not reachable for the users; as EMPS is a commercial software that is intended to simply produce outputs given certain inputs, without requiring modifications in the core code. This entire chapter is based on the user guides by SINTEF of EMPS Brukerveiledning Samkjøringsmodellen [41]
and EOPS Brukerveiledning Vansimtap [44], and the compendium of the course ELK15, Hydro Power Scheduling [6], from the Norwegian University of Science and Technology.
EMPS is a power market simulator widely used in the Nordics, specially for forecasting and planning in the context of electricity markets. The main reason behind its popularity in the Nordics is because it is well suited for simulating systems with large shares of hydropower generation, combined with thermal generation.
According to SINTEF, some of the tasks that can be solved with EMPS are: [5]
• Forecasting of electricity prices and reservoir operation
• Long term operational scheduling of hydropower
• Maintenance planning (transmission or production)
• Calculation of energy balances (supply, consumption and trade)
• Utilisation of transmission lines and cables
• Analysis of overflow losses, and probability for curtailment
• Analyse interplay between variable generation, hydro and thermal power
• Investment analysis and system development studies
• Calculation of CO2-emissions from power generation
3.1 Model Concept
The model concept of EMPS is based on multiple areas modelling [41], this concept is shown in Figure 3.1; which represents an example system of nine areas. The electrical connection between areas is described by the transmission capacity, and each area has modules describing the demand and generation. The amount of areas in EMPS is not fixed, which means that it provides flexibility to users to add new areas as required. In this project, the idea is to add hydropower generation modules to new areas representing Continental Europe.
Figure 3.1: Example of EMPS model concept, nine areas with respective transmission capacity between areas [41].
The objective of the EMPS model is to minimise the expected cost in the system, taking into consideration the constraints [5] and optimal usage of hydro resources given the uncertainty of future inflows. The time resolution of the model is one
week [41], but it can be divided into different load periods to simulate the variations at a more detailed step [6]. A week can be divided into up to 168 sections (hourly resolution), in each section it is possible to provide information about the load and transmission capabilities [6]. The main elements in each area are: hydropower, thermal power, other generation sources such as wind or solar, consumption and transmission constraints with neighbouring areas.
The EMPS model is composed by two parts, the strategy phase and the simulation phase. In the strategy phase, the marginal value of stored water (which will be referred from now on as the water values method) is calculated for each reservoir using stochastic dynamic programming (SDP), this process will be further explained in Section 3.3. For the water values method, the modules in each area are aggregated to obtain a simplified model representation of the hydropower composed by an equivalent reservoir and an equivalent hydropower station [41]. It is also important to mention that EMPS utilises a heuristic approach to model the interaction between every area [5]. The simulation part consists of, as the name suggests, simulating the system based on the strategy defined in the previous step [6], to define the water allocation according the individual reservoirs [43]. In the simulation part the total costs are minimised in a weekly basis for each climate scenario using a linear problem formulation (LP) [5].
Figure 3.2: Modelling Concept of EMPS [42].
Figure 3.2 summarises the modelling concept of EMPS. The main inputs for the strategy calculation are the details of generation modules and the electricity market specification, in case of the hydropower there is an aggregation procedure done to obtain the water values. The strategy calculation also takes into consideration stochastic weather conditions such as temperature and inflow. Then, the water values result is used as part of the inputs of the simulation phase, which also receives market data and stochastic profiles. The results of the market simulation are also used to calibrate the model.
3.2 Hydropower Modules
Hydropower systems in EMPS are represented by modules describing reservoirs and stations. Each module contains a corresponding inflow series profile (for both regulated and unregulated inflow), as well as parameters describing the watercourse like discharge, bypass and spillage. The hydrological coupling between the modules can also be defined in the standard modules. Figure 3.3 presents the concept of a hydropower module in EMPS.
Figure 3.3: Standard hydropower module in EMPS [41].
The standard concept can easily be applied to large systems, such as the Norwegian
hydropower system, this example is presented in Section 3.6. Different topologies such as cascade and parallel can be represented in EMPS by connecting modules.
The modules can be more detailed depending on the data availability, some parameters are not necessary and if not given, EMPS uses standard values already programmed in the model. More detailed input data yields better results, as the behaviour is represented closer to reality.
3.2.1 Model Elements
The simplest hydropower modules are able to function with very few data. The only parameters that are considered strictly necessary are the reservoir size, the discharge and capacity curve, annual inflow and an inflow profile. EMPS provides a lot of freedom to the user when designing the model layout, there are several ways to model the same system. For example, a module can include a reservoir and a power plant, or it can be modelled in two modules (one for each element). If there is a reservoir without a plant, the water runs to the next level; in the opposite case, the reservoir capacity can be set to zero to represent only a power plant. The same concept applies to regulated and unregulated systems, they can be modelled in the same module or in separate modules. The main elements of a hydropower module are:
• Reservoir
• Power plant
• Inflows
• Topology
• Hydrological coupling
• Restrictions
• Pump
Reservoir: The main parameter to describe is the volume, which must be given in million cubic meter units (M m3), this value must always be specified in every module (it can be set to zero if there is no reservoir). Usually, storage hydropower has larger reservoir volumes and run-of-river plants are modelled with a small reservoir with a volume close to zero. Reservoirs can be buffer or regulation reservoirs, which is specified as an EMPS input; the buffer reservoirs have a low degree of regulation.
There are other parameters that can be given to describe in more detail the reservoirs, for example the reservoir height-volume curve, which is given as a piece- wise linear curve as shown in Figure 3.4. If this curve is given, it is used to correct the production accordingly to the height of the reservoir, and to calculate the height that must be pumped in case of a system with a pump.
[Mm3]
Reservoir content
Lowest allowable
reservoir height Reservoir height
[m.a.s.l]
[MW]
Production
Discharge [m3/s]
Figure 3.4: Example of reservoir curve and power-discharge curve of a hydropower module, adapted from [41].
Power plant: The power plant is described by two elements, the energy equivalent and the power-discharge curve. The general equation to calculate the energy equivalent is presented in equation (2), this expression is obtained directly from equation (1), and it determines how much electricity is produced with a determined amount of water. The energy equivalent can be set to zero if the desired module is only a reservoir.
e = 1
3.6· 106 · ρ · g · h · η (2) Where
e energy equivalent [kW h
m3
]
ρ water density [kg
m3
]
g gravity acceleration [m
s2
]
h plant head [m]
η plant efficiency
3.6· 106 unit correction to obtain kWh
The power-discharge curve creates a relationship between the power output with
a specific discharge, an example curve is presented in Figure 3.4. However, this curve is usually given as a piece-wise linear curve. The rated discharge refers to the discharge that produces the nominal power of the turbine. The rated discharge and the energy equivalent together provide the plant description; it is important to mention that the energy equivalent is fixed in the EMPS model, when in reality this parameter is affected by the efficiency which is at the same time altered by other factors such as discharge and plant head. If the reservoir curve and the backwater level are specified, the output of the power plant is adjusted in EMPS to obtain more realistic results.
Inflow: There are different inflow parameters required by EMPS. The first one consists of a yearly mean regulated inflow, and a yearly mean unregulated inflow;
these correspond to a numeric value in million cubic meter per year (
M m3 year
) . The regulated inflow refers to the inflow that can be stored in a reservoir, this parameter is very important for hydropower storage plants. The unregulated cannot be stored and must be used continuously or if it exceeds the discharge capacity of the plant, it must be spilled; this parameter is vital when defining a run-of-river power plant.
Inflow series must also be provided to characterise the behaviour through the year, and perform the different weather scenarios. These inflow series are desired to cover many years and the data resolution should be daily and in cubic meter per seconds (m3
s
)
. Different inflow series can be linked to the regulated and unregulated inflows, but usually they have the same inflow profile. It is important that the inflows correspond to the area where the power plant or reservoir is located, because this will be determinant for the spring floods; and in general, to define the times with higher or lower inflows.
Topology: The topology of the river system in EMPS is given by simply defining the watercourse. For instance, if module A goes to module B, one defines in module A that the discharge goes to module B. The watercourse is defined for the plant discharge, the spillage and the bypass flow; all these variables could eventually go to different destinations (modules or sea) depending on the real system topology.
Hydrological coupling: Different hydrological coupling configurations can be specified in EMPS. Figure 3.5 presents a summary of these configurations, these differ in the way the reservoirs are connected for example through canals or tunnels, and if there is a controllable hatch between them to limit the capacity and apply different draining strategies.
Code 200 Code 300
Code 100 Code 120
Code 130
Figure 3.5: Different configurations for hydrological coupling between reservoirs in EMPS, adapted from [41].
Restrictions: Specific constrains can be linked to each module, this is particularly useful to describe the real behaviour and respect the operative constraints. The main restrictions that can be implemented are:
• Maximum and minimum reservoir level
• Maximum and minimum discharge
• Maximum and minimum bypass
Modules can have all of the previous constraints simultaneously, and it is important to mention that EMPS deals with them as soft constraints, meaning that the model will take them into consideration and try to satisfy them; if they cannot be satisfied there is a penalisation defined by a penalty function, which increases the objective function in the optimisation problem. The constraints are given by data points corresponding to a curve representing a piece-wise linear variation over time.
Pump: The hydropower modules also support pumps. The pumps are described by the relationship between the lifting height and the maximum amount of water that can be pumped. The main parameters that describe pumps is the pump capacity in megawatt (M W ), the pump flow capacity (
m3 s
), the height in meters
(m) that the water needs to be pumped and the respective reservoirs where the water will be pumped from and into. EMPS also offers the flexibility to model two different kinds of pumping modules, these are:
• Reversible pumps: hydropower stations with turbines that can be reversed to pump water into the upper storage.
• Pump station: a pump that is specifically used to pump water between reservoirs.
It is important to mention that the pumps in EMPS are not included in a formal optimisation process, instead they are characterised and delimited by the heuristics inside the model. Therefore, the pump generation results may not be fully accurate for daily operation reference.
3.2.2 Inputs
Figure 3.6 presents the EMPS dialog window, specifically showing the main parameters of a hydropower module. The EMPS version used is in Norwegian, therefore a translation for each parameter is provided below. The restrictions (17- 22 in the list) improve the results of the model, however, they are optional as they are very specific and it can be difficult to have the exact curves for each power plant;
if these are not provided, EMPS uses default descriptions.
Figure 3.6: EMPS dialog window showing main parameters of a hydropower module.
Optional parameters are marked with *.
1. Reservoir volume [M m3] 2. Energy equivalent [kW h
m3
]
3. Maximal discharge [
m3 s
]
4. Mean head [m]*
5. Elevation of station [m.a.s.l]*
6. Production discharge to [module number]
7. Spillage to [module number]
8. Bypass to [module number]
9. Code for hydrological coupling [0, 200, 300, 100, 120, 130]*
10. Coupling factor [0, 50, 100]*
11. Maximal levelling flow [
m3 s
]*
12. Yearly mean regulated inflow [M m3
year
]
13. Regulated inflow series name 14. Yearly mean unregulated inflow
[M m3 year
]
15. Unregulated inflow series name 16. Station name*
17. Maximal reservoir curve [(week, %)]*
18. Minimal reservoir curve [(week, %)]*
19. Maximal reservoir discharge curve [(
week,ms3 )]*
20. Minimal reservoir discharge curve [(
week,ms3 )]
* 21. Bypass curve [(
week,ms3 )]*
22. Maximal bypass [
m3 s
]
* 23. Reservoir drainage curve[(
m,ms3 )]*
24. Reservoir curve [(m.s.n.l, M m3)]*
25. Draining strategy of reservoir 26. Power-discharge curve[(
m3 s , M W
)]
27. Pump description*
3.3 Water Value Method
The water value method is described before going into detail into the EMPS strategy and simulation phase because the water values is the main element of the strategy part. The water value corresponds to the expected future value of the marginal unit (kWh) of water in the reservoirs [45]. The water value depends on different stochastic elements such as the demand, prices and inflows, which is why these values are calculated with the water value method, a variant of stochastic dynamic programming. When the water values are used as the marginal costs of water the system operation costs are minimised [45].
The water values facilitate the operation of reservoirs, as it can be determined when it is best to use the water; for instance, use water when the prices are higher and store when the prices are lower. Not planning the operation of reservoirs could lead to unnecessary spillage or even rationing by the lack of water to produce electricity. The water values process is done aggregating the model of the hydropower system, which is further explained in Section 3.4.1. The water value method requires simulations to determine the possible consequences of operating with water value.
There are four main factors that affect the water value:
• Reservoir level and generation capacity
• Demand expectation
• Price expectation
• Inflow expectation (including rainfall and snow melting process)
These are stochastic variables which is why it cannot be modelled in a deterministic way. The reservoir levels and inflows of neighbouring areas also affect an specific area, all these are extra sources of uncertainty in the water values calculation. This
