Study of the Impact of Hydropower
Production on Daily Electricity Prices
Pedro Jorge Teixeira Lages Torres
F
ORJ
URYE
VALUATIONMestrado Integrado em Engenharia Eletrotécnica e de Computadores Supervisor: Prof. Doutor Cláudio Domingos Martins Monteiro
O principal objetivo dos operadores não é apenas garantir uma energia e um abastecimento de ele-vada qualidade, mas também conseguir um preço de mercado competitivo para a energia fornecida. Nos últimos anos, com um aumento do número de agentes envolvidos no mercado e, consequente-mente, o aumento da concorrência, a necessidade de maximizar as receitas mantendo um preço competitivo tem vindo a tornar-se cada vez mais uma prioridade. Tanto em Portugal como em Espanha, a presença de geração utilizando fontes renováveis é muito significativa, pelo que é pos-sível afirmar que a disponibilidade de geração renovável tem um impacto significativo nos preços de mercado.
Com a capacidade instalada de energia hidroeléctrica a crescer a cada ano, principalmente devido às qualidades dinâmicas e despachabilidade (permitindo a entrada no mercado) que a car-acterizam e que estão relacionadas com a previsibilidade e capacidade de gestão da fonte (água), torna-se crucial perceber a influência da produção hidroeléctrica sobre os preços da electricidade, com o objectivo de permitir uma melhor previsão dos preços de mercado. Apesar da vasta quan-tidade de bibliografia disponível, nenhuma ou muito pouca aborda este assunto. Esta dissertação incidirá, portanto, sobre a relação entre a produção hidroeléctrica na Península Ibérica, com maior foco em Portugal, e os preços da electricidade no Mercado Ibérico de Electricidade (MIBEL).
The main goal of energy operators is not only to ensure a high-quality energy and a reliable supply, but also to reach a competitive market price for the supplied energy. In the past few years, with more agents entering the market and thus increasing the competition, the need for maximising revenues while keeping a competitive price has become a priority. In both Portugal and Spain, the presence of generation using renewable sources is very significant, so it is possible to affirm that renewable generation availability has a significant impact on market prices.
With installed hydropower capacity growing every year, mostly due to its dynamic qualities and dispatchability (allowing it to enter the market), which are related to the predictability and management ability of the source (water), it becomes crucial to understand the influence that hydropower production has in electricity prices, in order to achieve better EPF. Despite the vast quantity of bibliography available, none or very few of it studies this matter. This Master thesis will then cover the relation between the hydropower production in the Iberian Peninsula, with more focus on Portugal, and the electricity prices in the Iberian Electricity Market (MIBEL).
In the first place, I would like to thank Prof. Cláudio Monteiro, my supervisor, for his dedication, whose knowledge and willingness to go further were fundamental for my understanding of the problem and development of critical thinking, and for always having a word of motivation.
To my family for the restless support and motivation during this semestre, namely my mother Maria, my father Jorge, my sister Marta, and my grandmother Glicínia. Without them, this would have been a lot more difficult.
Finally, to my closest friends who, during the time I have been working on this thesis, as well as during all my academic life, were always there contributing with their amazing company, motivation, support, and who made for all the good moments I will always take with me.
Pedro Torres
Ayrton Senna
1 Introduction 1
1.1 Context and Motivation . . . 1
1.2 Objectives . . . 1
1.3 Structure . . . 2
1.4 Data . . . 2
1.5 Tools and Techniques . . . 2
2 State of The Art 3 2.1 Market Model . . . 3
2.1.1 The Iberian Electricity Market - MIBEL . . . 4
2.2 Hydro Units and Market . . . 10
2.3 Sectorial Pricing Forecasting . . . 13
2.4 Hydropower Generation Forecasting . . . 21
2.5 Hydropower Scheduling . . . 23
2.5.1 Short-Term Scheduling Model in Deregulated Market . . . 23
2.5.2 Long-Term Scheduling Model in Deregulated Market . . . 24
3 Data Treatment and Methodologies 27 4 Results and Discussion 35 4.1 Analysis per Type of HPP . . . 35
4.1.1 Influence of Electricity Prices on Aggregated Hydropower Generation . . 35
4.1.2 Market Value of the Different Types of HPP . . . 42
4.1.3 Analysis Based on Daily Average Values for Every Day of the Data Spectrum 43 4.1.4 Analysis Based on Monthly Average Values . . . 49
4.2 Analysis per Balance Area . . . 52
4.2.1 Market Value of the Different Balance Areas . . . 53
4.2.2 Analysis Based on Daily Average Values for Every Day of the Data Spectrum 54 4.2.3 Analysis Based on Monthly Average Values . . . 59
5 Conclusions and Future Work 61 5.1 Conclusions . . . 61
5.2 Future Work . . . 62
A 63 A.1 Run-of-River Hydropower Plants . . . 63
A.2 Special Generation Regime Hydropower Plants . . . 65
A.3 ACAVADO Balance Area . . . 73
A.4 ADOURO Balance Area . . . 77
A.5 AGUADIA Balance Area . . . 81
A.6 ALIMA Balance Area . . . 91
A.7 AMONDEG Balance Area . . . 101
A.8 ATEJZEZ Balance Area . . . 111
2.1 Aggregated Curves in Market Clearing Process in Day-ahead Market [1] . . . 6
2.2 Time framework of the EMPF model for price forecasting [2] . . . 16
2.3 Time framework of the REMPE model for price estimation [2] . . . 17
2.4 Time framework of the PPFM model [3] . . . 20
3.1 Diagram of Total Consumption for 21st August 2018 . . . 28
3.2 Diagram of Hydro Generation for 21st August 2018 . . . 28
3.3 Diagram of Special Regime Hydro Generation for 21st August 2018 . . . 28
3.4 Excerpt of the Processed Data Set Related to Reservoir Type of HPP . . . 29
4.1 Hydro Generation (MW) and Price (e/MWh) over Time for Reservoir HPP . . . 36
4.2 Hydro Generation (MW) and Price (e/MWh) over Time for Run-of-River HPP . 37 4.3 Hydro Generation (MW) and Price (e/MWh) over Time for Special Regime Gen-eration HPP . . . 38
4.4 Monthly Data Analysis for Reservoir HPP . . . 39
4.5 Weekly Data Analysis for Reservoir HPP . . . 39
4.6 Hourly Data Analysis for Reservoir HPP . . . 40
4.7 S(PR/HCF) per Type of HPP . . . 41
4.8 Sensitivity vs. Dispatchability Index per Type of HPP . . . 42
4.9 Comparison of Captured Value per Type of HPP and the Average Market Price . 42 4.10 Captured Value vs. Dispatchability Index per Type of HPP . . . 43
4.11 Influence of Hydro Capacity Factor (HCF) on the Market Price (MP) . . . 44
4.12 DI vs. HCF per Type of HPP . . . 44
4.13 Influence of Hydro Capacity Factor (HCF) on the Dispatchability Index (DI) . . . 45
4.14 S(HCF/PR) vs. HCF per Type of HPP . . . 45
4.15 Influence of Hydro Capacity Factor (HCF) on the Sensitivity (S(HCF/PR)) . . . . 46
4.16 Influence of the Dispatchability Index (DI) on the Sensitivity (S(HCF/PR)) . . . . 46
4.17 PCV vs. HCF per Type of HPP . . . 47
4.18 Influence of the Hydro Capacity Factor (HCF) on the Percentage Captured Value (PCV) . . . 47
4.19 PCV vs. DI per Type of HPP . . . 47
4.20 Influence of the Dispatchability Index (DI) on the Percentage Captured Value (PCV) 48 4.21 PCV vs. S(HCF/PR) per Type of HPP . . . 48
4.22 Influence of the Sensitivity (S(HCF/PR)) on the Percentage Captured Value (PCV) 49 4.23 Relation Between the Hydro Capacity Factor (HCF) and the Market Price (MP) per Month . . . 49
4.24 Relation Between the Hydro Capacity Factor (HCF) and the Dispatchability Index (DI) per Month . . . 50
4.25 Relation Between the Hydro Capacity Factor (HCF) and the Sensitivity (S(HCF/PR))
per Month . . . 50
4.26 Relation Between the Dispatchability Index (DI) and the Sensitivity (S(HCF/PR)) per Month . . . 51
4.27 Relation Between the Dispatchability Index (DI) and the Percentage Captured Value (PCV) per Month . . . 51
4.28 Relation Between the Sensitivity (S(HCF/PR)) and the Percentage Captured Value (PCV) per Month . . . 52
4.29 Comparison of Captured Value per Balance Area and the Average Market Price . 53 4.30 Captured Value vs. Dispatchability Index per Balance Area . . . 54
4.31 Influence of Hydro Capacity Factor (HCF) on the Market Price (MP) . . . 55
4.32 DI vs. HCF per Balance Area . . . 55
4.33 Influence of Hydro Capacity Factor (HCF) on the Dispatchability Index (DI) . . . 55
4.34 S(HCF/PR) vs. HCF per Balance Area . . . 56
4.35 Influence of Hydro Capacity Factor (HCF) on the Sensitivity (S(HCF/PR)) . . . . 56
4.36 S(HCF/PR) vs. HCF per Balance Area . . . 57
4.37 Influence of the Dispatchability Index (DI) on the Sensitivity (S(HCF/PR)) . . . . 57
4.38 PCV vs. HCF per Balance Area . . . 57
4.39 Influence of the Hydro Capacity Factor (HCF) on the Percentage Captured Value (PCV) . . . 58
4.40 PCV vs. DI per Balance Area . . . 58
4.41 Influence of the Dispatchability Index (DI) on the Percentage Captured Value (PCV) 58 4.42 PCV vs. S(HCF/PR) per Balance Area . . . 59
4.43 Influence of the Sensitivity (S(HCF/PR)) on the Percentage Captured Value (PCV) 59 4.44 Relation Between the Dispatchability Index (DI) and the Sensitivity (S(HCF/PR)) per Month . . . 60
4.45 Relation Between the Dispatchability Index (DI) and the Percentage Captured Value (PCV) per Month . . . 60
4.46 Relation Between the Sensitivity (S(HCF/PR)) and the Percentage Captured Value (PCV) per Month . . . 60
A.1 Hydro Generation and Market Price per Month for Run-of-River . . . 63
A.2 Hydro Generation and Market Price per Weekday for Run-of-River . . . 64
A.3 Hydro Generation and Market Price per Hour for Run-of-River . . . 64
A.4 Hydro Generation and Market Price per Month for Special Regime Generation . 65 A.5 Hydro Generation and Market Price per Weekday for Special Regime Generation 65 A.6 Hydro Generation and Market Price per Hour for Special Regime Generation . . 66
A.7 Influence of the Hydro Capacity Factor (HCF) on Market Price for Special Regime Generation . . . 66
A.8 Influence of the Hydro Capacity Factor (HCF) on the Dispatchability Index (DI) for Special Regime Generation . . . 67
A.9 Influence of the Hydro Capacity Factor (HCF) on Sensitivity of Production to Price (S(HCF/PR)) for Special Regime Generation . . . 67
A.10 Influence of the Dispatchability Index (DI) on Sensitivity of Production to Price (S(HCF/PR)) for Special Regime Generation . . . 68
A.11 Influence of the Hydro Capacity Factor (HCF) on Percentage Captured Value (PCV) for Special Regime Generation . . . 68
A.12 Influence of the Dispatchability Index (DI) on Percentage Captured Value (PCV) for Special Regime Generation . . . 69
A.13 Influence of Sensitivity of Production to Price (S(HCF/PR)) on Percentage Cap-tured Value (PCV) for Special Regime Generation . . . 69
A.14 Influence of the Hydro Capacity Factor (HCF) on Market Price for Special Regime Generation per Month . . . 70
A.15 Influence of the Hydro Capacity Factor (HCF) on the Dispatchability Index (DI) per Month for Special Regime Generation . . . 70
A.16 Influence of the Hydro Capacity Factor (HCF) on Sensitivity of Production to Price (S(HCF/PR)) per Month for Special Regime Generation . . . 71
A.17 Influence of the Dispatchability Index (DI) on Sensitivity of Production to Price (S(HCF/PR)) per Month for Special Regime Generation . . . 71
A.18 Influence of the Hydro Capacity Factor (HCF) on Percentage Captured Value (PCV) per Month for Special Regime Generation . . . 72
A.19 Influence of the Dispatchability Index (DI) on Percentage Captured Value (PCV) per Month for Special Regime Generation . . . 72
A.20 Influence of Sensitivity of Production to Price (S(HCF/PR)) on Percentage Cap-tured Value (PCV) per Month for Special Regime Generation . . . 73
A.21 Hydro Generation and Market Price for ACAVADO Balance Area . . . 74
A.22 Hydro Generation and Market Price per Month for ACAVADO Balance Area . . 75
A.23 Hydro Generation and Market Price per Weekday for ACAVADO Balance Area . 75
A.24 Hydro Generation and Market Price per Hour for ACAVADO Balance Area . . . 76
A.25 Influence of the Hydro Capacity Factor (HCF) on Market Price per Month for ACAVADO Balance Area . . . 76
A.26 Influence of the Hydro Capacity Factor (HCF) on the Dispatchability Index (DI) per Month for ACAVADO Balance Area . . . 77
A.27 Influence of the Hydro Capacity Factor (HCF) on Sensitivity of Production to Price (S(HCF/PR)) per Month for ACAVADO Balance Area . . . 77
A.28 Hydro Generation and Market Price for ADOURO Balance Area . . . 78
A.29 Hydro Generation and Market Price per Month for ADOURO Balance Area . . . 79
A.30 Hydro Generation and Market Price per Weekday for ADOURO Balance Area . . 79
A.31 Hydro Generation and Market Price per Hour for ADOURO Balance Area . . . . 80
A.32 Influence of the Hydro Capacity Factor (HCF) on Market Price per Month for ADOURO Balance Area . . . 80
A.33 Influence of the Hydro Capacity Factor (HCF) on the Dispatchability Index (DI) per Month for ADOURO Balance Area . . . 81
A.34 Influence of the Hydro Capacity Factor (HCF) on Sensitivity of Production to Price (S(HCF/PR)) per Month for ADOURO Balance Area . . . 81
A.35 Hydro Generation and Market Price for AGUADIA Balance Area . . . 82
A.36 Hydro Generation and Market Price per Month for AGUADIA Balance Area . . 83
A.37 Hydro Generation and Market Price per Weekday for AGUADIA Balance Area . 83
A.38 Hydro Generation and Market Price per Hour for AGUADIA Balance Area . . . 84
A.39 Influence of the Hydro Capacity Factor (HCF) on Market Price for AGUADIA Balance Area . . . 84
A.40 Influence of the Hydro Capacity Factor (HCF) on the Dispatchability Index (DI) for AGUADIA Balance Area . . . 85
A.41 Influence of the Hydro Capacity Factor (HCF) on Sensitivity of Production to Price (S(HCF/PR)) for AGUADIA Balance Area . . . 85
A.42 Influence of the Dispatchability Index (DI) on Sensitivity of Production to Price (S(HCF/PR)) for AGUADIA Balance Area . . . 86
A.43 Influence of the Hydro Capacity Factor (HCF) on Percentage Captured Value (PCV) for AGUADIA Balance Area . . . 86
A.44 Influence of the Dispatchability Index (DI) on Percentage Captured Value (PCV) for AGUADIA Balance Area . . . 87
A.45 Influence of Sensitivity of Production to Price (S(HCF/PR)) on Percentage Cap-tured Value (PCV) for AGUADIA Balance Area . . . 87
A.46 Influence of the Hydro Capacity Factor (HCF) on Market Price per Month for AGUADIA Balance Area . . . 88
A.47 Influence of the Hydro Capacity Factor (HCF) on the Dispatchability Index (DI) per Month for AGUADIA Balance Area . . . 88
A.48 Influence of the Hydro Capacity Factor (HCF) on Sensitivity of Production to Price (S(HCF/PR)) per Month for AGUADIA Balance Area . . . 89
A.49 Influence of the Dispatchability Index (DI) on Sensitivity of Production to Price (S(HCF/PR)) per Month for AGUADIA Balance Area . . . 89
A.50 Influence of the Hydro Capacity Factor (HCF) on Percentage Captured Value (PCV) per Month for AGUADIA Balance Area . . . 90
A.51 Influence of the Dispatchability Index (DI) on Percentage Captured Value (PCV) per Month for AGUADIA Balance Area . . . 90
A.52 Influence of Sensitivity of Production to Price (S(HCF/PR)) on Percentage Cap-tured Value (PCV) per Month for AGUADIA Balance Area . . . 91
A.53 Hydro Generation and Market Price for ALIMA Balance Area . . . 92
A.54 Hydro Generation and Market Price per Month for ALIMA Balance Area . . . . 93
A.55 Hydro Generation and Market Price per Weekday for ALIMA Balance Area . . . 93
A.56 Hydro Generation and Market Price per Hour for ALIMA Balance Area . . . 94
A.57 Influence of the Hydro Capacity Factor (HCF) on Market Price for ALIMA Bal-ance Area . . . 94
A.58 Influence of the Hydro Capacity Factor (HCF) on the Dispatchability Index (DI) for ALIMA Balance Area . . . 95
A.59 Influence of the Hydro Capacity Factor (HCF) on Sensitivity of Production to Price (S(HCF/PR)) for ALIMA Balance Area . . . 95
A.60 Influence of the Dispatchability Index (DI) on Sensitivity of Production to Price (S(HCF/PR)) for ALIMA Balance Area . . . 96
A.61 Influence of the Hydro Capacity Factor (HCF) on Percentage Captured Value (PCV) for ALIMA Balance Area . . . 96
A.62 Influence of the Dispatchability Index (DI) on Percentage Captured Value (PCV) for ALIMA Balance Area . . . 97
A.63 Influence of Sensitivity of Production to Price (S(HCF/PR)) on Percentage Cap-tured Value (PCV) for ALIMA Balance Area . . . 97
A.64 Influence of the Hydro Capacity Factor (HCF) on Market Price per Month for ALIMA Balance Area . . . 98
A.65 Influence of the Hydro Capacity Factor (HCF) on the Dispatchability Index (DI) per Month for ALIMA Balance Area . . . 98
A.66 Influence of the Hydro Capacity Factor (HCF) on Sensitivity of Production to Price (S(HCF/PR)) per Month for ALIMA Balance Area . . . 99
A.67 Influence of the Dispatchability Index (DI) on Sensitivity of Production to Price (S(HCF/PR)) per Month for ALIMA Balance Area . . . 99
A.68 Influence of the Hydro Capacity Factor (HCF) on Percentage Captured Value (PCV) per Month for ALIMA Balance Area . . . 100
A.69 Influence of the Dispatchability Index (DI) on Percentage Captured Value (PCV) per Month for ALIMA Balance Area . . . 100
A.70 Influence of Sensitivity of Production to Price (S(HCF/PR)) on Percentage Cap-tured Value (PCV) per Month for ALIMA Balance Area . . . 101
A.71 Hydro Generation and Market Price for AMONDEG Balance Area . . . 102
A.72 Hydro Generation and Market Price per Month for AMONDEG Balance Area . . 103
A.73 Hydro Generation and Market Price per Weekday for AMONDEG Balance Area 103
A.74 Hydro Generation and Market Price per Hour for AMONDEG Balance Area . . . 104
A.75 Influence of the Hydro Capacity Factor (HCF) on Market Price for AMONDEG Balance Area . . . 104
A.76 Influence of the Hydro Capacity Factor (HCF) on the Dispatchability Index (DI) for AMONDEG Balance Area . . . 105
A.77 Influence of the Hydro Capacity Factor (HCF) on Sensitivity of Production to Price (S(HCF/PR)) for AMONDEG Balance Area . . . 105
A.78 Influence of the Dispatchability Index (DI) on Sensitivity of Production to Price (S(HCF/PR)) for AMONDEG Balance Area . . . 106
A.79 Influence of the Hydro Capacity Factor (HCF) on Percentage Captured Value (PCV) for AMONDEG Balance Area . . . 106
A.80 Influence of the Dispatchability Index (DI) on Percentage Captured Value (PCV) for AMONDEG Balance Area . . . 107
A.81 Influence of Sensitivity of Production to Price (S(HCF/PR)) on Percentage Cap-tured Value (PCV) for AMONDEG Balance Area . . . 107
A.82 Influence of the Hydro Capacity Factor (HCF) on Market Price per Month for AMONDEG Balance Area . . . 108
A.83 Influence of the Hydro Capacity Factor (HCF) on the Dispatchability Index (DI) per Month for AMONDEG Balance Area . . . 108
A.84 Influence of the Hydro Capacity Factor (HCF) on Sensitivity of Production to Price (S(HCF/PR)) per Month for AMONDEG Balance Area . . . 109
A.85 Influence of the Dispatchability Index (DI) on Sensitivity of Production to Price (S(HCF/PR)) per Month for AMONDEG Balance Area . . . 109
A.86 Influence of the Hydro Capacity Factor (HCF) on Percentage Captured Value (PCV) per Month for AMONDEG Balance Area . . . 110
A.87 Influence of the Dispatchability Index (DI) on Percentage Captured Value (PCV) per Month for AMONDEG Balance Area . . . 110
A.88 Influence of Sensitivity of Production to Price (S(HCF/PR)) on Percentage Cap-tured Value (PCV) per Month for AMONDEG Balance Area . . . 111
A.89 Hydro Generation and Market Price for ATEJZEZ Balance Area . . . 112
A.90 Hydro Generation and Market Price per Month for ATEJZEZ Balance Area . . . 113
A.91 Hydro Generation and Market Price per Weekday for ATEJZEZ Balance Area . . 113
A.92 Hydro Generation and Market Price per Hour for ATEJZEZ Balance Area . . . . 114
A.93 Influence of the Hydro Capacity Factor (HCF) on Market Price for ATEJZEZ Bal-ance Area . . . 114
A.94 Influence of the Hydro Capacity Factor (HCF) on the Dispatchability Index (DI) for ATEJZEZ Balance Area . . . 115
A.95 Influence of the Hydro Capacity Factor (HCF) on Sensitivity of Production to Price (S(HCF/PR)) for ATEJZEZ Balance Area . . . 115
A.96 Influence of the Dispatchability Index (DI) on Sensitivity of Production to Price (S(HCF/PR)) for ATEJZEZ Balance Area . . . 116
A.97 Influence of the Hydro Capacity Factor (HCF) on Percentage Captured Value (PCV) for ATEJZEZ Balance Area . . . 116
A.98 Influence of the Dispatchability Index (DI) on Percentage Captured Value (PCV) for ATEJZEZ Balance Area . . . 117
A.99 Influence of Sensitivity of Production to Price (S(HCF/PR)) on Percentage Cap-tured Value (PCV) for ATEJZEZ Balance Area . . . 117
A.100Influence of the Hydro Capacity Factor (HCF) on Market Price per Month for ATEJZEZ Balance Area . . . 118
A.101Influence of the Hydro Capacity Factor (HCF) on the Dispatchability Index (DI) per Month for ATEJZEZ Balance Area . . . 118
A.102Influence of the Hydro Capacity Factor (HCF) on Sensitivity of Production to Price (S(HCF/PR)) per Month for ATEJZEZ Balance Area . . . 119
A.103Influence of the Dispatchability Index (DI) on Sensitivity of Production to Price (S(HCF/PR)) per Month for ATEJZEZ Balance Area . . . 119
A.104Influence of the Hydro Capacity Factor (HCF) on Percentage Captured Value (PCV) per Month for ATEJZEZ Balance Area . . . 120
A.105Influence of the Dispatchability Index (DI) on Percentage Captured Value (PCV) per Month for ATEJZEZ Balance Area . . . 120
A.106Influence of Sensitivity of Production to Price (S(HCF/PR)) on Percentage Cap-tured Value (PCV) per Month for ATEJZEZ Balance Area . . . 121
2.1 MIBEL in Iberian Peninsula [4] . . . 5
2.2 MIBEL in Portugal [5] . . . 5
2.3 Percentage of MIBEL PT Energy Transactions in the MIBEL . . . 5
2.4 Intraday Market Sessions . . . 8
2.5 Hydropower Units by Hydrographic Basin in Continental Portugal [6] [7] . . . . 12
2.6 Installed Capacity in Continental Portugal (MW) [8] [9] [10] [11] . . . 12
2.7 Power Generation and Demand in Portugal (GWh) [8] [9] [10] [11] . . . 12
2.8 Explanatory variables of the explanatory model for price forecast (EMPF) model [2]. 18 2.9 Explanatory variables of the reference explanatory model for price estimations (REMPE) model [2]. . . 19
2.10 Explanatory variables of the PPFM model [3]. . . 21
3.1 Aggregated Hydropower Generation by Type of HPP . . . 29
3.2 Hydropower Plants and Respective Installed Capacity and Type by Balance Area 30 3.2 Hydropower Plants and Respective Installed Capacity and Type by Balance Area 31 3.3 Aggregated Hydropower Generation by Balance Area . . . 31
4.1 Sensitivity and Dispatchability Indexes per Type of HPP . . . 40
4.2 Sensitivity and Dispatchability Indexes per Balance Area . . . 53
Abbreviations and Symbols
AMP Average Market PriceBIPCIC Basic Intraday Program of Continuous Incremental Cassation BIPIC Basic Intraday Program of Incremental Cassation
CFS Continuous Final Schedule CV Captured Value
DI Dispatchability Index
DSMPF Daily Session Model for Price Forecasts DVDS Definitive Viable Day-ahead Schedule EMAR Electricity Market Activity Rules EMPF Explanatory Model for Price Forecast EPF Electricity Price Forecasting
ERSE Energy Services Regulatory Authority FiT Feed-in Tariffs
FOS Final Operational Schedule FS Final Schedule
GA Genetic Algorithm
GFS Global Forecasting System HCF Hydro Capacity Factor
HPI Hydroelectric Producibility Index HPP Hydro Power Plant
IPH Índice de Produtibilidade Hidroeléctrica ISMPF Intraday Session Model for Price Forecasts ISO Independent System Operator
MCP Market Clearing Price MIBEL Iberian Electricity Market MO Market Operator
NEMO Nominated Electricity Market Operator NW-KDE Nadaraya-Watson Kernel Density Estimator NWP Numerical Weather Prediction
P Power
P2P Peer-to-Peer
PCV Percentage Captured Value PHOF Programa Horário Operativo Final PPA Power Purchasing Agreement
PPFM Probabilistic Price Forecasting Model
REMPE Reference Explanatory Model for Price Estimation RMPE Reference Model for Price Estimation
ROR Run-Of-River S Sensitivity
SHP Small Hydro Power SRG Special Regime Generation TSO Transmission System Operator
Introduction
1.1
Context and Motivation
The main goal of energy operators is not only to ensure a high quality energy and a reliable supply, but also to reach a competitive market price for the supplied energy. In the past few years, with more agents entering the market and thus increasing the competition, the need for maximising revenues while keeping a competitive price has become a priority. In both Portugal and Spain, the presence of generation using renewable sources is very significant, so it is possible to affirm that renewable generation availability has a significant impact on market prices.
A lot of research has been made in the last decade to allow market and system operators to have a better knowledge of their businesses in order to maximise profits and improve the service. From electricity price forecasting (EPF) to power production forecasting, various models have been developed with those objectives in sight, some of which being approached in this thesis.
With hydropower production installed capacity growing every year, mostly due its dynamic qualities and dispatchability (allowing it to enter the market), which is related to the predictability and management ability of the source (water), it becomes crucial to understand the influence that hydropower production has in electricity prices, in order to achieve better EPF. Despite the vast quantity of bibliography available, none or very few of it studies this matter.
This Master thesis will then cover the relation between the hydropower production in the Iberian Peninsula and the electricity prices in the Iberian Electricity Market (MIBEL).
1.2
Objectives
In order to establish reliable models of relation between hydropower production and market prices, it is mandatory to take into account some variables of major importance, such as the seasonal variability of the source, the type of hydro unit and the region or basin where it is planted, as well as relevant chronological variables, such as values for prices and production.
Taking into account the scope of this work, the objectives are: 1
• Study the seasonal variability of the influence of hydropower production on spot market prices;
• Validate the previous objective considering prices variation over the years; • Analyse the relation between prices and hydraulicity and water levels; • Analyse the relation between prices and weekly production;
• Analyse the relation between prices and hourly production;
• Analyse the relation between prices and hydropower production for different levels of an-other types of production.
1.3
Structure
This thesis is divided in five chapters. Chapter 1 presents an overview on the context and the motives that led to the elaboration of this work, as well as the objectives proposed to achieve by the time it is finished. The Introduction will also make reference to the data which the method-ology was based on and the tools used for the treatment and analysis of such data. Chapter 2
includes the State of The Art, making reference to methodologies and literature review related to market models, namely within the MIBEL, the relation between hydropower units and the market, market price and hydro generation forecasting, and hydropower scheduling. Chapter3covers the methodologies used for data treatment and presents the indexes developed in order to achieve the desired results, which are presented and analysed on chapter4. Finally, chapter5includes the final conclusions and a suggestion for future works.
1.4
Data
For this work, it were used two detailed sets of data provided by REN - Redes Energéticas Na-cionais. One consists on hourly data of power generation in Portugal and MIBEL market prices, from January 2016 to July 2018, for a total of 22442 hours, for each of the sources available, being the hydro sources divided by type (reservoir, run-of-river, and special regime generation). The second one consists on the same type of data, but for each one of the different balance areas (ACAVADO, ADOURO, AGUADIA, ALIMA, AMONDEG, and ATEJZEZ).
1.5
Tools and Techniques
The data was organised and manipulated with the software Microsoft Excel, which revealed itself to be very useful and with full capabilities of processing the data the way it was idealised in the first place.
State of The Art
This chapter is focused on the detailed presentation of the various factors to be taken into account in the study of the impact of hydropower generation on electricity prices. On section 2.1, an overview of the different energy trading processes and relevant values related to those processes. On section2.2, the relation between HPPs and their position and role in the market, with an ap-proach on the different types of HPPs and respective characteristics, as well as values of installed capacity and generation for the different hydrographic basins. On sections2.3 and2.4, electric-ity prices and hydropower generation forecasting models are overviewed. Lastly, section2.5 is focused on the scheduling methods for hydropower generation.
2.1
Market Model
Electricity is, after all, a commodity like many others. Therefore, as with happens with other commodities, electricity must be able to be traded between sellers, such as generation companies, and buyers, such as distribution companies, retailers, and eligible consumers.
There are different ways of trading electricity: • Day-ahead Market
• Intraday Market • Bilateral Contracts
• Peer-to-Peer (P2P) Trading
Both day-ahead and intraday markets are integrant parts of the Pool model. The Pool is an entity that establishes a relation between supply and demand entities with the objective of running a centralised dispatch according to bids placed by both sides.
The day-ahead market has the purpose of handling the sale and purchase bids that will lead to electricity transactions for the following day. In this model, the generators place offers for the supply of electricity in the form of price-quantity pairs, creating an aggregated supply curve.
In the other hand, the buyers place price-quantity bids, forming a demand curve [12]. Then, the Market Operator initiates and supervises a market clearing process, where the bids from both sides are sorted, prioritising the generators with the lowest marginal cost disregarding the constraints imposed by the network, and both curves are submitted to a matching process, which functions as an auction. When the two curves match, the Market Clearing Price (MCP) is reached. If there is a congestion, more expensive generators will replace some of the cheapest ones on dispatch, with an impact on the MCP. The buyers then pay and the generator are paid the MCP [12] [13] [14].
Once the day-ahead market ends, the intraday market is held, allowing market participants to place bids for electricity transactions in order to address the adjustments to the Definitive Viable Day-ahead Schedule (DVDS), whose programming basis is the result of the day-ahead market, according to updated forecasts [15].
As alternative to the Pool market, the participants can also establish Bilateral Contracts in order to fulfil their energy needs. With these contracts, a buyer and a seller agree with each other the payment and the delivery, respectively, of electric power. This market model has some advantages over the spot market, namely the customised nature of the transactions, less volatile prices (the MCP is usually volatile, especially in countries where hydro and other renewable energies are present in large scale, as in MIBEL [16]), and the support they give to renewable energy. Given the fact that renewable generation is strongly dependent on weather conditions and has high marginal costs, investors need to guarantee future revenues. With bilateral contracts, they ensure the supply and, consequently, the revenue, regardless the market situation [12] [13].
At last, the Peer-to-Peer trading concept is strongly related with deregulated markets, micro-grids, and distributed generation. Even though P2P trading is still growing, in an early phase, it has a great potential, as the penetration of distributed generation is also increasing. This model is characterised for the trading of electricity between end consumers connected to the same grid, without intermediation, usually by ICT-based online services. They become prosumers (some works also refer to P2P as Prosumer-to-Prosumer). Each node of the distribution grid is assumed to play a role in both generation and consumption of electricity, as the prosumers have the ability to make their overproduced energy available to other prosumers [17] [18].
2.1.1 The Iberian Electricity Market - MIBEL
In the Iberian Peninsula (mainland Portugal and Spain), most of the electricity is transacted within a spot market (MIBEL - Iberian Electricity Market), with day-ahead and intraday markets, and a share is negotiated through bilateral contracts.
According to [19], the existence of bilateral contracts has a noticeable impact on the gradient of the market’s bidding curve, being that the price to pay for a certain quantity of energy will be higher if bilateral contracts are established.
Table2.1presents an overview on Average Market Prices, transactions of electricity in both day-ahead and intraday markets, and the share of energy negotiated through bilateral contracts, in the Iberian Peninsula, in the period between 2013 and 2017, according to the last data provided by REN for the entire MIBEL. In table 2.2, the same kind of values are shown, for the period
between 2015 and 2018, but concerning only to the Portuguese side, in order to better understand the difference, in terms of transacted energy, between the two countries, as well as the impact that MIBEL PT has in the whole Iberian Market. In table2.3, percentage values of transacted energy in MIBEL PT relatively to the entire MIBEL are presented [4] [20].
Table 2.1: MIBEL in Iberian Peninsula [4]
2013 2014 2015 2016 2017 Average Market Price PT (e/MWh) 43,6 41,9 50,4 39,4 52,5 Average Market Price ES (e/MWh) 44,3 42,1 50,3 39,7 52,2
Acquired Energy (TWh) 273 258 257 266 278
Day-ahead Market 235 223 226 233 243
Intraday Market 39 35 32 32 36
Bilateral Contracts (GWh) 13 873 17 328 19 414 20 522 20 932
Table 2.2: MIBEL in Portugal [5]
2014 2015 2016 2017 2018 Average Market Price PT (e/MWh) 41,9 50,4 39,4 52,5 57,5 Acquired Energy PT (GWh) 53 191 53 205 53 512 54 605 56 008 Day-ahead Market 49 498 49 655 49 501 50 640 51 537
Intraday Market 3 693 3 550 4 011 3 965 4 166
Table 2.3: Percentage of MIBEL PT Energy Transactions in the MIBEL 2014 2015 2016 2017
Acquired Energy PT (%) 20,6 20,7 20,1 19,6 Day-ahead Market (%) 22,2 22,0 21,2 20,8 Intraday Market (%) 10,6 11,1 12,5 11,0
As it is possible to see in table2.3, the Portuguese market represents approximately 20% of the energy transactions in the MIBEL.
2.1.1.1 Day-ahead Market
The day-ahead market is the main electricity trading market and it is managed by OMIE (Spanish Iberian Market Operator). As most of the European day-ahead markets, it is a marginal pricing market in which the price-quantity pair for each hour is set according to the point of equilibrium between supply and demand, as illustrated on fig. 2.1. On a daily basis, bids for the purchase and sale of electricity are received for the next day up until 12 noon (Spanish time), which is the deadline for the submission of bids. These are then processed using a European algorithm called
EUPHEMIA, and the energy that will be produced and purchased in each one of the hours of the following day is then publicly announced by OMIE [15].
Figure 2.1: Aggregated Curves in Market Clearing Process in Day-ahead Market [1]
Bids from both sale and purchase can be made considering between 1 and 25 hourly energy blocks. In the case of sales, the bid price increases with the block number; in the case of purchases, the bid price decreases with the block number.
Electricity sale bids presented by sellers to the MO may be simple or incorporate complex conditions in terms of their content. Sellers for each hour and generation unit present simple bids, indicating a price and an amount of power. Complex bids are those that incorporate complex sale terms and conditions and those which, in compliance with the simple bid requirements, also include one or some the following technical or economic conditions:
• Indivisibility • Load gradients • Minimum income • Scheduled stop
The indivisibility condition enables a minimum operating value to be fixed in the first block of each hour. This value may only be divided by applying distribution rules if the price is other than zero.
The load gradient enables the maximum difference between the energy in one hour and the energy in the following hour of the generation unit to be established, limiting maximum matchable energy by matching the previous hour and the following hour, in order to avoid sudden changes in the generation units that the latter are unable to follow due to technical restrictions.
The condition of minimum income allows bids to be presented in all hours, although, it is known beforehand that the generation unit will not participate in the daily matching result if its total generation for the day does not exceed a minimum income level, in euros, plus a variable remuneration, also in euros, for every MWh to be supplied.
The condition of scheduled stop enables generation units that have been withdrawn from the matching process because they fail to comply the stipulated minimum income condition to carry out a scheduled stop for a maximum period of three hours, avoiding stoppages in their schedules from the final hour of the previous day to zero in the first hour of the following day by accepting the first slot of first three hours of their bids as simple bids, the only condition being that energy offered in bids must drop in each hour.
The purpose of the Euphemia matching algorithm is to optimise what is referred to as welfare which corresponds to the sum for the combined total of all the hourly periods in the programming horizon of the gain from the purchase bids, plus the gain from the sale bids, plus the congestion charge. Gain in terms of purchase bids is understood to be the difference between the price of the matched purchase bid and the marginal price received, and the gain in terms of sale bids is understood to be the difference between the resulting marginal price and the price of the matched sale bid.
The Euphemia algorithm considers stepwise aggregate curves, which correspond to the curves for which the start price for the acceptance of a block of energy and the full acceptance price for said block of energy coincide, and to interpolated aggregate curves, which are those curves for which the start price for the acceptance of a block of energy and the full acceptance price for said block of energy differ in at least one minimum step between bid prices. For the processing of both types of curves, the Euphemia algorithm performs the matching process with the accuracy of the price values and energy values exceeding the ceiling of decimals specified for the submission of bids. Once the matching process has been completed, the figures for energies and prices are rounded off according to the accuracy specified in each market. The accuracy established for the Iberian market is two decimal points for prices, stated ine/MWh, and one decimal point for energies, stated in MWh.
The Euphemia algorithm considers each one’s specific complex conditions or block condi-tions, with the conditions for bids in the Iberian market being as stipulated in the operating rules of the corresponding day-ahead and intraday markets.
The result of the Euphemia algorithm is limited to the interchange established in each market between generation zones. Accordingly, the net flow bidding zones (flow between Spain and Portugal, between Spain and France, and between Spain and Morocco), will be restricted to the capacity available for the market as notified by the corresponding system operators.
The Euphemia algorithm treats all simple bids as a single bid, being the combined total of all the simple bids in the generation zone. Once the matching process has been completed, the MO shall proceed to allocate the matched and unmatched blocks of the simple bids in each generation zone.
Once Euphemia’s matching process has been made, allocation will be made of the values of the matched and unmatched energy blocks of all the bids that have declared any one of the complex conditions, excluding indivisibility condition, as well as the values of the matched and unmatched energy blocks for the sum of bids that have not declared a complex condition or have only stated the condition of indivisibility [21].
2.1.1.2 Intraday Market
The intraday auction market is currently structured in six sessions with different programming horizons for each session and manages the price areas of Portugal and Spain, and the free capacity of the interconnections: Spain-Portugal, Spain-Morocco and Spain-Andorra. The schedule result-ing from each session of the intraday auction market is the Basic Intraday Program of Incremental Cassation (BIPIC). The system operator, based on this program, publishes the resulting program, the Final Schedule (FS).
The six sessions of the intraday auction market are established in the Electricity Market Ac-tivity Rules (EMAR) and are organised as shown in table2.4.
Table 2.4: Intraday Market Sessions
1st Session 2nd Session 3rd Session 4th Session 5th Session 6th Session Auction Opening Time 17:00 21:00 01:00 04:00 08:00 12:00 Auction Closing Time 18:50 21:50 01:50 04:50 08:50 12:50 Matching Process 18:50 21:50 01:50 04:50 08:50 12:50 Publication of Results 18:57 21:57 01:57 04:57 08:57 12:57 TSO Publication (FS) 19:20 22:20 02:20 05:20 09:20 13:20 Schedule Horizon (Time Periods) 27 hours (22-24 and 1-24) 24 hours (1-24) 20 hours (5-24) 17 hours (8-24) 13 hours (12-24) 9 hours (16-24)
The continuous intraday market, like the intraday auction market, offers market agents the possibility of managing their energy imbalances with two fundamental differences with respect to the auction:
• In addition to gaining access to market liquidity at the local level, agents can benefit from the liquidity available in markets in other areas of Europe, provided that cross-border transport capacity is available between the zones.
• The adjustment can be made up to one hour before the moment of delivery.
The continuous intraday market is managed by the designated market operators (in the first phase, the operational NEMOs are OMIE, EPEX SPOT, and EMCO), to create an intraday con-tinuous European cross-border market. The purpose of this market is to facilitate energy trade between different bidding zones of Europe in a continuous manner and increase the overall effi-ciency of transactions in intraday markets throughout Europe.
The program resulting from each round of the continuous intraday market is the Basic Intra-day Program of Continuous Incremental Cassation (BIPCIC). The system operator, based on this program, publishes the resulting program, the Continuous Final Schedule (CFS).
All the agents authorised to submit offers for sale or purchase of electricity in the day-ahead market and who participated in the corresponding day-ahead market session or established a bilat-eral contract, or who did not participate because they were unavailable and only could be present later, are entitled to participate in the intraday market only for scheduling periods corresponding to those included in the day-ahead market session in which they participated or would participate if they were available.
As in the case of the day-ahead market, the sale bids presented to the MO can be simple or include some complex conditions.
Simple bids are economic bids of sale or acquisition of energy, of 1 to 5 stretches, that the sellers present for each time period and unit of sale or acquisition of which they are holders. These simple bids express a price and an amount of energy, with the price increasing in each tranche.
Purchase or sale bids that include complex conditions are those that, fulfilling the requirements demanded for the simple bids, incorporate all or some of the following complex conditions:
• Load gradient.
• Minimum income; Maximum payments.
• Full acceptance in the matching of the first section of the sale bid.
• Full acceptance in each hour in the matching of the first section of the sale bid.
• Condition of minimum number of consecutive hours of full acceptance of the first section of the sale bid.
• Maximum energy.
The conditions of load gradient and minimum income are the same as those described in the day-ahead market. The condition of maximum payments is equivalent to the minimum income applied to energy purchases, which will not be matched if the cost is higher than a fixed value, plus one variable value per kWh.
The condition of complete acceptance in the matching of the first section of the sale bid allows the sale bids to set a profile for all the hours of the intraday market, which can only be matched if it also matches in the first section of all the hours. This allows to adjust the programs of generation or acquisition units to a new profile, or if it is not possible in one part, to keep the previous program without modifying some of the hours individually. This option is used when it is only possible to schedule a few hours if others are also possible, as in the case of advancing the start-up or shut-down process, avoid boiler congestion, etc..
The condition of full acceptance in each hour in the matching of the first section of the sale bid implies that the first section on an hour will only be scheduled if it is fully matched, with all sections being withdrawn at that hour, not being withdrawn for the remaining hours. This option is useful for scheduling groups that produce (technical minimum) or consume (pumping consumption), a minimum value or nothing. It can also be useful for consumers to express a similar situation.
The condition of the minimum number of consecutive hours with full acceptance of the first section of the bid could be applied when the generation or acquisition unit must produce or stop consuming consecutively at least a number of hours. The same condition would apply to a con-sumer who, for example, cannot start a factory for a number of hours lower than the one specified in the bid.
The maximum energy condition allows supply or acquisition units that have a limitation on the availability of energy to bid at all hours but limiting the matched value to an overall energy maximum. This condition is necessary due to the volatility of the prices of the intraday market between hours, which does not allow knowing the hours in which the generation or acquisition units can match, and yet there is a limit to the energy that can be sold, such as the case of pumping generation units.
The bid offers for each intraday market session must respect the unit limitations corresponding to the offer unit made available to the market operator by the system operators at the beginning of the session.
The simple matching method is that which independently obtains the marginal price, as well as the volume of electric energy that is accepted for each purchase and sale offer, for each pro-gramming period.
The complex matching method obtains the matching result from the simple matching method, to which the load gradient condition is added, obtaining the simple conditioned matching. Through an iterative process, several conditioned simple casings are executed until all the married sales and acquisition units meet the declared complex conditions, being this the first provisional final solution.
Through an iterative process, it is obtained the first definitive final solution that respects the maximum capacity of international interconnection with the electrical systems external to the Iberian Market.
In case of internal congestion in the Iberian Market (congestion in the interconnection between the Spanish and Portuguese electricity systems), the previously described process is repeated, mak-ing a market splittmak-ing that obtains a price in each zone of the Iberian Market, without congestion between both electrical systems.
Both in the simple matching method and in the complex, it will be ensured that any bid that does not comply with the limitations imposed by the MOs is not matched, for safety reasons [22].
2.2
Hydro Units and Market
Hydropower can be obtained from the kinetic energy of a body of water flowing through a water turbine, promoting the rotation of an axis, which is usually attached to the rotor of the generator, and its installed capacity has experienced a steady growth in power systems all around the world. This is due, in large part, to the relatively predictable nature of the energy source, water, being this a major advantage over other renewable sources, such as wind and sunlight.
[23]In what concerns to the type of hydropower plants, these can be divided in two major types:
• Run-Of-River (ROR)
• Reservoir (with or without Pumped Storage)
Run-Of-River HPPs are hydroelectric systems that harvest the energy from flowing water to generate electricity in the absence of a large dam and reservoir, although a small dam may be used to ensure enough water goes in the penstock1, and possibly some storage for same day use, making them generally more reliable as they compensate for any discrepancies in water flow. The primary difference of this type of hydroelectric generation when compared to others is that ROR primarily uses the natural flow rate of water to generate power, instead of exploring the power of water falling a large height. However, water may still experience some vertical drop in this kind of system, as in the case of existing a small dam.
For a ROR system to be viable in a given location, two specific geographical features are required: a reasonably substantial flow rate, and a slope that allows the water to speed up signif-icantly. Therefore, ROR systems are best implemented in bodies of water with a fairly constant flow rate, preferably without peaks, as the excess water would be wasted, given the low storage capacity of this kind of hydroelectric facilities.
In ROR systems, running water from a river is guided down a channel or penstock. There can be some change in altitude at this point (from a small dam or natural slope) so there may still be some contribution from "falling water." The diverted water is brought to an electricity generating house, where it drives a turbine, running a generator. After being used, the water is fed to the river downstream [25].
On the other hand, reservoir HPPs rely on a height difference (hydraulic head) between the level of the reservoir, created by a dam, and the river downstream, exploring the potential and kinetic energy of the stored water, which is then directed into the turbines through the penstock. In theory, the bigger the hydraulic head, the more potential energy the reservoir has and thus the more electricity can be generated [26].
Both ROR and Reservoir HPPs are highly efficient since the energy of moving water is purely mechanical, with only minor losses appearing, related to friction in the turbines’ components. Although, each one has its advantages and disadvantages.
Run-Of-River systems are usually less expensive to build and can be built over a shorter pe-riod of time. ROR facilities also avoid some of the environmental problems associated with the flooding, since the pondage is much smaller than the lakes created by dams. In its turn, reservoir facilities have a significantly larger power output as well as a lower cost per kWh. Furthermore, the existence of a dam and a reservoir means these systems have a larger storage capacity, al-lowing a better management of available resource, making them more reliable for generation of electricity [25].
1Pipes or long channels that carry water down from the hydroelectric reservoir to the turbines inside the actual power
In what concerns to the size of HPPs, two classes were established: • Small power plants (SHP) - up to 10 MW
• Large power plants - more than 10 MW
In Continental Portugal, there are eight hydrographic basins with installed hydropower capac-ity, as shown in table2.5.
Table 2.5: Hydropower Units by Hydrographic Basin in Continental Portugal [6] [7]
Type of HPP Size Basin ROR Reservoir
(w/ Pumped Storage) SHP Large
Installed Capacity (MW) Average Annual Productivity (GWh) Lima 1 2 (-) - 3 696 1 008 Cávado - 9 (5) - 9 1 633 2 924 Ave 1 4 (-) 5 - 26 71 Douro 7 9 (3) 1 15 2 828 4 462 Vouga 2 1 (-) 2 1 83 144 Mondego 7 5 (1) 6 6 469 498 Tejo 3 7 (-) 3 7 595 1 566 Guadiana - 2 (2) - 2 514 n/a
Table2.6presents the evolution in the values for total installed power in Continental Portugal, on a period from 2015 to the first quarter of 2018:
Table 2.6: Installed Capacity in Continental Portugal (MW) [8] [9] [10] [11] 2015 2016 2017 1st Q. 2018
Total 18 560 19 530 19 800 19 820
Renewable 12 045 13 065 13 400 13 415
Hydro 6 160 6 945 7 190 7 200
Pumps 1 640 2 435 2 700 2 700
Dispatchable Power Stations 11 320 12 110 12 370 12 375 Non-Dispatchable Power Stations 7 240 7 420 7 430 7 445
Table2.7 shows the variation on power generation and demand in Portugal for a period be-tween 2015 and the first quarter of 2018, which is also compared with the first quarter of 2017:
Table 2.7: Power Generation and Demand in Portugal (GWh) [8] [9] [10] [11] 2015 2016 2017 1st Q. 2017 1st Q. 2018 Total 48 165 55 875 54 545 14 740 15 100 Renewable 23 165 31 060 21 175 7 330 8 875 Hydro 8 453 15 415 5 535 2 695 3 535 Pumped Storage 1 160 1 215 1 805 445 510 (Consumption of Pumps) 1 465 1 520 2 225 555 635 Import Balance 2 265 -5 085 -2 685 -1 195 -865 Total Demand 48 965 49 270 49 670 12 990 13 600
Unlike other types of renewable energy, such as wind and solar, hydropower has a factor of predictability associated to it, as well as, in the case of reservoir HPPs, the ability to store energy. Therefore, hydropower generation is dispatchable, being traded in the market, aggregated by hydrographic basin.
Generation companies should identify the most interesting operation strategy for the next day or week. This identification should be driven by the maximisation of their profits and the results of the corresponding short term planning exercises can then be used to build a set of selling bids to be communicated to the day-ahead market operator. These bids are often formulated in terms of the maximum possible power to be generated namely when it is expected that electricity market prices are larger thus yielding larger profits. Therefore, generation periods tend to be located when the electricity prices are larger and pumping when they are more reduced [27].
Hydro station agents’ strategy is very dependent on the type of reservoir and inflows. De-pending on the hydro type, the bidding price strategy is determined by the water value on the reservoir. Having an accurate model for hydro pumping units is very important in the Portuguese case because these units often behave as price makers [16].
2.3
Sectorial Pricing Forecasting
Since the inception of competitive electricity markets two decades ago, Electricity Price Forecast-ing (EPF) has gradually become a fundamental process for energy companies’ decision makForecast-ing mechanisms. Over the years, the bulk of research has concerned point predictions. However, the recent introduction of smart grids and renewable integration requirements has had the effect of in-creasing the uncertainty of future supply, demand, and prices. The research community has come to understand that probabilistic electricity price (and load) forecasting is now more important for energy systems planning and operations than ever before [28].
Depending on the purpose of the forecasting process, three forecasting horizons are often considered, although without well defined thresholds [29]:
• Short-term EPF - usually involves forecasts from a few minutes up to a few days ahead, and is essential in day-ahead and intraday markets operations.
• Medium-term EPF - with time horizons from a few days to a few months ahead, is generally preferred for balance sheet calculations and risk management.
• Long-term EPF - with lead times measured in months, quarters or even years, has as main objective investment profitability analysis and planning, such as determining the future sites or fuel sources of power plants.
[29] presents a variety of models that have been studied for electricity price forecasting (EPF) over the last two decades, with varying degrees of success:
• Multi-agent (multi-agent simulation, equilibrium, game theoretic) models, which simulate the operation of a system of heterogeneous agents (generating units, companies) interact-ing with each other, and build the price process by matchinteract-ing the demand and supply in the market. Equilibrium (game theoretic) approaches may be viewed as generalisations of cost-based models, amended with strategic bidding considerations. These models are espe-cially useful in predicting expected price levels in markets with no price history, but known supply costs and market concentration. On the other hand, the increasingly popular adap-tive agent-based simulation techniques can address features of electricity markets that static equilibrium models ignore.
• Fundamental (structural) methods. These models try to capture the basic physical and economic relationships which are present in the generation and trading of electricity. The functional associations between fundamental drivers (loads, weather conditions, system pa-rameters, etc.) are postulated, and the fundamental inputs are modelled and predicted in-dependently, often via statistical, reduced-form or computational intelligence techniques. Moreover, many of the EPF approaches considered in the literature are hybrid solutions with time series, regression and neural network models using fundamental factors – like loads, fuel prices, wind power or temperature – as input variables.
• Reduced-form (quantitative, stochastic) models, which characterise the statistical properties of electricity prices over time, with the ultimate objective of derivatives evaluation and risk management. their main intention is not to provide accurate hourly price forecasts, but rather to replicate the main characteristics of daily electricity prices, like marginal distributions at future time points, price dynamics, and correlations between commodity prices. Such models lie at the heart of derivatives pricing and risk management systems. If the price process chosen is not appropriate for capturing the main properties of electricity prices, the results from the model are likely to be unreliable.
• Statistical (econometric, technical analysis) approaches, which are either direct applications of the statistical techniques of load forecasting or power market implementations of econo-metric models. These models forecast the current price by using a mathematical combina-tion of the previous prices and/or previous or current values of exogenous factors, typically consumption and generation figures, or weather variables. The two most important cate-gories are additive and multiplicative models. They differ in whether the predicted price is the sum (additive) of a number of components or the product (multiplicative) of a number of factors. The former are far more popular.
• Computational intelligence (artificial intelligence-based, non-parametric, non-linear sta-tistical)techniques, which combine elements of learning, evolution and fuzziness to create approaches that are capable of adapting to complex dynamic systems, and may be regarded
as ‘intelligent’ in this sense. Artificial neural networks, fuzzy systems, support vector ma-chines (SVM) and evolutionary computation (genetic algorithms, evolutionary program-ming, swarm intelligence) are unquestionably the main classes of CI techniques. CI models are flexible and can handle complexity and non-linearity. This makes them promising for short- term predictions, and a number of authors have reported their excellent performance in EPF.
Monteiro et al. [2] present two important models influenced by explanatory variables for day-ahead price forecasting in the MIBEL:
• An Explanatory Model for Price Forecast (EMPF), at a regional level, which integrates a wide set of explanatory variables, which include regional aggregation of power multi-generations and power demands, recent prices, broad hourly time series records of weather forecasts as well as other chronological information. This EMPF model constitutes the best explanatory model for forecasting purposes from point of view of including the most complete and suitable set of explanatory input variables.
• An innovative Reference Explanatory Model for Price Estimation (REMPE model) for hourly price estimations based on actual power multi-generations and actual power demands of such day, that is, based on real data of explanatory variables. It allows the calculation of the best hourly price estimations and the corresponding error that represents the lowest limit of forecast error values reachable with the used explanatory variables. It should be noticed that the REMPE model is not a forecasting model because actual information of multi-generations and power demands are not available for day-ahead hourly price forecasting. This REMPE model will determine the lowest limit of the forecasting error than the EMPF model could achieve what will allow evaluating the quality of the forecasting performance of the EMPF model.
The EMPF model follows the typical time framework of short-time price forecasting, provid-ing hourly prices for the day-ahead, as illustrated in fig.2.2.
Figure 2.2: Time framework of the EMPF model for price forecasting [2]
Figure2.3illustrates the time framework for the REMPE model for price estimation, and it allows to understand the differences between EMPF and REMPE models in what concerns to variables to be considered.
Figure 2.3: Time framework of the REMPE model for price estimation [2]
These two models use different kinds of explanatory variables. Tables2.8and2.9present the variables that influence each one of the models according to [2], respectively, the EMPF and the REMPE models:
a) Actual recorded hourly data electricity prices, that is, real information known up to day D. b) Chronological variables: hour, week day, holiday, week number and month number, for
both past and future periods, although in [2] these models are influenced only by the “hour” and “week day” variables, including a value to identify holidays in the “week day” variable. c) Actual recorded hourly power system data, mainly regional aggregated hourly power
de-mands and regional hourly power generations aggregated by generation type. d) Hourly weather forecasts, including windspeed, solar irradiance, and temperature.
e) Power system hourly variable forecasts: power demand forecasts, wind power forecasts, solar power forecasts, hydropower forecasts, independent cogeneration forecasts, thermal power forecasts, etc. However, due to the fact that such information is only available to some actors, the models in [2] used the hourly weather forecast information previously described in paragraph (d), since such information is related partially to some of the power system hourly variable forecasts.
f) Power market restriction variables: unavailable capacity for power generation, reserves of power generation and interconnection, volume of electric energy allocated in other elec-tricity markets, and elecelec-tricity futures market and bilateral contracts. Since this is inside information, whose usefulness and usability have not been studied yet, the models in [2] do not include it.
Table 2.8: Explanatory variables of the explanatory model for price forecast (EMPF) model [2].
Variable Description
V1 hour
V2 week day
V3 hourly price D V4 hourly price D – 6
V5 hourly power demand D - 1 V6 hourly power demand D - 6
V7 hourly hydropower generation D - 1 V8 hourly hydropower generation D - 6
V9 hourly cogeneration and solar power generation D – 1 V10 hourly cogeneration and solar power generation D – 6 V11 hourly coal power generation D – 1
V12 hourly coal power generation D – 6 V13 hourly nuclear power generation D – 1 V14 hourly nuclear power generation D – 6
V15 hourly combined cycled power generation D – 1 V16 hourly combined cycled power generation D – 6 V17 hourly forecasted temperature D + 1
V18 hourly forecasted wind speed D + 1 V19 hourly forecasted irradiance D + 1
Table 2.9: Explanatory variables of the reference explanatory model for price estimations (REMPE) model [2]. Variable Description V1 hour V2 week day V3 hourly price D V4 hourly price D – 6
V5R hourly power demand D + 1
V6R hourly wind power generation D + 1 V7R hourly hydropower generation D + 1
V8R hourly cogeneration and solar power generation D + 1 V9R hourly coal power generation D + 1
V10R hourly nuclear power generation D + 1
V11R hourly combined cycled power generation D + 1 V12R hourly forecasted temperature D + 1
V13R hourly forecasted wind speed D + 1 V14R hourly forecasted irradiance D + 1
In what concerns to Probabilistic Price Forecasting Models (PPFM), Monteiro et al. [3] present a model based on the Nadaraya-Watson Kernel Density Estimator (NW-KDE), which consider a historical dataset composed of n cases, corresponding to past instants p (p = 1, 2, ... n), of m price explanatory variables, x, and the corresponding dependent variable (hourly electricity price variable, y). This set of n historical cases, composed by hourly time series of the mentioned variables, constitutes the knowledge base dataset (matrix of knowledge).
Figure 2.4: Time framework of the PPFM model [3]
For the development of the PPFM model, the following kinds of data were considered: a) Chronological data (hour, day of the week).
b) Actual hourly data prices of the electricity market (MIBEL).
c) Actual hourly data of the power system: power demand, hydropower generation, wind power generation, cogeneration and solar power generation, coal power generation, nuclear power generation, combined cycle power generation and power exchanged with France. d) Data of hourly forecasts of the power system: wind power generation forecasts and power
demand forecasts.
e) Data of hourly weather forecasts: weighted average temperature, solar irradiance and wind speed.
Table 2.10: Explanatory variables of the PPFM model [3]. Variable Description V1 hour V2 week day V3 hourly price D V4 hourly price D – 6
V5 hourly power demand D - 1 V6 hourly power demand D - 6
V7 forecasted hourly power demand D + 1 V8 forecasted hourly temperature D + 1 V9 forecasted hourly wind speed D + 1 V10 forecasted hourly irradiance D + 1
V11 forecasted hourly wind power generation D + 1 V12 hourly wind power generation D - 1
V13 hourly hydropower generation D - 1
V14 hourly cogeneration and solar power generation D - 1 V15 hourly coal power generation D - 1
V16 hourly nuclear power generation D - 1
V17 hourly combined cycle power generation D - 1
2.4
Hydropower Generation Forecasting
Generation companies having hydro power plants in their portfolio have to identify the most ade-quate operation strategy in order to maximise their profit. In a competitive environment, they have to build selling bids (and buying when they have pumping) and send them to the day-ahead mar-ket operator. In addition to the uncertainty associated to the hydro conditions, the optimisation of hydro power plants is a complex and nonlinear problem namely due to their nonlinear nature (non-linear relationship between the discharge volume, the net head, and the hydropower) together with the number of constraints and variables to consider and the interdependencies between several stations in cascades. [16] [27].
2.4.0.1 The H4C2 Model
In past years, short-term forecasts of meteorological variables in many different applications have been frequently obtained by NWP models. Such models are used to obtain meteorological fore-casts of water precipitation values, which are input values of the H4C2 model.
Water precipitation in the hydrographic basins where the hydroelectric power plants are lo-cated ultimately determines the principal water flows related to the hydroelectric power genera-tion of such plants. Analysing the accumulagenera-tion of precipitagenera-tion over time, hydropower generagenera-tion increases when water precipitation occurs and decreases in periods without precipitation.
The objective of the H4C2 model is to forecast the hourly hydropower generation Ph
(fore-casted power of the aggregated generation of the hydropower plants, for hour h), based on meteo-rological precipitation forecasts (hourly water precipitation values using NWP tools) and based on other independent causal variables, which are the aggregated load demand of the electric power system, the aggregated wind power generation of the system, and another additional variable (hour h).
The hourly hydropower generation Ph, in MW, to be forecasted by the H4C2 model, is defined
by [30]:
Ph= Ph0+ Dh
where Ph0 is the “monthly” hourly hydropower generation, for hour h, in MW, and Dh is the
hourly power deviation, for hour h, expressed in MW.
The H4C2 model forecasts are carried out in any hour of the current day, D, and they corre-spond to the hourly hydropower generation for the 24 h of the next day, D + 1, from its first hour, z, to its last one, z + 23. The forecasts of the hourly water precipitation are obtained beforehand from the NWP meteorological model. The stages for obtaining the forecasts of the hydropower generation are the following ones:
• First stage. This stage is optional and it consists on calculating the monthly actual power generation (monthly moving average of the values of the actual hourly power generation) for the hour x, using historical data. Afterwards, the Hydrological power potential for all the hours from the hour x + 1 to the hour 23 of the day D (that is, the hour z − 1) are calculated. If this first stage is applied, then the value of the hourly Hydrological power potential for the last hour, z − 1, of the current day, D, will be different from the value of such hourly Hydrological power potential obtained in the forecasting process carried out in the previous day, D − 1.
• Second stage. The second stage consists on obtaining the hourly hydropower generation forecasted values for the 24 hours of the day D + 1.
Presented in [30] as an original model, the H4C2 model comprises three modules:
• The first module provides forecasts of the monthly hourly power generation of the hy-dropower plants, taking into account the water precipitation forecasts and the Hydrological power potential values.
• The second module provides forecasts of the hourly power deviation values, which repre-sent temporal variations of the monthly hourly power generation of the hydropower plants (from the first module) with respect to the hourly forecasted power of the aggregated hy-dropower generation. Causal variables (aggregated load demand and wind power genera-tion) are utilised by this module.
