Commonly, the day-ahead and intraday market on the electricity exchange are treated separately in academia. However, a model that forecasts the direction of the price spread between these two markets creates an opportunity for a market participant to leverage the price spread. In the neighbouring domain, electricity price forecasting, deep learning has proven to excel. Therefore, it is hypothesised that it will do so in directional price spread forecasting as well. A quantitative case study was performed to investigate how accurately a deep learning approach could be in directional electricity price spread forecasting. The case study was conducted on the Nordic electricity exchange Nord Pool in the SE3 region. The deep learning approach was compared with previously suggested machine learning models and a naive heuristic. The results show no statistical difference in error rate between the deep learning model and the machine learning model or naive heuristic. The results suggest that deep learning might not be a suitable approach to the task or that the implementation did not fully exhaust the potential of deep learning. / Vanligtvis behandlas marknaden för day-ahead och intraday på elbörsen separat i den akademiska litteraturen. En modell som prognostiserar riktningen för prisskillnaden mellan dessa två marknader skapar dock en möjlighet för en marknadsaktör att utnyttja prisskillnaden. I grannområdet elprisprognoser har djupinlärning visat sig överträffa andra typer av modeller. Därför antas det att djupinlärning även kommer göra det i riktade prisskillnadsprognoser. En kvantitativ fallstudie utfördes för att undersöka hur precis en djupinlärningsmetod kan vara i prognos för riktad elprisskillnad. Fallstudien genomfördes på den nordiska elbörsen Nord Pool i SE3-regionen. Djupinlärningsmetoden jämfördes med tidigare föreslagna maskininlärningsmodeller och en naiv heuristik. Resultaten visar ingen statistisk skillnad i fel-andel mellan djupinlärningsmodellen och maskininlärningsmodellen eller naiv heuristik. Resultaten antyder att djupinlärning kanske inte är ett lämpligt tillvägagångssätt för uppgiften eller att implementeringen inte helt utnyttjar potentialen för djupinlärning.

Cette thèse s'intéresse à plusieurs aspects des marchés de gros de l'électricité. L'achat et vente d'électricité se négocient sur les marchés à terme et sur le marché day-ahead. Sur ce dernier se pratique un type d'enchère très spécifique, où les enchères des acteurs prennent la forme de fonctions d'offre et de demande. Chapitre 2 prend comme point de départ un résultat de Zachmann et von Hirschhausen (2008) qui constatent une réponse asymétrique du prix de gros de l'électricité en Allemagne au changement du prix des permis d'émission négociable ( EUA ). Cependant, en contradiction avec les résultats existants, il est démontré que l'asymétrie a disparu suite à la publication d'un rapport d'enquête par l'autorité de la concurrence. Chapitre 3 porte sur l'interaction des marchés à terme et day-ahead dans un jeu d'oligopole répété. L'effet du marché à terme sur la stabilité des collusions est étudié dans le cas où les stratégies sur le marché spot prennent la forme des fonctions d'offre. Il est démontré que la simple existence d'un marché à terme peut élargir l'intervalle des valeurs du facteur d'actualisation pour lesquelles la collusion est soutenable. Chapitre 4 examine si une réaction asymétrique au changement du prix du C02 est également présente dans les fonctions d'offre du marché d'électricité day-ahead. À cette fin, les outils de l'analyse des données fonctionnelles sont adoptés et appliquées à des données des enchères. Chapitre 5 développe un test pour l'auto-corrélation dans un panel d'observations fonctionnelles. Une simulation Monte-Carlo montre une bonne puissance du test dans des échantillons de taille habituellement utilisé dans la recherche appliquée. / This thesis addresses aspects of interacting markets in electricity wholesale. Electricity is traded in forward markets and in day-ahead auctions, which implement a very specifie market design. The bids of market participants take the fonn of supply and demand functions. Chapter 2 builds upon a finding of Zachmann and von Hirschhausen (2008) who report an asymmetric response of electricity wholesale prices for Gennany to changes in the price of EV Emission Allowances (EVA). ln contrast to the fonner contribution, it is shown that the asymmetry disappeared in response to a report on investigations by the competition authority. Chapter 3 addresses the interaction offorward markets and day-ahead auctions in a repeated oligopoly game. The effect offorward trading on the sustainability of collusion is studied for the case that spot market strategies take the fonn of supply functions. It is shown that the existence of forward markets enlarges the range of discount factors for which collusion can be sustained. Chapter 4 examines if an asymmetric reaction to EVA prices can also be found in the supply functions from the day-ahead market. To this end, tools from the field of functionaJ data analysis are adopted and applied to observed bids from the day-ahead auction. Chapter 5 develops a test for autocorrelation in functional panel data. Asymptotic nonnality of the statistic is proved, and Monte-Carlo simulation sho\l good power of the test in sample sizes which frequently prevail in applied research.

One of the key steps of the liberalization of the Turkish electricity market has been the estab- lishment of PMUM (Turkish day ahead electricity market). The aim of this study is to explore the dynamics of electricity prices observed in this market and their relation with temperature observed in Turkey. The electricity price process is studied as a univariate process and the same process is studied along with temperature together as a two-dimensional process. We give a fairly complete model of temperature. We observe that the electricity prices in Turkey exhibit many of the features that similar prices exhibit in other world markets. In particular, Turkish day ahead prices are seasonal / every year the price seems to follow a path similar to the one years preceding it. To simplify our analysis we focus our study to a 35 day pe- riod where every year the prices show a relatively simple behavior. We study the effects of the fluctuations in temperature in this period on the fluctutations in the day ahead electricity price.

Over the past several decades, many techniques and approaches have been proposed and implemented for load and price forecasting. The objective of all of these methods was load and price forecasting with minimal error. However, researchers face several challenges in achieving this goal. For price forecasting, the main challenge is to forecast electricity prices accurately in a deregulated electric power market with volatile aspects. Decentralized or deregulated markets are very volatile systems. Hence, pattern following and accurate forecasting of electricity prices are difficult tasks using ordinary methods. In this thesis, a novel approach is introduced and implemented to overcome the challenges inherent in accurate price forecasting. This novel approach involves innovations in forecasting to improve the spot power price forecasting accuracy in a competitive market. To investigate the applicability and effectiveness of this technique, Multiple Linear Regression (MLR) and Artificial Neural Networks (ANN), two well-known forecasting techniques, are developed.

In deregulated electricity markets, profit driven electricity retailers compete to supply cheap reliable electricity to electricity consumers, and the electricity consumers have free will to switch between the electricity retailers. The need to maximize the profits of the electricity retailers while minimizing the electricity costs of the electricity consumers has therefore seen a drastic increase in the research of electricity markets. One of the factors that affect the profits of the electricity retailers and the energy cost of the consumers in electricity retail markets is the supply and demand. During high-supply and low-demand periods, the excess electricity if not managed, is wasted. During low-supply high-demand periods, the deficit supply can lead to electricity blackouts or costly electricity because of the volatile electricity wholesale spot market prices. Research studies have shown that electricity retailers can achieve significant profits and reduced electricity costs for their electricity consumers by minimizing the excess electricity and deficit electricity. Existing studies developed load forecasting models that aimed to match electricity supply and electricity demand. These models reached excellent accuracy levels, however due to the high volatility character of load demand and the rise of new electricity consumers, load forecasting alone is unable to mitigate excess and deficit electricity. In other studies, researchers proposed charging the electricity consumers’ batteries with excess electricity during high-supply low-demand periods and supplying their deficit electricity during low-supply high-demand periods. Electricity consumers’ incorporating batteries resulted in minimized excess and deficit electricity, in turn, maximizing the profits for the electricity retailers and minimizing the electricity costs for the electricity consumers. However, the batteries are consumer centric and only provide battery energy for the battery-owned consumer. Electricity consumers without battery energy during low-supply highdemand periods have electricity blackouts or require costly electricity from the electricity wholesale spot market. The peer-to-peer (P2P) energy sharing framework which allows electricity consumers to share their energy resources with one another is a viable solution to allow electricity consumers to share their battery energy. P2P energy sharing is a hot topic in research because of its potential to maximize the electricity retailers’ profits and minimize the electricity consumers’ electricity costs. Due to the increased profits for the electricity retailer and reduced electricity costs for the electricity consumers from implementing battery charging and P2P energy sharing, this dissertation proposes a day-ahead electricity retail market structure in which the electricity retailer supplies consumers’ batteries with excess electricity during high-supply low-demand periods, and during low-supply highdemand periods the electricity retailer discharges the consumers’ batteries to supply their deficit supply or supply their peers’ deficit supply. The electricity retailer aims to maximize its profits and minimize the electricity cost of the electricity consumers in its electricity retail market, by minimizing the excess and deficit electricity. The problem is formulated as a non-linear optimization model and solved using game theory. This dissertation compares the profits of the electricity retailer and electricity costs of the consumers that charge their batteries with excess electricity, discharge their batteries and purchase electricity from their peers to supply their deficit supply, with consumers that only charge their batteries with excess electricity but do not share their battery energy with their peers, consumers that only purchase electricity from their peers to supply their deficit supply but do not employ a battery, and consumers that neither employ a battery nor purchase electricity from their peers to supply their deficit supply. The results show that the consumers that charge their batteries with excess electricity, discharge their batteries and purchase electricity from their peers to supply their deficit supply achieved the lowest electricity cost and highest profits for the electricity retailer. / Dissertation (MEng)--University of Pretoria, 2020. / Electrical, Electronic and Computer Engineering / MEng / Unrestricted

Within forecasting electricity pricing, we analyse whether adding various vari- ables improves the predictions, and if shorter time intervals between observa- tions enhance accuracy of the forecasting. Next, we focus on proper selection of lagged observations, which has not been thoroughly covered in the past litera- ture. In addition, many papers studied electricity prices in larger markets (e.g. United States, Australia, Nord Pool, etc.) on datasets limited in scope, with 2-3 years timespan. To address these gaps in literature, we obtain one daily and one hourly dataset, both spanning 6 years (January 1, 2015 - December 31, 2020), from four Central and Eastern European countries - the Czech Repub- lic, the Slovak Republic, Hungary, and Romania. These contain information on the electricity prices, and information on our observed added variables - temperature and cross-border electricity flows. For the forecasting, we use two different methods - Autoregression (AR) and Seemingly Unrelated Regression (SUR). The thorough selection of lagged observations, which we accustom to the closing time of the auction-based electricity market system, serves further studies as a guidance on how to avoid possible errors and inconsistencies in their predictions. In our analyses, both AR and SUR models show that...

L’objectif de cette thèse est d’étudier les nouvelles dynamiques et leurs impacts dans le secteur de l'électricité. Elle discute des sujets critiques d’après les perspectives de la macroéconomie, de la configuration structurelle, et de la transition vers des sources d'énergie renouvelables. Plus précisément, trois sujets se dégagent: le lien entre la consommation d'électricité et la croissance économique, les impacts de l'intégration verticale entre les producteurs et les détaillants, et les impacts d'intégration de production d'énergie renouvelable intermittente. En mettant en jeu ces trois sujets, elle tente d’apporter des réponses aux défis principaux de la sécurité d'approvisionnement, de la compétitivité, et de la durabilité du développement énergétique. En donnant de nouvelles orientations dans la recherche sur l’économie de l’énergie, elle servira à éclairer des débats politiques. / The objective of this thesis is to study the new dynamics and their impacts in the electricity sector. It discusses the critical issues from the perspectives of macroeconomics, structural configuration, and a transition to renewable energy sources. More precisely, three topics emerge: the nexus between electricity consumption and economic growth, the impacts of vertical integration between power generators and retailers, and the market impacts and integration issues of intermittent renewable generation. By studying these three topics, it provides answers to the key challenges of supply security, competitiveness and sustainable development in the energy sector. By giving new research directions of energy economics, it serves to inspire related policy debates.

The unpredictability and variability of wind power generation can pose an economical risk to the wind power producer when participating in the day-ahead market and delivering the committed generation. These risks come from the creation of imbalances due to a mismatch between the sold and real generation fed to the grid. Energy Storage System (ESS) are a good solution for the wind power producer to plan the operation of the wind farm once the day-ahead market prices are cleared. However, depending on the price forecasts and wind generation, one type of storage technology might be more optimal than others. This is due to the fact that lithium-ion batteries have costs, power and energy ratings and limits that differ from other ESS (vanadium redox flow batteries, supercapacitors, pumped hydro or even other lithium-ion batteries with different chemistries). Hence, a multi-energy storage system technology solution can be proposed to be combined with a wind farm in order to both optimise the bids in the day-ahead market and to take part in current and emerging electricity markets. For this purpose, a mathematical model has been developed, and it provides the optimal bidding strategy to the day-ahead market and the most convenient operational planning for the energy storage systems. Based on the expected daily profits, a yearly stream of revenues is obtained and an overall techno-economical assessment is provided. The results show that, with the current capital costs of energy storage systems, the multi-ESS hybrid wind farm would recover the initial investment after 2-5 years depending on the ESS combinations. Moreover, the wind power producer would need an extra stream of revenues in order for it to be more profitable than the wind farm operating without storage blocks. / Den oförutsägbara och varierande vindkraftsproduktionen kan utgöra en teknisk och ekonomisk risk för vindkraftsproducenten när denne deltar i dayahead-marknaden och levererar den sålda energin. Dessa risker beror på att det uppstår obalanser på grund av bristande överensstämmelse mellan den sålda och den verkliga produktionen som matas in i nätet. Energilagringssystem (ESS på engelska) är en bra lösning för vindkraftsproducenten för att planera driften av vindkraftparken när priserna på dagen före marknaden är klara. Beroende på prisprognoserna och vindkraftsproduktionen kan dock en typ av lagringsteknik vara mer optimal än andra. Detta beror på att litiumjonbatterier har kostnader, effekt- och energimärkningar och gränser som skiljer sig från dem som gäller för vanadiumredoxflödesbatterier, superkondensatorer, pumpad vattenkraft eller till och med andra litiumjonbatterier med olika kemiska sammansättningar. Därför kan man använda en teknisk lösning med olika typer av energilager som kombineras för att både optimera budgivningen på day-ahead-marknaden och för att delta i nuvarande och nya elmarknader. För detta ändamål har en matematisk modell utvecklats som ger den optimala budstrategin för day-ahead-marknadenochdenmestpraktiskadriftsplaneringen för energilagringssystemen. På grundval av de förväntade dagliga vinsterna erhålls en årlig intäktsström och en övergripande teknisk-ekonomisk bedömning görs. Resultaten visar att med de nuvarande kapitalkostnaderna för energilagringssystem skulle återbetalningstiden för en vindkraftpark med flera olika energilager vara 2-5 år beroende på vilka energilager som kombinerats. Dessutom skulle vindkraftsproducenten behöva en extra intäktsström för att bli mer lönsam än en vindkraftpark som drivs utan lagringsblock.

This thesis investigates the relation between the day-ahead electricity market prices and the electricity production in the Nordic synchronous area of the European electric power system by looking into the market data ranging from 2015 to the current time. The increasing penetration of variable renewable energy sources, coupled with the deeper electrification in various sectors of the economy, has led to a higher volatility in the market, e.g. in the market prices. Since all power plant owners plan their production depending on prices, price forecasts, availability, it is necessary to better understand the relation between price signals and the production variations. Firstly, balancing contributions, considered as a suitable tool, are found unfit to apply to this research. Secondly, spectral analysis is used to highlight frequencies in the signals, leading to the determination of three time scales of variations: daily, two-weekly and yearly. It is used to define three timehorizons which are used to apply different mathematical tools on the market data. Thirdly, the correlation coefficient between a type of production and the day-ahead prices assesses the linearity of their respective variations. Lastly, the relative market value of the production gives insight into the , e.g. with regard to flexibility. Results of this research show that the correlation between the day-ahead prices and the production depends on the time-horizon, the production type, as well as the area. E.g., if SE2’s and SE1’s hydro production are highly correlated to prices in the daily and two-weekly patterns, the former has not enough storage to follow prices on the yearly horizon, while the latter can, due to its bigger reservoirs. Results also show that wind power is one of the drivers of the day-ahead prices, especially in the two-weekly time horizon. The increasing share of wind power will hence lead in more price variations as well as lower average price levels.. This study aims to provide insights to the plant owners with flexibility possibility on how to face those future challenges. / Detta examensarbete undersöker sambandet mellan day-ahead-priser på elmarknaden och elproduktion i den nordiska synkrona delen av det europeiska elsystemet. Analysen bygger på data mellan 2015 och maj 2021. I och med att andelen väderberoende förnybara energikällor ökar samtidigt som olika sektorer elektrifieras har elmarknaden och elpriserna blivit mer volatila. Den volatiliteten behöver elproducenter förhålla sig till när de planerar sin produktion beroende på elpriser, elprisprognoser, tillgänglighet m.m. och därför är det avgörande att belysa förhållandet mellan prissignaler och anpassning av produktionsnivåer. Arbetet inkluderar olika steg: Först undersöktes balansbidrag och det kunde visas att detta mått inte lämpar sig för att beskriva sambandet mellan produktion och elpriser. Sedan gjordes en spektralanalys för att identifiera de grundläggande variationerna: dygns-, tvåveckors- och årsvariationer. Baserad på dem specificerades tre tidshorisonter som olika matematiska mått utvärderas på. Ett sådant mått är korrelationskoefficienter som användes för att beskriva sambandet mellan produktionsdata per kraftslag och elpriser på ett linjärt sätt. Slutligen användes relativa marknadsvärden på samma sätt för att kunna undersöka olika kraftslags karakteristik och få inblick i deras flexibilitet samt körsättet. Resultaten visar att korrelationen mellan day-ahead-priser och produktion beror påtidshorisonten, kraftslag och prisområde. Till exempel är vattenkraften i norra Sverige (SE1 och SE2) starkt korrelerad med elpriserna på dyngs- och tvåveckorsnivå medan korrelationen på årsnivå är starkare för SE1 eftersom vattenmagasin i SE1 har större reglerutrymme på den långa tidshorisonten än magasinen i SE2. Resultaten visar dessutom att vindkraften är den drivande faktorn för elprisvariationerna, särskilt när det gäller tvåveckorshorisonten. Vindkraften kommer öka prisvariationerna och samtidigt sänka prisnivån. Målet med examensarbetet är att få mer inblick i kraftslagens flexibilitet och därmed öka förståelsen för framtida utmaningar.

Keeping a contingency reserve in power systems is necessary to preserve the security of real-time operations. This work studies two different approaches to the optimal allocation of energy and reserves in the day-ahead generation scheduling process. Part I presents a stochastic security-constrained unit commitment model to co-optimize energy and the locational reserves required to respond to a set of uncertain generation contingencies, using a novel state-based formulation. The model is applied in an offer-based electricity market to allocate contingency reserves throughout the power grid, in order to comply with the N-1 security criterion under transmission congestion. The objective is to minimize expected dispatch and reserve costs, together with post contingency corrective redispatch costs, modeling the probability of generation failure and associated post contingency states. The characteristics of the scheduling problem are exploited to formulate a computationally efficient method, consistent with established operational practices. We simulated the distribution of locational contingency reserves on the IEEE RTS96 system and compared the results with the conventional deterministic method. We found that assigning locational spinning reserves can guarantee an N-1 secure dispatch accounting for transmission congestion at a reasonable extra cost. The simulations also showed little value of allocating downward reserves but sizable operating savings from co-optimizing locational nonspinning reserves. Overall, the results indicate the computational tractability of the proposed method. Part II presents a distributed generation scheduling model to optimally allocate energy and spinning reserves among competing generators in a day-ahead market. The model is based on the coordination between individual generators and a market entity. The proposed method uses forecasting, augmented pricing and locational signals to induce efficient commitment of generators based on firm posted prices. It is price-based but does not rely on multiple iterations, minimizes information exchange and simplifies the market clearing process. Simulations of the distributed method performed on a six-bus test system showed that, using an appropriate set of prices, it is possible to emulate the results of a conventional centralized solution, without need of providing make-whole payments to generators. Likewise, they showed that the distributed method can accommodate transactions with different products and complex security constraints.