The effects of renewable energy sources and interconnection capacity on electricity wholesale price volatility

(1)

The effects of renewable energy sources and interconnection

capacity on electricity wholesale price volatility

Master Thesis Economics

FEB

University of Groningen

Abstract

Member states of the EU have set objectives to increase renewable energy shares and interconnection levels. Variable renewable power generation causes variability in supply, increasing the price volatility. On the other hand integration of electricity markets enhances reliability of supply, which could lead to a lower electricity price volatility. In this thesis volatility effects related to the objectives of increasing renewable energy shares and increasing interconnection levels in the EU are assessed by means of an empirical model using data of six European electricity markets between 2010 and 2017. Market-specific effects were found for the renewable energy share and interconnection capacity share on annual price volatility, varying in size and in sign. A monthly analysis of the Dutch and German market has led to positive marginal effects for both renewable energy production and electrical transfer capacity. With the current level of interconnectivity price fluctuations are more likely to be spilled over, rather than productivity gains lead to supply of cheap electricity constantly. Therefore this paper gives insight on the determinants of the electricity price volatility, which can be used to further explore opportunities for the energy transition in Europe by working towards an integrated European electricity grid.

Supervisor:

(2)

2

1. Introduction

To tackle global warming a structural change in the energy system must be made. The earth heats because of the enhanced greenhouse effect (Fourier, 1824, Flemming, 1999), caused by emissions of greenhouse gases, in particular CO2 (EPA, 2017). The electricity sector accounts for a large part of total CO2 emittance. Therefore there is pressure on electricity producers to switch from traditional, fossil fuel, sources to renewable ones. However, using renewable energy sources (RES) has significant consequences. As electricity is a peculiar, non-storable good, with large public value (Green & Newberry, 1992, Joskow, 1998a, Helm, 2002, Mulder, 2017) this is a big challenge.

Renewable energy sources are often characterised by weather dependency. The main sources for renewable power are solar, wind, hydro and biomass. Solar power can only be produced during the day. Moreover, when it is winter, demand for heating increases, while supply from solar panels decreases as days are shorter. Wind power does not follow such a straightforward production cycle, but remains to be a variable power source. Although hydropower and biomass are a flexible and controllable sources, they are often not abundant enough to be a main power source, but could be used as an additional flexible power source. Nevertheless, renewable energy has enormous potential to reduce greenhouse gas emissions. If the wind blows, wind replaces CO2 intensive production technologies, particularly coal in periods of low demand, and gas in periods of high demand (Morthorst et al., 2010). As marginal costs of renewable production are lower compared to fossil fuel power plants, the electricity price drops when demand is saturated by renewable power supply. Overall, the effect of increasing RES shares in a particular generation portfolio causes the electricity price to fluctuate due to weather cycles, making the electricity price more volatile. Another innovation observed in the electricity sector is the interconnection of regional electricity grids. The EU has a policy for integration electricity grids (EC, 2014), building to a system that some call the European Supergrid already (Toretti, 2014). To integrate electricity grids so-called cross-border-capacity is installed, allowing for cross-border electricity flows. When demand needs to be met with a high merit order, marginally expensive, power source (such as coal-fired or gas-fired power plants) it is favourable to import lower merit order produced, marginally cheaper, electricity. Hence, the electricity prices converge. Thus, by increasing cross border capacity low merit order produced electricity can be distributed to a larger market, which leads to a more efficient productive allocation of resources. The larger market enhances reliability of supply (Valeri, 2009) and leads to more stable electricity prices, hence the volatility of the electricity price decreases.

(4)

4

analysis. Price reductions can be achieved at times of low demand, but remain small at times of high demand, such that the electricity price volatility is increased. The level of integration and installed renewable capacity is thus too low to obtain continuous distributional effects decreasing the electricity price volatility.

(5)

5

2. Overview of the electricity market

2.1 Physical aspects of the electricity market

In this paragraph the physical aspects of the electricity market and therefore the constraints it is subject to are explained, while the next paragraph describes how electricity markets are shaped to cope with these constraints. Power is the rate of electric work exercised and goes with the unit Watt. Electricity and power are often referred to as the same, although electricity is the set of phenomena associated with charge, this confusion is grounded as electric power is the main use of our current electricity system. In this thesis this convention is adopted. When power is transferred a current, the physical flow of electrons (Drude, 1900), flows through a cable by means of an induced voltage. Electricity grids make use of the so called alternating current. The voltage alternates between a positive potential and a negative potential at a high frequency, typically around 50 Hz.1_{Alternating current allows for efficient long distance transportation as electrons}

only vibrate back and forth instead of physically travel long distances (Backhaus & Chertkov, 2013). To keep the frequency of the system stable, the generated power should equal power consumption instantaneously. Any imbalance between the two is initially compensated by a change in the kinetic energy of the connected motors and generators. As a result, the grid frequency will start to deviate from its rated value, which is unfavourable. Ultimately, deviations between load and supply cause a black out. The system can be seen as a flywheel system, see figure 1. As long as the sum of accelerating (electricity supply) and decelerating (electricity consumption) powers is zero, the wheel will continue to spin at a constant frequency. If however, a mismatch between generation and load appears, the accelerating and decelerating powers do not cancel out, leading to a detrimental change of the rotational frequency of the wheel (Frunt 2011).

Figure 1: Flywheel representing interaction between load, production and frequency of the grid. P stands for the production, or supply, of electricity and L for the load or electricity consumption. Picture from Frunt (2011).

(6)

6

2.2 Players and market structure

In the electricity market multiple players are active, power is produced by power producers, distributed by transmission and distribution system operators and consumed by both large industrial and small domestic consumers. In the 20th century electricity producers were centrally coordinated and electricity prices were collectively determined based on the costs of the entire production, transport and distribution system (EC, 2015). However, at the end of the 20th century liberalisation of the electricity market started, with the centrally driven effort of the European Commission being the driving force (Jamasb, Pollitt 2005). The European “Energy Act” of 1989 had aimed to establish two goals: competition in supply and efficient generation by means of coordinated production (Van Damme, 2005). Though it paved the way for privatisation and liberalisation of utility industries (Thomas, 2004), the limited competition allowed by the law proved to be conflicting with the efficiency gains aimed for (Van Damme, 2005). In 2003, the European Commission introduced a new Electricity Directive (EC/2003/54/EC) with much stronger requirements on competition. This resulted in multiple large power producers instead of vertical integrated monopolists2_and

decentralised decision-making about investments in and deployment of power plants, a key precondition for efficient markets, as competition arises (Mulder, 2017).

To ensure non-discriminatory network access, the generation activities are separated from transmission and distribution activities, which is the most effective unbundling form (Jamasb, Tollitt, 2005). Transmission is managed by transmission system operators (TSO), where after power is distributed to neighbourhoods by Distribution System Operators (DSO). TSOs manage the operation of the network and schedule generation to meet demand and to maintain the physical parameters of the network (frequency, voltage, stability) (Joskow, 2008), they manage program responsibility of players in the electricity market. A decentralised organisation is not feasible here due to the natural monopoly nature of the electricity networks and the public value ascribed to grid quality. Creating a number of neighbouring networks competing for grid usage demand is not cost-effective, because of high fixed costs. Therefore a monopoly-like organisational structure of the networks is the best solution, as it keeps costs low (Mulder, 2017). The varying electricity demand must be in balance with supply continuously to avoid frequency changes. As many (large) generators cannot easily be switched on or off, their output needs to be scheduled in advance. This is done in the processes of unit commitment and economic dispatch (Frunt, 2011). For this, producers and operators cooperate, as producers know their costs (maintenance costs, the marginal generation costs by different generators, start-up costs, ramping rates and minimum on and off times) and the TSO bares the responsibility of balancing. The continuous process of unit commitment and economic dispatch causes a fluctuating electricity wholesale price.

The prices in the wholesale market are based on the marginal cost of production and opportunity costs and marginal willingness to pay (Mulder, 2017). The opportunity costs arise as traders can also trade in several forward markets, spanning from hours ahead of operation up to months or even years ahead of operation (Frunt, 2011, Mulder, 2017). Getting closer to the actual delivery time of power, better predictions can be made for load and generation. In practice the major time frames are the day-ahead market, the intraday and balancing market (Crampes, von der Fehr, Steele, 2017).

2_{In England one monopolist CEGB split into three (Joskow, 2008), in Italy ENEL split in three (Dessenibus et al.,}

(7)

7

Marginal costs of producing electricity varies from hour to hour, demand fluctuates continuously and the set of power producers own a portfolio of generators facing a multitude of marginal costs. Such a portfolio can consist of generators such as coal-fired power plants, gas-fired power plants, nuclear power plants, hydroelectric dams, wind and solar farms. Different techniques have their advantages and disadvantages. Capital costs for fossil fuel plants are relatively low, while they are high for RES technology and very high for nuclear power plants. The fuel costs however are high for fossil fuel plants, low for nuclear and zero for RES technologies (Heptonstall 2007, EIA 2019). Also large generators can generally not be switched on or off easily, that is nuclear plants are very inflexible (I. Pavić, T. Capuder, I. Kuzle, 2016), while gas-fired power plants are flexible. It is thus so that, by economic dispatch, (only) in times of high demand power is produced by means of high marginal costs and flexible techniques.

Large consumers deserve an extra note, by flexible combined heat and power (CHP) production a significant share of industry produces their own heat and power. This is used to balance the system by flexible adoption to over- or undersupply, and is settled in a separate market: the balancing market. This balancing market generally has a time span of less than an hour. When expected imbalance arises, the electricity imbalance market price causes an incentive for player to react on the market. With higher expected imbalances higher prices occur, such that participants (Balance Responsible Parties) supplying in the opposite direction of the imbalance are rewarded (Mulder, 2017). The system operator will solve the remaining imbalance during the ISP using own reserve power.

Households, but also (small) firms, schools etc. are small consumers. Their program responsibility is faced by balance responsible parties, retailers. Having all individual households face their program responsibility would be too complicated, yet aggregation of individual consumers increases predictions of demand (Frunt, 2011). In return retailers often obtain a mark-up. Moreover the pricing of retailers is often not in line with fluctuations in the wholesale market3_{, but goes via a linear pricing curve. However, a consequence is that}

this leads to a less efficient market. On the one hand household lack the willingness or ability to search for contracts that best fit their needs (Ofgem, 2011a). On the other hand the lack of transparency of retailers, while price differences between retailers are high, and negative spill-overs of bad experiences with one retailer (bankruptcy, fraudulent billing, and bad customer support), reduces consumers’ faith in the market (Mulder, Willems 2019). Moreover, the number of retailers is often limited and the number of entrants remain small, in Europe for example 70% of the aggregate market share is in hands of only 3 firms. Retailers offer multiple contracts, for example for green or traditional power. It can therefore be hard to compare contract, and it is sometimes used for retailers to avoid price-setting market behaviour. Nevertheless, product innovation is strong, particularly in the green energy sector, this might lead to welfare benefits. However, this welfare increase is mainly distributed over active households, while inactive consumer might be harmed (Mulder, Willems 2019).

Summing up, the electricity market is a complex market with high public value. Therefore a lot of regulations are in place. Although regulations have changed over time, since 2003 the regulation on the European market is relatively stable. Since then competition between power producers has been established. Consequently, the wholesale electricity price depends largely on the marginal costs of producers. The

(8)

8

(9)

9

3. Literature Review

3.1 Price effects of renewable energy

Renewable power generation is characterised by 1) (almost) zero marginal costs of production (Heptonstall, 2007, Tietjen et al., 2016, Mulder, 2017, EIA, 2019), and 2) weather dependence (Sensfuß, Ragwitz, Genoese, 2008, Von Bremen, 2010, Schleicher-Tappeser, 2012, Staffell, Pfenninger, 2018). When the share of renewable power production increases in a certain power portfolio (the collection of power plants in a geographical area), the electricity price will be affected significantly (Morthorst et al., 2010), especially as electricity demand is inelastic. Consequences of increased renewable energy source shares can be broken down in two effects. The first is the merit order effect (MOE) (Sensfuß, Ragwitz, Genoese, 2008). The electricity generated by renewable energy sources has to be bought by supply companies in advance, such that the residual demand to be met with other sources is reduced. This lowers the electricity price when renewable energy sources are operational. The second effect is due to flexibility. Traditional power sources remain to be required to meet demand on days when the weather limits renewable power production (Moraga, Mulder, 2018), this could lead to temporary higher electricity prices.

Due to the low marginal costs, wind power enters the supply curve, or merit order curve, near the bottom, shifting the supply curve to the right, see figure 2. Therefore the power price is reduced when wind power is operational. However, the exact reduction depends on the demand level. In general, at times of high demand the price reduction is fiercer, as the supply curve is steep in that regime. In order to analyse the price reduction a comparative study, with a fixed reference, corresponding to a situation with zero contribution from renewable sources must be made. However, this fixed reference is purely theoretical, usually calculations only show how renewable production influences power prices, but doesn’t answer the question what the power price would have been without renewable power production enabling weather (Morthorst et al., 2010). The literature describes several methodological approaches like OLS (O'Mahoney & Denny, 2011, Clò, Cataldi, Zoppoli, 2015) combined with load weighted averages (Cludius et al., 2014), or more complicated comparative simulations using PowerACE models (Sensfuß, Ragwitz, Genoese, 2008) and mixed integer linear programming (Delarue et al., 2009), due to a high set of constraints.

(10)

10

Figure 2: The effect of wind power shifting the supply curve at different times of the day. When wind power is operational it enters at the left of the supply curve, reducing the electricity price. Source: EWEA The Economics of Wind Energy (2009).

In an empirical study on the German spot market, Cludius et al. (2014) show that electricity generation by wind and PV has reduced spot market prices considerably by 6 €/MWh in 2010 rising to 10 €/MWh in 2012 and forecast to reach 14-16 €/MWh in 2016. The estimated merit order effects range from − 0.97 to − 2.27 €/MWh for wind and from − 0.84 to − 1.37 €/MWh for PV. Unfortunately they don’t give insights on the installed capacities, making it hard to compare with the final result of Sensfuß, Ragwitz, Genoese (2008). Moreover they mention that at very low levels of conventional generation, the price of electricity can fall below €0, where low demand coupled with a lot of wind, depressed prices to below − 200 €/MWh. On the other hand, prices can climb to above 100 €/MWh, during times of high demand and low renewable feed-in, as generators with very high marginal cost (gas turbines, oil generators) set the price.

In Italy, Clò, Cataldi and Zoppoli (2015) find that an increase of 1 GWh in the hourly average of daily production from solar and wind sources has, on average, reduced wholesale electricity prices by respectively 2.3€/MWh and 4.2€/MWh over the period 2005–2013 and has amplified their volatility. They endorse the conclusion of Sensfuß, Ragwitz, Genoese (2008), the impact of RES power production on the electricity price has declined as RES production has increased.

In Ireland, O'Mahoney, Denny (2011) find a reduction of €141 million in market dispatch savings due to wind energy for the year 2009 only. Without wind energy the electricity price would have been 12% higher. Renewables account for 10% of the market share. The average “shadow price”, the marginal cost of the most expensive unit required to meet demand in the same period, is 36€/MWh, leading to an absolute price reduction of 4.3€/MWh. Although the article doesn’t mention absolute effects, this lower price reduction with respect to the German market is due to a lower share of renewables.

(11)

11

al. (2014) for Spain, Würzburg et al. (2013) for Germany and Austria and also in the United States similar effects are found by Woo et al. (2016).

From this literature the conclusion can be drawn that renewables have significant effect on electricity prices, inducing a merit-order effect. Moreover, market values of RES are largely influenced by their share in overall electricity generation and the respective time of day in which variable RES feed-in takes place. Overall, the downward effect of variable RES on electricity spot prices seems to be unambiguous (Welisch, Ortner, Resch, 2016). Nevertheless, it is important to realize that wind and solar power lead to substantial variation in electricity prices, increasing the price volatility.

3.2 Price effects of grid integration

Another ongoing phenomenon in Europe on electricity markets is the integration of electricity grids. In October 2014, the European Council called for all EU countries to achieve interconnection of at least 10% of their installed electricity production capacity by 2020. (EC 2014) 15 European Union members have so far met this target (EC, 2015). However, the European project for integration of Member States’ electricity markets has historically relied on privatisation (Joskow, 2008), not on EU objectives. Torreti (2014) endorses this analysis, he finds that privatisation in most cases led to higher power exchange and benefits are more significant where privatisation measures have been in place for a longer period. Nevertheless, further integration of electricity grids could lead to more advantages, though most (European) markets are readily privatised. To give a sense of the need for integration Neuhoff (2004) states that from 1975 to 2000, electricity trade had doubled, also Torreti (2014) mentions that while cross border capacity hasn’t increased significantly from 2000, the traded volume has risen the following decade.

Three main advantages will be discussed in the following literature review: competitive effects (Borenstein et al. 1997) i.e. allocative efficiency, productive efficiency i.e. access to lower cost imports (Hobbs, Rijkers, Boots, 2005), supply security as well as balancing of historically grown power portfolio (Toretti, 2014) which could potentially decrease the need for spare generation capacity (Hobbs, Rijkers, Boots, 2005). Furthermore some argue that integration facilitates more efficient use of complementary resources and lowers congestions (Toretti, 2014).

(12)

12

market, divided in North- and South-California. In 2000, Borenstein et al. showed that increasing the transmission line capacity from 3.000 MW to 3.835 MW, the North-Californian price, where producers face competitive disadvantage, dropped from 169.3 $/MWh to 27.8 $/MWh.

Neuhoff (2004) published an article titled “Integrating transmission and energy markets mitigate market power”. He poses that in any unconstrained or partially constrained network, integration mitigates market power of strategic generators and avoid inefficient production decisions. The Dutch-German and Norway-Sweden interconnections are used as a case study and support the theory.

Hobbs, Rijkers, Boots (2005) conduct an empirical study on the Dutch/Belgian electricity market, connected to the German and French market. They find a yearly social benefit of 100 million € upon market coupling. This is obtained by a Cournot-Nash approach simulation. This benefit is however distributed. The enhancement of social welfare in Belgium is at cost of higher electricity prices in the Netherlands. If the Belgian incumbent is consistently a price taker, instead of a Cournot competitor, gains in social surplus are smaller but more evenly distributed between the Netherlands and Belgium. Also they note, the size of the efficiency improvements and their distribution depend on the amount of market power.

Ehrenmann and Neuhoff (2009), find that in a two-node scenario moving to integrated markets always reduces market power and increases welfare. In a three node scenario the theoretical results are ambiguous, but when the authors apply their findings to the case of interconnection between Germany, Belgium, France and the Netherlands, they conclude that wholesale prices are reduced.

It is hard to identify specific contributions of particular effects inside the overall framework of integration. The following articles contribute to the literature on electricity grid integration, while mentioning competitive effects as well. Nevertheless, their scope lies more on productive efficiency gains and security of supply.

(13)

13

Finon and Romano (2009) analyse interconnection effects from an entirely different perspective. The French market consists of 93% hydro and nuclear generation, low variable cost generation techniques. The integration of electricity markets poses a real problem with respect to consumers’ interests. They conclude, among others, that electricity market integration results in an increased electricity bill for consumers in countries enjoying a large share of power generation capacities at low variable costs when market integration occurs, whereas it does not result in decreased prices in neighbouring countries, contradicting the majority of the literature finding converging prices. Finon and Romano (2009) advocate that the interconnection capacity comes on top of the French demand curve, increasing the French electricity price, while interconnection capacity is not large enough to lower prices across the border (in Germany). Although Finon and Romano (2009) state that the price effect is not observed in the higher cost region, one could argue that increasing interconnections would lead to price reductions at some point. Therefore we conclude the literature review on interconnections with the notion that, if the interconnection capacity is large enough, electricity prices converge. By increased security of supply and the ability to optimally distribute low-merit-order generated power the volatility of the electricity price decreases as well.

3.3 Combining RES and interconnections: does volatility increase or

decrease?

In section 3.1 and 3.2 can be read that renewable power supply lowers the electricity price on average but results in high, weather dependent, variability, interconnections however can cause productive efficiency gains as low marginal costs produced electricity can be distributed to a larger market, together the effect becomes ambiguous. In November 2017 the European Commission Expert Group on electricity interconnection targets writes: “To achieve its climate and energy goals, Europe needs to improve cross-border electricity interconnections. Connecting Europe's electricity systems will allow the EU to boost its security of electricity supply and to integrate more renewables into energy markets.” Considerations made prior to this statement are the increasing weather dependence due to RES technology, while integration has positive effect on the security of supply by enlarging the efficiency of using supply portfolios and enlarging the market. The literature in section 3.2 often mentions potential for looking to the effects of grid integration in combination with RES effects as discussion point. This is a logical discussion as the combination of the two might lead to lower, yet stable, prices.

(14)

14

(15)

15

4. Theory

4.1 Electricity prices and volatility

In this chapter the price setting mechanisms in the electricity market are discussed and the consequences of increasing RES shares as well as interconnection levels are described. Prices in the wholesale market are based on the marginal cost of production, the willingness to pay of consumers and opportunity costs (Mulder, 2017), as electricity is traded across several forward markets. The electricity price depends crucially on the level of electricity demand and supply, which are fluctuating continuously. In figure 3, a typical supply curve is depicted. The marginal costs of production differ per technique, therefore the supply curve is called the merit order curve. Low marginal costs techniques are for example renewables and nuclear plants, while coal-fired and gas-fired power plants face significant fuel costs and belong to the high marginal costs techniques (Heptonstall, 2007, EIA, 2019). The lowest cost production technique is preferred over more expensive techniques, however electricity prices are often set by plants facing higher marginal costs (Tietjen, 2016), giving the low marginal costs techniques opportunities to remunerate on investments. Because low merit order production techniques lack either flexibility or controllability4_{and can adopt badly}

to the fluctuating demand, they are used to face the base load. When demand drops below the base load supply it is even more advantageous to sell electricity for negative prices than to screw down production, as is sometimes observed in Germany (Moraga, Mulder, 2018). As low costs marginal techniques face reliability problems, the electricity price is highly correlated with the fuel costs and the volatility of the electricity price thus depends on the volatility of gas and coal prices.

Figure 3: Electricity price setting with different generation techniques results in the merit order curve. The lowest marginal price setting technique has preference over higher marginal costs price setting techniques. Source: E&C

consultants (2014).

4_{Nuclear plants are inflexible: both technically (scaling up production leads to a supercritical chain reaction causing}

a meltdown and scaling down production causes a subcritical dying chain reaction (Martin, Shaw, 2019)) and economically (production must be high to remunerate on the high investment costs). Wind and solar are

(16)

16

Another determinant of the electricity price and price volatility is the total capacity of generation. When electricity demand exceeds total generation (and import) capacity another pricing setting regime is entered. At maximum supply, the supply curve is vertical, therefore an increase in (forecasted) demand only increases the price. This scarcity price allows all producers to obtain a mark-up and allows producers to get higher returns on investment. However, investment in generation capacity causes the duration of peak prices to be lower. Regulators require reserves to prevent power shortages, therefore selling electricity at scarcity prices is limited. The volatility of the electricity (wholesale) price, indicated by the standard deviation of the electricity price, depends not only on the price difference, but also on how often deviations from the mean occur. As scarcity prices are, in theory, unlimited and cause high variation in prices, the duration of scarcity is a determinant of the volatility of the electricity price, but in reality occur limited.

4.2 Price and volatility effects of increasing RES shares

With renewable power generation, supply fluctuations arise more frequent as the supply curve shifts in and out because of weather variations, resulting in an increase of the electricity price volatility. The fluctuations of the electricity price are dependent on the correlation between the weather and demand. En gros, supply of solar energy is correlated positively with demand in short term, as the sun shines during the day when most electricity consuming activities are operational, while it is negatively correlated in a long time span, seasonal effects dictate higher electricity demand during winter while days are shorter and solar production is limited (from a Western European perspective). Wind does not follow such a straightforward production cycle, but remains to be a variable power source. There is no evidence electricity demand rises when the wind blows, or not. As hydropower is a flexible source, hydropower supply is correlated positively with electricity demand, it is used to meet demand when necessary. Biomass has the issue that it is not abundant enough to be a main power source, but could certainly be used as a flexible power source (Pavić, Capuder, Kuzle, 2016). Therefore, by increasing the renewable energy share, specifically solar and wind, electricity price fluctuations are stronger and occur more often5_{(Mulder, 2017).}

The effects of increasing the renewable energy capacity share in one’s generation portfolio makes the electricity price more volatile, but the extent depends strongly on the installed portfolio in the market. At days when RES capacity is fully operational the merit order curve shifts to the right and electricity prices will be low, see figure 4. This is called the price effect as the share of renewable electricity increases (Hirth, 2013). However, when the weather does not allow for RES capacity to be operational the merit order curve shifts left again, increasing the electricity price ceteris paribus. Figure 4 and 5 describe the price difference for two arbitrary merit order curves, consisting of renewable, nuclear, coal-fired and gas-fired power generation. Indeed, renewable power increases the volatility as price differences arise with fluctuations in renewable supply. However, to which extent electricity prices differ is strongly dependent on the proportion of RES and steepness of the merit order curve, defined by the generation portfolio. Specifically, when fluctuations in RES supply cause a switch in price-setting generation technique, the price difference is large, causing higher price volatility. When renewables make a large part of the generation portfolio, switches

5_{Price volatility decreases from hour to hour, it increases from day to day. For our purpose, an annual and monthly}

(17)

17

between generation techniques are generally more likely to occur. This leads to the synthesis of the first hypothesis:

Hypothesis 1: The electricity price volatility increases when the renewable energy share rises in a country's generation portfolio.

Next to the hypothesis, the extent to which the variable RES capacity share influences the electricity price volatility will be examined. When a market is characterised by a small share of RES, volatility effects could be rather limited. However, volatility effects become increasingly apparent when the RES share is large. Therefore it is plausible that the RES capacity share has an exponential relationship with variations in the wholesale electricity price. Figure 6 depicts two relationships between the price volatility and the RES share, a linear relationship and a quadratic relationship. The nature of this relationship will be investigated empirically in this thesis.

Figure 4: Price differences subject to weather conditions at an arbitrary demand level, large price differences over time imply high price volatility. Operational production techniques are depicted below the supply curve in matching

(18)

18

Figure 5: The price difference is subject to the generation portfolio. At an arbitrary demand level, if (the absence of) renewable production doesn’t cause a transition of generation technology, price differences are limited and therefore the price volatility effect remains limited. Operational production techniques are depicted below the

supply curve in matching colour boxes.

Figure 6: Linear vs. quadratic relationship of increasing the RES share, where β1 denotes the price volatility due to

renewables with a fully renewable portfolio.

Two extra features having impact on the electricity price and the merit order are feed-in-tariffs (FiT) and CO2 pricing. The FiT, set in for example Spain, Germany, Denmark6_{, stimulates renewable generation}

capacity investments, but excludes renewable produced power from the market by promising a flat rate per

(19)

19

MWh. This causes a perverse incentive to keep renewable production high, even when prices are very low. Therefore the FiT increases the electricity price volatility.

CO2 pricing affects the structure of the merit order, as the marginal costs for fossil fuel intensive production techniques rise, the CO2 intensive part of the merit order curve is shifted upwards. CO2 pricing is implemented by CO2 taxation, or via auctioning schemes with a CO2 cap, as the ETS. The main price setting production techniques are affected by such measures, putting upward pressure on the electricity price. As coal-fired power plants emit more CO2 than their gas equivalents, the marginal costs lie closer to each other. Depending on the generation portfolio this could either increase or decrease the electricity price volatility. If switches from nuclear to coal or gas are made often in order to meet demand, the volatility increases as the price difference is higher, due to the upward shift of the fossil part of the merit order curve. However, if the demand is often met with coal when renewables are active, and additional gas-fired power plants need to run to meet demand at bad weather days, the volatility is less, see figure 7 (compared to figure 4).

Figure 7: CO2 pricing increases the marginal costs of coal-fired power production more than that of gas-fired power production, i.e. the price gap between coal and gas production diminishes. At an arbitrary demand level, a switch from coal to gas, by variability of renewable production, causes a limited price increase, depending on the CO2 price. Operational production techniques are depicted below the supply curve in matching colour boxes.

4.3 Price and volatility effects of increasing interconnection levels

(20)

20

the markets the electricity prices will converge, see figure 8. When production has low marginal costs in one market, market players will sell electricity in the more expensive market. Arbitraging of electricity prices will continue until one of the two following conditions is met: the electricity prices, corrected for possible transportation costs, in the markets are equal, or the interconnection capacity (IC) has reached it limit, it is congested (Neuhoff, 2004). Whether the interconnection on average increases or decreases the electricity price depends on the initial electricity price in the separated markets. This depends on technology differences between markets (Valeri, 2009), in the market where producers have larger competitive advantage, lower marginal costs, an upward price effect upon interconnecting will be observed. However this appears to be more of a rule of thumb when the time dimension is included, as commented on below.

Figure 8: Converging price effects of interconnecting markets. The low marginal costs supply technologies from market A partially serve demand in market B, depicted by a virtual shift.

(21)

21

decreases, called an increase in the supply security, and ultimately leads to less needed installed capacity to meet peak demand (Giesbertz, Mulder 2008).

The integration of electricity grids comes with two efficiency gains, both exerting downward pressure on the electricity price. An allocative efficiency gain arises as integrated markets face more competition. Even the threat of entry in a connected market, reduces the electricity price significantly, because the threat of entry becomes a threat of entry in all markets, causing each to produce nearer to competitive levels (Borenstein et al., 1997). Moreover, market integration improves productive efficiency (Hobbs, Rijker, Boots, 2005). Because of the interconnection, low merit order produced electricity can be used to meet a larger base load. The fluctuations in demand can therefore be coped with better. Especially as different markets face different demand curves, this effect can be substantial. A country with a relatively flat demand curve can distribute solar produced power to a country with more fluctuating demand, while vice versa nuclear power can be distributed more efficiently. This contradicts the representation depicted above slightly, as the time dimension is now included. Nevertheless, the productive efficiency gain stabilizes the electricity price and drives the price down, i.e. results in less price volatility.

However, when the interconnection is congested, the productive efficiency gain and reduction of peak load price duration fall away.7_{I.e. when the interconnection is congested low merit order produced electricity}

cannot be distributed further. It is therefore crucial that the interconnection has sufficient capacity. In reality these interconnections are often not of sufficient capacity. European countries have their own electricity grid with specific equilibrium prices, because interconnection levels remain limited. Nevertheless, in an empirical analysis of recent data a volatility decreasing effect on the electricity price is expected to be found, which leads to the second hypothesis:

Hypothesis 2: Higher levels of cross border capacity lead to lower volatility of the electricity price.

4.4 Combined RES and IC price and volatility effects

By arbitrage, electricity generated with low marginal costs in a connected market will be supplied rather than high merit order power in the local market. High merit order power becomes supervacaneous in that sense. As RES sources are characterised by almost zero marginal costs, power from these sources have priority over traditional power. With larger cross border capacity, RES generated power will have larger distribution opportunities. Green power can therefore be used more effectively. In Scotland operators of wind farms are paid large amounts by the government to stop producing at night, as the wind farms overproduce with respect to demand. In a larger market such situations might be avoided. As RES produced power is used more effectively, the electricity price will be lower. The more effective distribution of RES produced power increases the RES supply security, prices will therefore be lower on average and the electricity price volatility decreases.

7_{Borenstein et al. (1997) showed this is not the case for the allocative efficiency. As stated above, the threat of entry}

(22)

22

However, this distributional advantage is limited by weather correlations over space. At times that the sun is shining and the wind is blowing (or not) across the entire space of interconnected markets, all individual markets face peak RES supply (or no RES supply at all). Depending on the total demand and deployed RES capacity, the strength of this effect varies. Nevertheless, when integrating markets over a large enough spatial distance (for example the proposed European Supergrid (Toretti, 2014)) weather correlations and demand correlations diminish. Grams et al. define seven weather regimes designed to capture year-round, large-scale flow variability in the Atlantic-European region and show that the energy system could profit largely from exploiting the implications of these regimes for continent-scale wind generation patterns. Also Monforti et al. (2016) state that the non-homogeneous meteorological conditions across Europe cause different values and patterns of wind power generation. At any time, Europe is large enough to face multiple weather systems, as a consequence this can be exploited by better integration of trans-boundary power exchange in Europe. Thus, by expanding the region where renewable generation activities take place, total production could be balanced.

The same yields for the demand side of the electricity market. Across large space demand fluctuations may be balanced out. Moving across a longitudinal line, the sun rises and falls at different moment at different places. Therefore, time-dependent electricity demand curves differ. Peaks in demand are therefore spread out over a longer time span, while relatively being lower than for all separate markets individually. This is depicted in figure 9 for perfect sinusoid demand curves. One observes the peak aggregate demand to be less than 4, the sum of individual peak demands. For intra-day fluctuations this results in a (small) efficiency gain.

Figure 9: Representation of 24 hour perfect sinusoidal demand smoothing for 4 time zones. The peak aggregate demand is less than 4, the sum of individual peak demands.

(23)

23

temperatures in summer reach high values. To cool domestic and industrial venues, air conditioning is used. As these air conditioning systems use power, electricity demand increases. However, closer to the poles temperatures in summer are moderate and electricity demand doesn’t peak the way it does in warmer places. Here domestic air conditioning is not abundant. The cooling degree day in the south exceeds that of the north and therefore average demand in a fully integrated market stabilizes. On the other hand, in winter Nordic regions freeze out and demand for heating increases. Although not all heating is met with electric sources, one expects the electricity demand to increase. In warmer climates, heating is less of an issue, as they face lower heating degree days. In total, these phenomena balance each other. By integrating electricity markets, lower total capacity is needed as demand curves can be balanced, facing less fluctuating aggregate demand the electricity price will be less volatile.

Biomass and hydropower can also be exploited better upon integration of electricity grids and serve as excellent swing supply mechanisms. The integration of electricity grids cause better distribution of flexible renewable sources. Biomass can be burned closer to the excavation site. The obtained power can be distributed almost without costs (Borenstein et al., 2000), which saves transportation costs, increasing the efficiency of the system. Hydropower faces abundancy problems, as not every place is suited for hydroelectric dams, but can be used excellently as a power reserve. In this way fluctuations in demand or supply can be handled with a low merit order power source, reducing the electricity price and volatility, with respect to using traditional fossil-fuel plants.

The combination of renewable energy sources and an interconnected electricity grid thus leads to reliable electricity supply at a predominantly lower price. However, this is based on highly idealized models assuming power generation and distribution over large spatial zones without the infrastructure being ready in reality. Moreover, the use of biomass and hydropower as swing supply is not yet in practice. Nevertheless, this leads towards the third and last hypothesis:

Hypothesis 3: The combined effect of increasing RES shares and increasing interconnection capacity decreases the volatility of the electricity price.

(24)

24

5. Method

5.1 Econometric Model

In this chapter the method of the empirical research will be gone through. From the theory explanatory variables for the electricity price volatility are exposed. These are used in multiple models over different periodicities and variations in subject markets. A simplistic annual model is used to identify general effects of the explanatory variables and searches whether the effects can be extrapolated to additional markets. A less general angle is used for the monthly interval as it includes wind speed dependence and fluctuating transfer capacities to make an comparative analysis between two countries. The data will be described in section 2, after which the results of the econometric tests on the data will be reported in section 3 of this chapter.

The objective of this study is to find the effects of the renewable energy share and the interconnection capacity on the electricity wholesale price volatility. These effects are strongly dependent on the generation portfolio. It is therefore important to compare and combine analyses from several countries with different characteristics. The volatility of the electricity price depends on fuel cost fluctuations, demand fluctuations, CO2 price fluctuations, RES capacity and interconnection levels. Most are expected to have a positive influence on the price volatility, except for the interconnection level. Moreover the effect of CO2 volatility is ambiguous as a mitigating effect arises when a switch between coal-fired and gas-fired power generation is made often. To assess the three hypotheses,

Hypothesis 1: The electricity price volatility increases when the renewable energy share rises in a country's generation portfolio.

Hypothesis 2: Higher levels of cross border capacity lead to lower volatility of the electricity price.

Hypothesis 3: The combined effect of increasing RES shares and increasing interconnection capacity decreases the volatility of the electricity price.,

(25)

25

However, for monthly purposes the transfer capacity varies. The transfer capacity differs from the interconnection capacity as TSOs account for reliability margins. These reliability margin are amongst others subject expected loop flows and temperature, making it fluctuate throughout the year. In this way the transfer capacity thus varies, without additional capacity to be installed, and allows for setting up a fixed effects model for the monthly time span.

The period of interest is crucial, as volatility effects of RES differs per time span. For example, daily consumption patterns are in line with solar production, while over a year the two are anti-correlated. In order to restrain from these daily and seasonal effects, the annual price volatility is used, while these effects are incorporated in the monthly model by choosing slightly different explanatory variables.

The pooled OLS derives best estimates while ignoring individual effects. This is in principle no problem, as we are looking at the aggregate outcome of a dynamic market setting instead of behaviour. The consequence is that a Breusch-Pagan test should be conducted to find out whether the estimates will be efficient. A pooled OLS with heteroscedasticity still finds unbiased estimates. Possibly the pooled OLS model would need to be estimated with heteroscedasticity-consistent standard errors.

The annual pooled OLS models have the following structure:

𝑝𝑝𝑖𝑖𝑖𝑖𝑣𝑣 = 𝛽𝛽1 ∗ 𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖+ 𝛽𝛽2 ∗ 𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽3 ∗ 𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖∗ 𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽4 𝐷𝐷𝑖𝑖𝑖𝑖𝑣𝑣 + 𝛽𝛽5 ∗ 𝐼𝐼𝐶𝐶2 𝑖𝑖𝑖𝑖 𝑣𝑣 + 𝛽𝛽6 ∗ 𝐺𝐺𝐺𝐺𝑅𝑅𝑖𝑖𝑖𝑖𝑣𝑣 + 𝜀𝜀 [1]

and,

𝑝𝑝𝑖𝑖𝑖𝑖𝑣𝑣 = 𝛽𝛽1 ∗ 𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖2+ 𝛽𝛽2 ∗ 𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽3 ∗ 𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖∗ 𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖+ 𝛽𝛽4 𝐷𝐷𝑖𝑖𝑖𝑖𝑣𝑣 + 𝛽𝛽5 ∗ 𝐼𝐼𝐶𝐶2 𝑖𝑖𝑖𝑖 𝑣𝑣 + 𝛽𝛽6 ∗ 𝐺𝐺𝐺𝐺𝑅𝑅𝑖𝑖𝑖𝑖𝑣𝑣 + 𝜀𝜀 [2]

Where 𝑝𝑝_{𝑖𝑖𝑖𝑖}𝑣𝑣_{is the electricity price volatility in country i in period t, 𝑅𝑅𝑅𝑅𝑅𝑅}

𝑖𝑖𝑖𝑖 is the share of renewable energy

supply capacity to the total capacity, 𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖 is the interconnection capacity as a share of total capacity, 𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖∗

𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖 is the interaction term, 𝐷𝐷𝑖𝑖𝑖𝑖𝑣𝑣 is the demand volatility, 𝐼𝐼𝐶𝐶2 𝑖𝑖𝑖𝑖 𝑣𝑣 is the CO2 emission allowance price volatility

and 𝐺𝐺𝐺𝐺𝑅𝑅_{𝑖𝑖𝑖𝑖}𝑣𝑣_{is the gas price volatility. The 𝑅𝑅𝑅𝑅𝑅𝑅}

𝑖𝑖𝑖𝑖2 term in equation [2] represents the squared ratio of

renewable energy supply capacity to the total capacity. The annual model estimations allow to examine generic relationships between the price volatility and RES capacity and IC, by abstracting from seasonal effects, and makes it easy to expand the analysis to additional markets.

For monthly purposes the specification of the models differ slightly but is expected to generate more specific results. As for a monthly analysis seasonal effects kick in, the variable renewable energy supply share is multiplied by the standard deviation of wind speeds (WINDSTD). Moreover, variability in available interconnection capacity is measured by the Net Transfer Capacity (NTC). Dividing the NTC with the total generation capacity results in the Net Transfer Capacity Ratio (NCTR). The specifications of the models are given in equation 3 to 5, where the last two differ by using a linear or quadratic relationship between the variable RES capacity share and price volatility.

𝑝𝑝_{𝑖𝑖𝑖𝑖}𝑣𝑣_{= 𝛽𝛽}

(26)

26 𝑝𝑝_{𝑖𝑖𝑖𝑖}𝑣𝑣_{= 𝛽𝛽} 1 ∗ 𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖+ 𝛽𝛽2 ∗ 𝑊𝑊𝑊𝑊𝐼𝐼𝑅𝑅𝑖𝑖𝑖𝑖 + 𝛽𝛽3 ∗ 𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖∗ 𝑊𝑊𝐼𝐼𝑊𝑊𝑅𝑅𝑖𝑖𝑖𝑖+ ⋯ 𝛽𝛽4 ∗ 𝐷𝐷𝑖𝑖𝑖𝑖𝑣𝑣+ 𝛽𝛽5 ∗ 𝐼𝐼𝐶𝐶2 𝑖𝑖𝑖𝑖 𝑣𝑣 + 𝛽𝛽6 ∗ 𝐺𝐺𝐺𝐺𝑅𝑅𝑖𝑖𝑖𝑖𝑣𝑣 + 𝜀𝜀 [4] 𝑝𝑝𝑖𝑖𝑖𝑖𝑣𝑣 = 𝛽𝛽1 ∗ 𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖2 + 𝛽𝛽2 ∗ 𝑊𝑊𝐼𝐼𝑊𝑊𝑅𝑅𝑖𝑖𝑖𝑖 + 𝛽𝛽3 ∗ 𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖∗ 𝑊𝑊𝐼𝐼𝑊𝑊𝑅𝑅𝑖𝑖𝑖𝑖+ ⋯ 𝛽𝛽4 ∗ 𝐷𝐷𝑖𝑖𝑖𝑖𝑣𝑣+ 𝛽𝛽5 ∗ 𝐼𝐼𝐶𝐶2 𝑖𝑖𝑖𝑖 𝑣𝑣 + 𝛽𝛽6 ∗ 𝐺𝐺𝐺𝐺𝑅𝑅𝑖𝑖𝑖𝑖𝑣𝑣 + 𝜀𝜀 [5]

Equation 3 will be used to compare the effects of the RES capacity share and the intercept term between the Dutch and the German market. By including the wind speed volatility in the first explanatory variable renewable production fluctuations are simulated rather than the renewable capacity. With this extra specification allows stronger conclusions can be drawn based on difference in the results and exploiting the transfer capacities for the two markets. Model 4 and 5 will again be used to draw conclusions about the relevant estimates and to investigate whether the model still holds when pooling the data of the two countries.

The specification of the variables is as follows. The annual electricity price volatility is the standard deviation of the average daily electricity wholesale price over a period of a year. The monthly price volatility is the standard deviation of the average daily electricity wholesale price over a period of a month. The share of renewable energy source capacity is taken as the ratio of renewable energy techniques when fully operational to the total generation capacity. When using only the realised share of renewable energy production one might cause an endogeneity problem, as weather dependence enters the equation on the left hand side and on the right hand side. Indeed, this is more of a problem for shorter time spans as the aggregate weather characteristic (climate) is relatively stable over a year time. By choosing the renewable share a share of capacity (in MW) and not of production (in MWh), the effect of the RES share to the electricity price volatility is independent. Moreover, it is wise to define two types of renewable energy shares, namely the total RES share and the variable RES share, denoted as the corrected RES share RESCit as hydropower

and biomass don’t face flexibility problems. The variable RES share is also used for the monthly analysis. To increase the effectiveness of estimation this variable RES share is multiplied by the volatility of wind speeds. Also here, the volatility of wind speeds are taken as the standard deviation of the average daily wind speed over a month. As wind speeds are not subject to the volatility of the electricity price the product of the variable capacity share and wind speed volatility remains endogeneous.

(27)

27

can be wheeled through the interface between the two systems, which does not lead to network constraints in either systems. (ENTSO-E, 2000) The NTC becomes an important parameter when reducing the time frame to monthly.

Demand, or more specifically consumption, volatility causes an endogeneity problem, as it reflects the equilibrium price volatility directly by moving along the merit order curve. When demand rises the electricity price rises, but at the same time, when the electricity price rises demand diminishes at the rate of the price-demand elasticity. Though the relation is different across subject markets and not of constant proportion it is wise to seek for an explanatory variable that has a more independent character. By letting the exogenous cooling degree days (CDD) and heating degree days (HDD) be a proxy for demand volatility, direct effects are filtered out as the temperature is not influenced by the electricity price, obviously. CDD and HDD are quantitative indices demonstrated to reflect demand for energy to heat or cool buildings. With high values for both in a year electricity demand is thus higher than average. This reflects the volatility in the way that electricity demand peaks during certain times of the day, and therefore is explanatory for the price volatility, after all, on a warm day cooling is more necessary during the day, than at night. For monthly purposes the two are decoupled, to incorporate seasonal effects.

CO2 price volatility is regarded as an exogenous variable as the electricity price volatility does not affect the volatility of the price of emission allowances. However, there are several ways pricing the emission of CO2, for example by levying a tax or using a capped auctioning scheme such as the European emission trading scheme (ETS). For the purpose of this thesis the ETS prices are used. The consequence is however that possible local taxation could lead to an omitted variable bias, especially for countries where fossil techniques are continuously the price setting technique and levy an additional local tax.

The gas price volatility is chosen to be a measure of input price variability as gas-fired power plants face the highest fuel cost, making the technique to be the rightmost one in the merit order. Together with the flexible character of gas-fired power plants it makes that this technique is often the price setting technique. Nevertheless, when assessing volatilities care should be taken in the interpretation. Two things must be addressed, as the price setting method depends on multiple variables, a downward change in one variable together with an upward price shift in another variable results in no change in the dependent variable, thus theoretically the volatilities might cancel each other out. Moreover, gas-fired production techniques do not account as a 100% price setting technique, it could therefore be that an omitted variable bias arises as other input prices can also vary in time.

(28)

28

5.2 Data Description

The countries of interest are Spain, France, Germany, the Netherlands, Denmark and Norway because of market properties discussed below. An overview of characteristics is given in table 1, the second column alternately denotes realized renewable production and renewable supply capacity. These countries are of interest for several reasons. Firstly, for all countries sufficient data is available. Secondly, the share of renewable energy generation differs strongly between the countries of interest. The almost fully renewable power portfolio in Norway is in sharp contrast with the largely fossil Dutch portfolio. Also it is interesting that different countries in the dataset follow different trends regarding the renewable energy share, while Nordic countries have historically high shares of renewables, the share in other countries is rising in the investigated time frame. For the interconnectivity the dataset allows an alike analysis. The extremely low level of transfer capacity of Spain is in sharp contrast with the highly interconnected characteristic of the Danish market, where electricity supply actually relies on import as controllable production is replaced by import during the summer season when the heat demand is low and most thermal power plants are out of service for maintenance.

(29)

29

Table 1: Overview of electricity market characteristics for the purpose of this study. For category renewable share: low, moderate, high stands for <20%, 20%-50%, >50%, respectively, of total capacity. For interconnectivity low, moderate, high yields <10%, 10%-25%, >25%, respectively.

Country Renewable share Interconnectivity Demand

characteristics Traditional Supply Methods Spain8 _{Moderate share of}

renewables, 47% capacity in 2013: hydro and wind.

Extremely low interconnectivity (2% 2014).

Summer peaking

country. Petroleum and gas

France Moderate level renewables, 40% capacity in 2013, mostly hydro power.

Low level of interconnectivity (10% 2014).

Weather dependent demand. Electricity used for both cooling and heating.

High share of nuclear power

Germany Moderate share of wind energy; 27% of total generation in 2016. Low level of interconnectivity (10% 2014). Germany is a net energy exporter. Diminishing electricity consumption from 2008-2014. Winter peaking country.

Lignite and Coal

The Netherlands Low share of renewables, 6% generation in 2017 (most carbon intensive portfolio of EU). Moderate (17% in

2017). Winter peaking country. Coal and Gas

Denmark High share of renewables (60% capacity, of which 40% wind). Highly interconnected. In July 2017 40% of consumption import. Winter peaking

country. Fossil and wind (the latter already since 1970s)

Norway Traditionally high share of hydropower, 98% in 2014. Low level of interconnectivity (7% in 2014).

Much flexibility for demand response. Winter peaking country.

Hydropower

Sources: Deloitte market analyses (2014), smartgrid Norway (2014), energinet Denmark, Bach (2017), Mulder (2017), General Electric (2018), Agora Energiewende (2015), rte France (2014), Romero-Jord, del Río, Peñasco (2014).

(30)

30

To find yearly volatilities of the electricity price the standard deviation is calculated as a measure of dispersion. The data from Bloomberg was hourly indexed for each day for the entire timespan for each country. The average day-ahead price was calculated to find the price volatility benchmarked on day-to-day fluctuations. For Spain, France and Germany, of which data allow to estimate a model with an eight year time span, the price volatilities are shown in figure 10. The data from the Nordpool group was daily indexed per year, allowing to calculate the price volatilities directly. A description of the average electricity price and the annual price volatility for all countries is given in figure 11 and 12, respectively. For Denmark an extreme value for the volatility in Denmark in 2013 is observed, this is the product of extreme electricity prices on June 7 of up to 2000 €/MWh because production capacity was unable to enter the market (DERA, 2014). When deleting this outlier on June 7, the electricity price volatility in Denmark drops from 22.6 to 8.8. When one compares the average and the volatility of the wholesale day-ahead spot prices one observes that fluctuations of the electricity price are rather significant. For monthly purposes the Dutch and German electricity markets are examined, the price volatilities using daily prices are shown in figure 13.

(31)

31

Figure 11: The average electricity wholesale day-ahead spot price for Spain, France, Germany, the Netherlands, Denmark and Norway. Sources: Bloomberg, Nordpool.

(32)

32

Figure 13: Monthly price volatilities in the Netherlands and Germany measured as standard deviation of the daily electricity price. Source: Bloomberg.

(33)

33

Figure 14a: Renewable power capacity shares for Spain, France, Germany, the Netherlands, Denmark and Norway as obtained from Eurostat by calculating the ratio of renewable power capacity to full generation capacity.

Figure 14b: Renewable power capacity shares corrected for hydropower and biomass power generation for Spain, France, Germany, the Netherlands, Denmark and Norway as obtained from Eurostat by calculating the ratio of corrected renewable power capacity to full generation capacity. For Spain and France the miniature amount of tidal power is incorporated, in other countries tidal power generation is not deployed.

(34)

34

Service using the station in Offenbach, where the official institute is located. Again the volatilities are measured as the standard deviation of the average daily wind speed in a month. The wind speed volatilities are plotted in figure 15. The average wind speed volatility in the Netherlands is 1.26 m/s and in Germany 0.96 m/s and will be used to calculate marginal effects.

Figure 15: Wind speed volatilities for the Netherlands and Germany measured as the standard deviation of average day wind speeds at De Bilt and Offenbach, respectively. Sources: KNMI, German Weather Service.

The interconnection levels are obtained from a report in 2015 from the European Commission “Achieving the 10% electricity interconnection target”. The data reports only the interconnection level in 2014, and is based on information from ENTSO-E, Scenario Outlook and Adequacy Forecast 2014. Unfortunately contact with ENTSO-E has not resulted in a time series dataset, and the electrical interconnection level of 2014 is used for all years. As expansions of interconnection capacities are limited, this suffices. The EC does not report data on the interconnection level of Norway, the value of interconnection for Norway is therefore calculated on its own. Norway is connected to the Nordic market and the Netherlands, with a capacity of 1700 MW and 700 MW respectively (IEA, 2016). In 2014, the total capacity of Norwegian producers was 33,144.000 MW (Eurostat), which results in an interconnection level of 7(.24)%. The interconnection levels of all countries in question are given in table 2, in the model the interconnection levels are adapted to a number equivalent between 0 and 1.

(35)

35

of the respective market the NTCs are divided by the total generation capacity. The average NCTR for the Netherlands is 0.134 and for Germany 0.055 and will be used to calculate marginal effects.

Table 2: Interconnection levels in percentages in 2014.

Spain France Germany the Nether-

lands Denmark Norway

Interconnec-tion level 2% 10% 10% 17% 44% 7%

Sources: EC (2015) and IEA (2016).

Figure 16: Monthly Net Transfer Capacity ratios for the Netherlands and Germany. Sources: TenneT, TransnetBW.

The Cooling Degree Days and Heating Degree Days were retrieved from Eurostat on the 3rd_{of May 2019,}

(36)

36

demand country, the HDD values exceed that of the CDD. An overview of the data is presented in appendix A1 and A2, in order to set all explanatory variables at approximately the same order of magnitude in the model the HCDD values given in appendix A1 are divided by 1000.

For the CO2 price volatility data from the European Climate Exchange OTC 1 Year Forward CO2 Emission EU ETS Px, reported by Bloomberg, are used. Note that Norway is a part of the European Economic Area and is linked to the ETS. The dataset runs from 2010 (with exclusion of the first week of January) to 2018. Also here, forward prices in weekends and public holidays do not exist. The values of the volatility are low for 2012 to 2017, this is a reflection of low prices for emission rights of ~5€/tonne, as the price before and after this period is significantly higher, up to 24€/tonne in December 2018, this is reflected in the volatility. All values are reported in appendix A3 and A4.

The gas prices are also obtained from the Bloomberg terminal which reports among others the PEGAS PEG (formerly PEG Nord) Spot - Day-Ahead prices. It enables to find daily prices with the exclusion of weekends and public holidays. Compared to data from Eurostat, where bi-annual gas prices are reported, this enables to find reasonable variance for the timespan of one year. As the gas prices in France is the longest dataset available, from 2010 to 2018 compared to starting from 2013 for the Netherlands and Germany, it is used as the independent variable for all gas price volatilities for annual purposes. For the monthly purpose the gas price volatility is country specific. The values are also given in appendix A3 and A4. The annual gas price volatility, with a typical value of about 2.5, is lower than the volatility of the wholesale day-ahead spot price volatility, which hints that indeed more than the marginal costs of gas-fired production determine the electricity price volatility.

5.3 Econometric tests in Stata

Before estimating the models the data should be tested. For panel data, non-stationarity could lead to biased results. To test for a unit root the Augmented Dicky Fuller test is used, the test statistics are listed in Appendix table B1, B3 and B4. For annual purposes from 2010 to 2017, for the countries Spain, France and Germany (SFG model), almost no variable is stationary according to the test. Only for the HCDD and the gas price volatility the null hypothesis that a unit root is present can be rejected. For the RES capacity share and the corrected RES capacity share this is of course no surprise, with present-day policy of increasing RES capacities. Subsequently a co-integration test as suggested by Engle and Granger (1987) is executed. The results reject the null hypothesis of no co-integration for Spain, France and Germany to at least a 10% level for both the RES and the RESC estimation results, see Appendix table B2, hence the results will not be spurious. Only in the case of Spain using RES and the quadratic dependency models for France the test statistic fails to reject the null hypothesis of no co-integration. For the analysis from 2013 to 2017 for all six countries, most variables show non-stationarity as well. A co-integration test could not be run as the number of variables exceed the number of years, however by inspection of the data a positive trend is observed for the price volatility and the RES capacity share and co-integration is assumed.

The effects of renewable energy sources and interconnection capacity on electricity wholesale price volatility