1
Does renewable energy generation impact the electricity
prices in Germany in the same way as it does in the
Netherlands?
Abstract
This paper assesses the impact of renewable, weather depended, energy resources on the day-ahead electricity prices in Germany. We take our lead from Mulder and Scholtens (2013), who studied the effect of German & Dutch wind/solar power on the Dutch electricity prices. The focus period for the present study is from January 2015 up to January 2017 and uses hourly data. There is no evidence found that energy generated by photovoltaic panels influences electricity prices; daylight duration and sun intensity do not appear to be factors. We do find that the average wind speed has a significant negative effect on the German day-ahead prices. These conclusions are partly in accordance with the Dutch results of our leading paper Mulder and Scholtens (2013). In Germany, the effect of wind power on electricity prices is significantly greater than the effect of wind power on electricity prices in the Netherlands. This could be explained by the increased wind turbine capacity compared to our leading paper.
Research Paper MSc Finance
Rudy van Deventer S2765063
Supervised by dr. prof. T.K. Dijkstra
University of Groningen
Faculty of Economics and Business Duisenberg Building Nettelbosje 2
9747 AE Groningen
Word count: 8899
This thesis will also be used for the Energy Certificate
2
1. Introduction
Renewable energy has become more and more important in the electricity market due to the environmental impact of fossil fuels. Many countries implement policies to stimulate renewable energy which leads to a larger share in renewables in their energy mix (Mulder & Scholtens, 2013). In Germany, they adopted the ‘’Erneuerbare-Energien-Gesetz’’ which are several German laws that provide a feed-in tariff scheme for renewables (Cludius et al., 2014). Feed-in tariff schemes allow renewable energy producers to sell their electricity at a fixed price (KleFeed-in et al., 2008). According to the Bundesministerium für Wirtschaft und Energie (2016) 31.6% of Germany’s energy consumption in 2015 was generated from renewable energy sources. In 2016 Germany was the third largest country in the world with almost 50 GW/h of power when it comes to installed wind energy capacity and the second when it comes to solar power with more than 40 GW/h of installed capacity (Fraunhofer Institute, 2017).
3 cross-border transmission capacity between two countries. Therefore, electricity can only be sold to a limited extend abroad.
Conducting research on the costs and benefits of increased renewable capacity is an important element of economic research (Würzburg et al., 2013). It is important to understand whether and how renewable energy generation effects electricity prices since renewable energy is becoming increasingly more important in our society and our society depends strongly on them.
This study investigates the reaction of the day-ahead wholesale prices of electricity in Germany to the influence of renewable electricity generation. It will investigate, through an approach that is based on Mulder & Scholtens (2013) who study this effect for the Netherlands, what happens to the wholesale prices of electricity when the influence of renewable energy increases as a result of an increase in wind speed and sunshine duration/intensity. There already has been done research on this subject (e.g.: Mauritzen, 2011, Mulder & Scholtens, 2013, Weber and Woll, 2007), therefore this study will try to strengthen the conclusions of earlier research. The general research question is:
Does renewable energy generation impact the electricity prices in Germany in the same way as it does in the Netherlands?
We will split this research question into two hypotheses.
1: 𝐇𝟎: Renewable energy generation does not have a significant impact on the electricity
prices in Germany.
2: 𝐇𝟎: Renewable energy generation does not have the same impact on the electricity
prices in Germany as it does in the Netherlands.
In a nutshell, this study finds that the average wind speed has a significant negative effect on the German day-ahead electricity prices. There is no evidence found that energy retrieved by photovoltaic panels, both duration of the daylight as sun intensity, have an effect on electricity prices. These conclusions are partly in accordance with the Dutch results of Mulder and Scholtens (2013). In Germany, the effect of wind power on electricity prices is significantly greater than the effect of wind power on electricity prices in the Netherlands.
4
2. Literature review
In this chapter, earlier literature about our subject is discussed. This chapter is divided into two parts. The first part discusses simulation based studies and the second part discusses empirical based studies.
2.1 Simulation based studies
Hirth (2013) concluded that the merit-order effect, that is the effect that occurs when the production of renewable energy with low marginal costs shifts the merit-order to the right, depresses the electricity prices when renewables produce electricity. They also conclude that this effect becomes stronger if the number of renewables in the electricity mix becomes more dominant. Another study about the merit-order effect is done by Sensfuß (2008). They used the calibrated PowerACE model to run a simulation that compares a situation with and a situation without renewable electricity production. They conclude that in Germany in the year 2006 the reduction of the unweighted average price reaches 7.8 €/MWh.
Weber and Woll (2007) also studied the German electricity system for the year 2006. They incorporated 34 technologies for electricity generation, prices of other energy goods and CO2 permits. They concluded that the electricity prices in Germany are 4.04 €/MWh higher when there is a no-wind scenario compared to a normal wind production scenario. A study done by Bode and Groscurth (2006) on Germany conclude a price drop of 0.5 to 0.6 €/MWh per GWh of additional renewable production and suggest that conventional thermal power stations with high marginal costs are forced out of the market due to the lowered electricity prices.
Holttinen (2004) determined that, based on a study that ran a series of simulations on increasing amount of wind power in the Nordic electricity system, high penetrations of wind power decreases the spot market prices with 2 €/MWh per 10 TWh/annum of added wind production on the Nordpool (which is a Scandinavian electricity market platform for multiple countries).
2.2 Empirical based studies
Cludius et al. (2014) used a time-series regression analysis and showed that electricity
generation by wind and solar panels reduced spot market prices in Germany considerably by 6 €/MWh in 2010 rising to 10 €/MWh in 2012. They also used their results to build a near-term forecasting tool for merit-order effects in Germany. It projected the result that the price effect would reach a reduction of 14-16 €/MWh in 2016.
5 the day-ahead prices. He also argues that because the wind power data is regressed on the system prices for the entire Nordic market, thus not only for the Danish market, that it is economically quite significant. Thus, the effect could become stronger when wind speeds are regressed against single country prices, as we will do in the case of Germany.
A study done by Gil et al. (2012) uses actual ex post wind power and electricity price data from Spain to calculate how much the market penetration of wind generation influences the Spanish electricity prices. This study, using an econometric inference approach, shows an 18% drop in prices (from 54.6 €/MWh to 44.9 €/MWh) with respect to a hypothetical no-wind industry model for the period 2007-2010. Weigt (2009) finds that the average electricity price becomes 10 €/MWh lower in Germany due to wind energy during their observation period. They also state that wind generation leads to a significantly lower market price during peak periods. Stefano et al. (2015) researched the merit-order effect on the Italian power market where they looked at the impact of solar and wind generation on national wholesale electricity prices. They conducted their research over the period 2005–2013 and concluded that an increase of 1 GWh in the hourly average of daily production from solar and wind sources, has reduced the wholesale electricity prices by 2.3€/MWh and 4.2€/MWh, respectively.
Mulder and Scholtens (2016) argue that the renewable energy producers connected to the German electricity market have a moderate impact on the Dutch electricity market. When renewable electricity production is high in Germany, the day-ahead electricity prices in the Dutch market are reduced. They conclude that the price elasticity of wind is about -0.03 when the cross-border capacity is not fully utilized. Another study done by Mulder and Scholtens (2013) also concludes a price elasticity of German wind power on the Dutch day-ahead prices of about -0.03. Furthermore, they also find that the Dutch windspeeds have an even more modest effect of below 0.01% on the Dutch electricity prices. In the same study, they researched the effects of electricity generated with solar panels on the Dutch day-ahead prices. They incorporate the variables length of the daylight and the intensity of the sunshine in their study. They do not find a robust significant connection between these two variables and the Dutch day-ahead prices.
6
3. Methodology
In this chapter, we outline the methodology that we used in our paper. In the first part of this chapter we are going to form our hypotheses. This is followed by a description of our
variables that we use to build our model. We close this section with discussing the statistics analysis that we have used for this paper.
3.1 Hypotheses
This study follows the methodology used by Mulder & Scholtens (2013) in their research about the impact of renewable energy on electricity prices in the Netherlands. Where they investigated the impact on electricity prices in the Netherlands, this study will do so for Germany. Therefore, one of the hypotheses of this study will compare the results from those two countries. Based on the described research, two main hypotheses can be formulated as follows:
1: 𝐇𝟎: Renewable energy generation does not have a significant impact on the electricity
prices in Germany.
2: 𝐇𝟎: Renewable energy generation does not have the same impact on the electricity
prices in Germany as it does in the Netherlands.
Mulder & Scholtens (2013) finds that electricity generated by wind, both German as Dutch windspeeds are considered in their paper, has a negative impact on the Dutch electricity prices. The connection between solar energy and the Dutch electricity prices could not be established. The share of renewable energy in the German energy mix has almost doubled compared to their researched period. Therefore, it is to be expected that the impact of renewable energy on the German price of electricity is bigger than it was on the Dutch electricity prices.
3.2 Model
7
3.2.1 Historical electricity day-ahead prices
Historical German electricity day-ahead prices are made available by the European Energy Exchange (EEX) on an hourly level. Prices used for this paper are an average of the prices during the time frame of an hour and are expressed in natural logarithms. A graphical representation of the German day-ahead prices is depicted in figure 3.1.
The average day-ahead price dropped from €31.63 in 2015 to €28.98 in 2016. This could be explained by the merit-order effect, that is described in our literature review, which is caused by the expansion of the capacity of renewables in Germany from 96.62 GW in 2015 to 103.15 GW in 2016. Especially the amount of wind mills, who influence the merit-order significantly, increased with 11%. More about wind energy will be discussed in chapter 3.2.5. On average, the day-ahead price was 29% higher during peak-hours compared to off-peak hours in 2015. In 2016 this fell by 2% to a 27% difference. A possible cause for this drop could be the increased capacity of photovoltaic panels. Photovoltaic panels only produce during daytime, which largely coincide with peak hours. Hence, solar energy can be used during the peak hours to smooth out load curve peaks (El-Khattam & Salama, 2004). Thus, increasing the supply during peak hours could depress the peak day-ahead prices, making the difference between peak-hours and off-peak hours smaller. More about solar energy will be discussed in chapter 3.2.6.
€-20 €-10 €0 €10 €20 €30 €40 €50 €60 €70 I II III IV I II III IV 2015 2016 P rice p er MW /h Quarters Fig. 3.1
8
3.2.2 Tightness of the market (D)
The tightness of the market is one of the factors that determine how much an electricity producer can raise its price when a higher demand occurs. The higher the demand, the less capacity remains to respond to further demand increases. We calculate this effect by taking the daily average level of residual demand minus the demand met by decentralized production. The residual demand can be calculated by taking the domestically centralized generators plus imports minus exports. This is equal to the net load of the energy grid. Another way to view this: the net load of an electricity grid is the amount of power in the grid when demand and supply are equal. The hourly data is derived from the transparency platform of the ‘’European network of transmission system operators for electricity’’ (Entsoe) and is expressed in natural logarithms.
During our focus period the average German hourly net load was 54.66 GW and stayed relatively stable during this period. Roughly, the net load was 25% higher during peak-hours compared to off-peak hours. Our leading paper expects that a lower demand and a higher supply met by decentralized production units both have a negative influence on electricity prices (see appendix A). Although this is generally true, the relation between demand (and therefore also the supply) and day-ahead prices could sometimes be the other way around. Market makers could adjust electricity prices downwards to send signals to the power producers when they are producing too much (oversupply) and to consumers that they consume too little to restore the balance in the electricity grid (EPEX Spot, 2017). Thus, increasing demand could cause the net load to go up and could theoretically mean lower prices.
3.2.3 Intensity of competition (C)
The intensity of competition can be measured by the Residual Supply Index (RSI). In our case, it will measure the aggregate supply capacity that remains in the market after subtracting firm i’s capacity, relative to the total demand. The formula for the RSI is:
RSI𝑖 =
∑ CAP𝑗− 𝐶𝐴𝑃𝑖
𝑛 𝑗=1
𝑇𝐷 (3.1)
9 of the largest firm on the electricity market and TD is the total demand (Sheffrin, 2002). The total demand is measured, just like in chapter 3.2.2, by the actual domestic generation plus import minus exports. The total capacity is measured by the sum of domestic generation capacity and import capacity which is equal to the net load of the German energy grid. The lower the RSI, the more market power a firm has. Also, the RSI is expressed in natural logarithms. According to Möst & Genoese (2009) it has to be remembered that the RSI only offers the necessary conditions for the exercise of market power; it is not a sufficient condition for the exercise of market power.
According to the data of the Bundesnetzagentur (2017), a department of the German Federal Ministry of Economics and Technology, 197 GW of total electricity generating capacity was installed by the end of 2016. This was 7.56 GW capacity more compared to 2015. The energy producer with the biggest installed capacity is the RWE Group which had an installed capacity of 23.86 GW in 2015 which grew to 23.96 GW in 2016 (RWE, 2017). A small footnote has to be made that the claimed capacity of the RWE Group only consists of power generators bigger than 10 MW since power generators smaller than 10 MW do not have to be reported to the Erneuerbare Energien Gesetz (Bundesnetzagentur, 2017) and are not available to us. Thus, the actual capacity of the RWE Group and therefore the RSI will depict a slightly higher value than it actually is. The hourly data for this variable also is derived from the transparency platform of Entsoe and reported in natural logarithms. In 2016 the RSI grew from 3.54 in 2015 to 3.67 which indicates that the market power that the RWE Group could exercise slightly dropped. This is in line with findings of the Bundesnetzagentur (2016) who state that the market power of the largest electricity producers has decreased significantly over the last few years. Sheffrin (2002) indicates that the RSI has a negative effect on the markup of the day-ahead price.
3.2.4 Marginal costs of production hard thermal coal (F)
10 electricity production in Germany is relatively small with 12.4% (4th place, after uranium). For that reason, we want to find another energy source as a reference point for the marginal costs of production. In 2016 more than 40% of the in Germany produced electricity is produced by the two coal sources lignite and thermal hard coal (Destatis, 2017). Lignite is with 23% of the total produced energy the biggest energy source after the renewables. This is followed by thermal hard coal with 17% of the total electricity production. Unfortunately, there is no public market for lignite; Lignite has a low calorific value which makes transport uneconomic over longer distances. This has the effect that the cost of lignite per MW/h including the transport, would become higher than hard thermal coal which is its biggest competitor. As a consequence, lignite mines often are located next to the powerplant that uses it as an energy source and often form one economic entity with the powerplant (Euracoal, 2017). For this reason, we will use the second biggest conventional energy source, thermal hard coal, as a reference point for the marginal costs of production since this is partly traded on a public market (Vattenvall, 2016). According to Trüby & Paulus (2012) Indonesia has become one of the biggest thermal coal exporting countries in the world. Thus, we will use the thermal coal prices published by the Indonesian government. These prices are noted as €MH/h, are expressed in natural logarithms, are only available on a daily level and are collected through DataStream.
In the first half of 2015 the average price per MW/h was €47.50, which fell to €37.99 in the first half of 2016 and ended at €59.25 on 30 December 2016. We expect a positive relationship between the prices of hard thermal coal and the German electricity prices.
3.2.5 Wind energy (W)
11 who work with a cut-in speed of 3-4 m/s and 5 m/s. For this study, we use an average based on the earlier mentioned studies of 3.2 m/s. Mulder & Scholtens (2013) use a cut-out speed of 24.5 m/s which comes close to the 25 m/s that is used by Markou & Larsen (2009) and Feng & Shen (2012). Hence, we will use the same cut-out speed of 25 m/s. Thus, when the wind speed lies out of this region, it is corrected to 0. Hourly data on wind speed is derived from the German metrological institute DWD and is based on the average wind measured across five different locations (Berlin-Tegel, Flensburg, Hannover, Dusseldorf and München). As Mulder & Scholtens (2013) take cube of the wind speed to convert wind speed to wind energy, since kinetic energy is half times the mass times the square of speed and that mass per second is proportional to speed, we will do the same. This is in line with other studies like Chen & Spooner (2001). The measured wind speeds are expressed in natural logarithms.
The year 2016 was less windy with an average measured wind speed of 3.52 m/s (after corrections for the cut-in and cut-out speed) than 2015 which had an average wind speed of 3.78 m/s. It is to be expected that the speed of the wind has a negative impact on the German day-ahead prices. Furthermore, the wind speed data will be checked by replacing the wind speeds by actual data on wind production (data is published by the German TSOs) in the model and compare them to the impact of our measured wind speeds. We will do this in the robustness check in chapter 5.
3.2.6 Daylight (S) and Sun intensity (SI)
12 across five different locations (Berlin-Tegel, Kiel, Hannover, Dusseldorf and München). Both the variables are not expressed in natural logarithms because these values often are zero.
The year 2015 was slightly sunnier than 2016 with 11 minutes per hour and 11.19 minutes, respectably. These variables will be checked by replacing them by the actual data on solar production (this data is also published by the German TSOs) in the model and compare them to the impact of our measured solar variables. We will do this in the robustness check in chapter 5. Since the literature is mixed about the effects of the sun on the electricity prices, it is hard to make a prediction about the influence of the sun on the German electricity prices.
3.2.8 Temperature of river water (RTR)
Many countries, including Germany, impose environmental restrictions on thermal powerplants that use river water for cooling purposes. In Germany those restrictions clamp down the energy sector when the temperature of the river water exceeds 23 degrees Celsius (Müller, 2007). When this threshold is exceeded, powerplants are forced to reduce their power production for environmental reasons. Hence, the supply curve will shift to the left. This variable is based on an average minimal day temperature measured across three major rivers in Germany: the Donau (made available by Gewässerkundlicher Dienst Bayern), the Main (which is an offshoot of the Rhine and also made available byGewässerkundlicher Dienst Bayern) and the Wezer (made available by Weser-Datenbank). This data is only made available on a daily basis. Since temperatures of major rivers do not vary much from hour to hour, we assume that the measured daily temperature is applicable to every hour of the day. Thus, transforming it into an hourly variable. The variable is measured in the amount of degrees Celsius above 23 degrees Celsius. Also, this variable cannot be transformed into a natural logarithm because it often is zero.
In 2015, 55 days had a river water temperature above 23 degrees compared to 31 days in 2016. Mulder & Scholtens (2013) find a positive but insignificant connection between the temperature of the river water and the Dutch electricity prices. We also expect a positive relationship between the temperature of the river and the German electricity prices.
13
Table 3.1
Correlation Matrix Amount of observations: 17544
EEX Coal Netload RSI RTR SUN (S) SUN (I) Wind
EEX 1 Coal 0,12 1 Netload 0,53 0,05 1 RSI -0,56 -0,06 -0,96 1 RTR 0,09 -0,16 -0,04 0,02 1 Sun (S) 0,02 -0,04 0,26 -0,29 0,11 1 Sun (I) 0,13 -0,03 0,37 -0,45 0,08 0,65 1 Wind -0,30 -0,01 0,31 -0,33 -0,06 0,14 0,25 1 Table 3.2
Descriptivestatistics Amount of observations: 17544 EEX Coal Netload RSI RTR SUN (S) SUN (I) WIND
Mean 30,30 45,10 54,66 3,61 0,32 30,71 11,33 129,78
Maximum 104,96 76,75 76,21 6,18 7,90 60,00 60,00 2844,28
Minimum -130,09 35,00 31,45 2,45 0,00 0,00 0,00 0,00
Std. Dev. 12,64 8,53 9,89 0,69 1,20 29,18 16,83 196,43
Adding these factors together yields the following regression:
𝐿𝑜𝑔(𝑝) = 𝛽0+ 𝛽1𝐿𝑛(𝐷) + 𝛽2𝑙𝑛(𝐶) + 𝛽3𝐿𝑛(𝐹) + 𝛽4𝐿𝑛(𝑊) + 𝛽5(𝑆𝐼) + 𝛽6(𝑆) + 𝛽7(𝑅𝑇𝑅)
+ 𝛽8𝐷𝑢𝑚𝑚𝑦 𝑆𝑢𝑛𝑑𝑎𝑦 … . . +𝛽14𝐷𝑢𝑚𝑚𝑦 𝐹𝑟𝑖𝑑𝑎𝑦 + 𝜀
3.4 Used statistical methods
The model is based on hourly data for the years 2015 and 2016. We follow the statistical tests that are used in our leading paper and we make necessary adjustments to the standard errors to correct for the found heteroscedasticity and autocorrelation.
3.4.1 Test on stationary
The first test that will be performed is a test in order to find if our time series variables are stationary; have a constant mean, constant variance and the covariance between any 𝑦𝑡 and
14 largely comparable with the results of Mulder & Scholtens (2013) who find that all their data passes this test. The results of these tests are available when requested.
3.4.2 Heteroscedasticity
The second test that will be performed is a test for heteroscedasticity; that is the case when the variance of a series is not constant over time (Brooks, 2014). In order to test for heteroscedasticity, we use White’s Heteroskedasticity Test on our data. If one starts with the following regression model (in this example it is a model with two variables):
𝑦𝑡= 𝛽0+ 𝛽1𝑥1𝑡+ 𝛽2𝑥2𝑡+ 𝑢 𝑡 (3.2)
White’s test asks whether the variance of û2𝑡 (estimated regressive residue) is related to its
regressors. Whites test is then carried out as an F-test on the significant of the coefficient in the following regression:
û𝑡2𝛼0+ 𝛼1 𝑥1𝑡+ 𝛼2 𝑥2𝑡+ 𝛼3 𝑥1𝑡2 + 𝛼4 𝑥2𝑡2 + 𝛼5𝑥1𝑡𝑥2𝑡 + 𝑣𝑡 (3.3)
Where the null hypothesis contains five restrictions:
H0: 𝛼1 = 𝛼2 = 𝛼3 = 𝛼4 = 𝛼5 = 0
There appears to be strong evidence that in our case we can reject the presence of homoscedasticity. We accept that our residuals are heteroscedastic. Results from Whites test are depicted in table 3.3.
Table 3.3
White Heteroscedasticity Test & Breusch-Godfrey test Amount of observations: 17544
Value P- Value
White 1441.230 0.0000
15
3.4.3 Autocorrelation
In the third test, we will check if there is evidence for autocorrelation; that is the case where error terms are correlated through time. In this case, we will perform the Breusch-Godfrey test which is carried out as an F-test for significance of the coefficients in the following regression. With k regressor and r lags (Brooks, 2014).:
û𝑡 = γ0 + γ1x1𝑡+ γ2x2𝑡+ … + γ𝑘x𝑘𝑡 + ρ1û𝑡−1+ ρ2û𝑡−2… + ρ𝑟û𝑡−𝑟+ v𝑡 (3.4)
The null hypothesis for this test is:
H0 = ρ1 = ρ2 = ⋯ = ρ𝑟
We perform the test with 1 lag. We can conclude that according to the test, it is likely that our data is autocorrelated. The results from this test are depicted in table 3.3.
3.4.4 Newey-West Robust standard errors
As we just showed it is likely that the residuals are autocorrelated and heteroscedastic. Hence, this can lead to incorrectly calculated standard errors for the estimated parameters (Brooks, 2014). Hence, the subsequent tests could be incorrectly calculated as well. To tackle these obstacles, the Newey-West robust standard errors are introduced to our tests and applied when calculating our p-values. These standard errors (also called heteroskedasticity and autocorrelation consistent standard errors) are calculated as follows (Newey & West, 1987):
Var(𝛽̂|X)̂ = { ∑( 𝑛 𝑡=1 𝑋𝑡− 𝑋̅)(𝑋𝑡− 𝑋̅)′} −1 x {∑ û𝑡2( 𝑛 𝑡=1 𝑋𝑡− 𝑋̅)(𝑋𝑡− 𝑋̅)′ + ∑ ∑ 𝑤𝑙 𝑛 𝑡=𝑙+1 û𝑡û𝑡−𝑙(𝑋𝑡− 𝑋̅)(𝑋𝑡−𝑙− 𝑋̅) 𝐿 𝑙=1 ′} (3.5) 𝑥 {∑( 𝑛 𝑡=1 𝑋𝑡− 𝑋̅)(𝑋𝑡− 𝑋̅)′} −1
where 𝑤𝑙 is the default lag length:
𝑤𝑙= [4( 𝑇
16
3.4.5 T-test
The statistical test that we are going to use to test the significance is the two-sided t-test. This t-test is defined as:
t-value = 𝛽̂ − 0
𝑆𝐸(𝛽̂) (3.6) where 𝛽̂ is the estimator of the coefficient and 𝑆𝐸(𝛽̂)is the Newey-West robust standard error that we explained in 3.4.4. We want to test against a significance level of 5%. The t-statistic has to be greater than 1.96 or below -1.96 in order to be significant and to reject H0. If the t-statistic is located between 1.96 and below -1.96 the relation between the tested variables can be interpreted as insignificant and H0 cannot be rejected.
4. Results
Table 4.1 presents the output of the regression of the variables during 2015 & 2016. It also shows standard errors, the results of the t-tests and their corresponding p-values.
As table 4.1 shows all the economic factors have a significant influence on the German day-ahead electricity prices. The price of coal follows our expectations and has a positive relationship with the electricity prices. Since this variable is expressed in a natural logarithm we can interpret it as elasticities of each other. Hence, during the total focus period a 1% increase in the price of gas will have rose the electricity price with 0.16%. A similar relationship can be established when we look at table 4.2. In the first half of 2016 coal has a significant impact of 0.65 on the electricity prices. This increases even further to 2.21 in the second half of 2016.
17
Table 4.1
Regression outputs complete focus period Amount of observations: 17544
Coefficient Std. Error t-Statistic p-value
Coal 0,161 0,049 3,249 0,001 Netload 0,778 0,313 2,487 0,013 RSI -0,366 0,111 -3,312 0,001 RTR 0,036 0,005 7,802 0,000 Sun duration 0,002 0,000 6,147 0,000 Sun intensity -0,003 0,000 -7,034 0,000 Wind -0,150 0,006 -23,671 0,000 Monday 0,011 0,028 0,399 0,690 Tuesday 0,015 0,032 0,465 0,642 Wednesday 0,017 0,032 0,527 0,598 Thursday 0,015 0,032 0,455 0,649 Friday 0,030 0,031 0,966 0,334 Sunday -0,112 0,031 -3,666 0,000
P-values expressed in bold are significant at 5% level
We find a significant negative relationship between the RSI and the German electricity prices. This relationship is to be expected and is in line with earlier research on RSI (Sheffrin, 2002) and Mulder & Scholtens (2013). A lowering of the RSI means a higher market power for the biggest energy producer, which translates to higher electricity prices. In this case we find a significant elasticity of -0.366. Furthermore, if we take a look at the semiannual figures in table 4.2, we see a constant significant relation that varies between -0.06 to -0.86.
The variable river temperature has a positive relationship with the German electricity prices. Although the coefficient is very small at 0.036, it is statically significant. Because this variable is not expressed in a natural logarithm, this coefficient cannot be interpreted as an elasticity. Looking at the semiannual figures, you will see that each first half of a year does not contain coefficients. This is done to prevent perfect collinearity that is caused by the fact that this variable is constantly 0 since the temperature of the river water does not exceed the threshold temperature of 23 degrees Celsius during this period. During 2015 and 2016 the effect of the river water temperature weakly increased from 0.02 to 0.03.
18 variable intensity of the sun does follow our expectations. It shows a minor negative relationship with the German electricity prices and becomes slightly stronger during our focus period. This could be caused by the installation of more photovoltaic panels during this period. Nevertheless, both variables are incredibly close to zero. Thus, one could ask himself the question if there is a real connection between these variables and the German electricity prices. The second form of renewable energy that we studied is energy generated by wind turbines. We find a strong negative connection between the German electricity prices and the average wind speed in Germany. During our focus period, we find an elasticity of -0.15. Thus, an increase of the windspeed by 1% causes a decrease of the German day-ahead electricity price of 0.15%. As table 4.1 depicts, the t-statistic is -23.67 and falls out our rejection region of -1.96 and 1.96. Hence, we can conclude that the measured coefficient is significant. This negative relationship is in line with our expectations and also partly in line with our leading paper Mulder & Scholtens (2013). Our leading paper finds a more modest effect of -0.03% of the German windspeeds on the Dutch electricity prices and an even more modest effect of below 0.01% of Dutch wind power on the Dutch electricity prices. Furthermore, when we look at the semiannual figures we cannot establish that the impact of the wind increases over time despite the more capacity that becomes available. The impact of the wind is the strongest in the second half of 2015 with an elasticity of -0.16 and the weakest in the second half of 2016 with an elasticity of -0.13. The lower coefficient in the second half 2016 might be caused by the lower windspeeds that were measured in that period. When comparing the second half of 2015 to the second half of 2016, measured average windspeeds were almost 18% lower.
Table 4.2
Regression outputs semiannual Amount of observations: 17544
Q1-2 2015 Q3-4 2015 Q1-2 2016 Q3-4 2016
Coeff. t-sta. p-val. Coeff. t-sta. p-val. Coeff. t-sta. p-val. Coeff. t-sta. p-val.
Coal 0,11 1,87 0,06 0,33 0,95 0,34 0,65 3,52 0,00 2,21 10,50 0,00 Netload 0,38 1,29 0,20 1,51 18,60 0,00 0,69 4,55 0,00 1,07 12,98 0,00 RSI -0,39 -3,64 0,00 -0,06 -2,80 0,01 -0,86 -5,71 0,00 -0,16 -3,90 0,00 RTR 0,02 5,05 0,00 0,03 3,69 0,00 Sun duration 0,002 5,25 0,00 0,002 6,19 0,00 0,001 1,58 0,11 0,002 6,03 0,00 Sun intensity -0,002 -2,44 0,01 -0,004 -7,33 0,00 -0,004 -5,45 0,00 -0,004 -8,50 0,00 Wind -0,15 -15,3 0,00 -0,16 -19,7 0,00 -0,14 -13,9 0,00 -0,13 -12,3 0,00 Monday 0,03 0,99 0,32 0,04 1,16 0,24 -0,05 -1,20 0,23 -0,03 -0,76 0,45 Tuesday 0,05 1,28 0,20 0,01 0,43 0,67 -0,01 -0,33 0,75 -0,04 -1,08 0,28 Wednesday 0,07 1,77 0,08 0,03 0,84 0,40 -0,03 -1,00 0,32 -0,07 -1,91 0,06 Thursday 0,05 1,26 0,21 -0,01 -0,33 0,74 -0,02 -0,61 0,54 -0,04 -1,23 0,22 Friday 0,04 1,10 0,27 0,01 0,30 0,76 0,00 0,12 0,90 0,01 0,32 0,75 Sunday -0,18 -4,09 0,00 -0,05 -1,22 0,22 -0,14 -3,28 0,00 -0,04 -0,94 0,35
19 We also added dummies of the days of the week to our regression. All the dummies, except for the dummy Sunday, are statistically insignificant. On Sunday, the average price becomes lower than the price on Saturday. This finding is in line with our expectations.
This study finds that the average wind speed has a significant negative effect on the German day-ahead prices. There is hardly any evidence found that energy retrieved by photovoltaic panels, both duration of the daylight as sun intensity, have an impact on electricity prices. Just like in our leading paper, the economic factors together (coal, netload, RSI) have more influence on the electricity prices then the environmental factors (RTR, daylight, sun intensity and wind).
5. Robustness check
We will perform a robustness check by replacing the measured windspeeds and sun-intensity/daylight variables with the actual output (actual electricity generated by these sources) that are published by the four German TSOs (Amprion, TenneT and Transnet, 50Hertz). The hourly data for this robustness check is derived from the transparency platform of Entsoe and reported in natural logarithms.
5.1 Wind
20
Table 5.1
Regression outputs with actual wind output Amount of observations: 17544
Coefficient Std. Error t-statistic p-values.
Coal 0,113 0,047 2,415 0,016 Netload 0,681 0,267 2,549 0,011 RSI -0,351 0,098 -3,593 0,000 RTR 0,033 0,005 6,986 0,000 Sun duration 0,000 0,000 0,088 0,930 Sun intensity -0,006 0,000 -12,475 0,000
Actual wind output -0,214 0,008 -26,464 0,000
Monday 0,021 0,026 0,824 0,410 Tuesday 0,024 0,029 0,809 0,419 Wednesday 0,029 0,030 0,986 0,324 Thursday 0,019 0,030 0,648 0,517 Friday 0,031 0,028 1,121 0,262 Sunday -0,124 0,029 -4,218 0,000
P-values expressed in bold are significant at 5% level
5.2 Solar
The second check that we are going to perform is replacing the variables daylight and sun intensity with the actual output of the solar panels. As table 5.2 depicts, the impact of the actual output of the sun is statically significant with a coefficient of -0.000011. This is close to zero just what our model with the original variables also suggests (0.002 and -0.003) and is materially not significant.
Table 5.2
Regression outputs with actual solar output Amount of observations: 17544
Coefficient Std. Error t-Statistic Prob.
Coal 0,132 0,055 2,404 0,016
Netload 0,930 0,375 2,481 0,013
RSI -0,371 0,133 -2,788 0,005
RTR 0,041 0,005 8,357 0,000
Actual sun output -0,000011 0,000 -4,728 0,000
Wind -0,141 0,007 -19,385 0,000 Monday -0,008 0,030 -0,276 0,782 Tuesday -0,010 0,037 -0,267 0,790 Wednesday -0,008 0,036 -0,227 0,820 Thursday -0,008 0,037 -0,219 0,827 Friday 0,008 0,034 0,224 0,823 Sunday -0,096 0,032 -2,981 0,003
21
6. Conclusion
In this paper, the effect of renewable energy on the German day-ahead electricity prices is discussed. Although Germany outperforms many other countries when it comes to the amount of renewable energy in its energy mix, still 69% of its energy consumption is generated by traditional energy resources. This is visible when we look at the traditional economic variables demand/supply, RSI and coal. These variables together have the most influence of the researched variables on the German electricity prices which is in line with Mulder & Scholtens (2013). Secondly, we find that the speed of the wind has a significant negative impact on the German electricity prices. When wind speeds go up by 1%, the German day-ahead electricity price will drop with -0.15%. This is partly in line with the findings of our leading paper who find a more modest effect of the windspeed on the Dutch electricity prices. When we replace our measured windspeeds with the actual output, we even find a more powerful effect of the wind speeds on the German electricity prices. Thirdly, we find a very small connection between the influence of the sun and the German electricity prices. This becomes even smaller when our variables are replaced with the actual output of solar power panels.
So, can we reject our hypotheses?
Renewable energy generation does not have a significant impact on the electricity prices in Germany
Rejected, electricity generated by wind turbines does have an influence the German electricity prices. We could not establish a materially significant relationship between solar power and the German electricity day-ahead price.
Renewable energy generation does not have the same impact on the electricity prices in Germany as it does in the Netherlands
Partly rejected, the impact of renewable energy is stronger in our paper than in our leading paper.
22
Appendix A Table 5.2
Regression outputs the Netherland versus Germany Amount of observations: 17544
The Netherlands Germany
2006 -2007 2008 - 2009 2010 -2011 2015 -2016 Coal/Gas 0,58 0,59 0,62 0,16 Netload 0,47 0,45 0,39 0,78 RSI -0,65 -0,16 -0,18 -0,37 RTR -0,01 0,04 0,01 0,04 Sun duration -0,0002 -0,00007 0,000009 0,002 Sun intensity NL 0,003 -0,03 -0,001 Sun intensity DE 0,03 0,03 -0,01 -0,003 Wind NL -0,004 -0,01 -0,001 Wind DE -0,03 -0,03 -0,02 -0,15 Monday 0,19 0,21 0,15 0,01 Tuesday 0,16 0,18 0,10 0,02 Wednesday 0,14 0,16 0,10 0,02 Thursday 0,12 0,16 0,09 0,02 Friday 0,08 0,13 0,09 0,03 Sunday -0,11 -0,12 -0,06 -0,112
23
References
Abbasi, T., & Abbasi, S.A., 2010. Biomass energy and the environmental impacts associated with its production and utilization. Renewable and Sustainable Energy Reviews, 14(3), 919-937.
Anderson, S., 2013. Comparing offshore and onshore wind. HSA 10–5—The Economics of Oil and Energy.
Bode, S., Groscurth, H.M., 2006. Zur wirkung des EEG auf ‘den strompreis’. HWWA Discussion Paper, 348.
Brooks, C., 2014. Introductory econometrics for finance. Cambridge university press, Cambridge.
Bundesministerium für Wirtschaft und Energie, 2016. The Energy of the Future. Fifth “Energy Transition” Monitoring Report, 2015 Reporting Year.
Bundesnetzagentur, 2016. Monitoring Report 2016.
Bundesnetzagentur, 2017. Retrieved from:
https://www.bundesnetzagentur.de/DE/Sachgebiete/ElektrizitaetundGas/Unternehmen_Institu
tionen/Versorgungssicherheit/Erzeugungskapazitaeten/Kraftwerksliste/kraftwerksliste-node.html; last visited on 10 May 2017.
Chen, Z., & Spooner, E., 2001. Grid power quality with variable speed wind turbines. IEEE Transactions on energy conversion, 16(2), 148-154.
Cludius, J., Hermann, H., Matthes, F.C., & Graichen, V., 2014. The merit order effect of wind and photovoltaic electricity generation in Germany 2008–2016: Estimation and distributional implications. Energy Economics, 44, 302-313.
Destatis, 2017. Retrieved from:
24 El-Khattam, W., & Salama, M. M., 2004. Distributed generation technologies, definitions and benefits. Electric power systems research, 71(2), 119-128.
Energinet, (2013). Retrieved from:
http://www.energinet.dk/EN/KLIMA-OG- MILJOE/Energi-og-klima/Naturgas-og-klimaet/Sider/Naturgassen-som-supplement-til-vedvarende-energi.aspx; last visited on the 20may 2017
Entsoe, 2015. Retrieved from: https://www.entsoe.eu/major-projects/network-code-development/electricity-balancing/Pages/default.aspx.; Last visited on 28 may 2017
EPEX Spot, 2017. Retrieved from: https://www.epexspot.com/en/company-info/basics_of_the_power_market/negative_prices.; Last visited on 12May 2017
Elliott, G., Rothenberg, T.J. and Stock, J.H. (1996), “Efficient Tests for an Autoregressive Unit Root,” Econometrica, 64, 813–836.
Euracoal, 2017. Retrieved from: https://euracoal.eu/coal/why-is-there-no-lignite-market/; last visited on 1 May 2017.
Evans, A., Strezov, V., & Evans T.J., 2009. "Assessment of sustainability indicators for renewable energy technologies." Renewable and sustainable energy reviews 13.5, 1082-1088.
Feng, J., & Shen, W.Z., 2012. Control of variable speed pitch-regulated wind turbines in strong wind conditions using a combined feedforward and feedback technique. The science of Making Torque from Wind 2012.
Fraunhofer Institute, 2017. Retrieved from: https://www.energy-charts.de/power_inst.htm; last visited on 28 May 2017.
Gil, H.A., Gomez-Quiles, C., & Riquelme, J., 2012. Large-scale wind power integration and wholesale electricity trading benefits: estimation via an ex post approach. Energy
25 Hirth, L., 2013. The market value of variable renewables: The effect of solar wind power variability on their relative price. Energy economics, 38, 218-236.
Hittinger, E., Whitacre, J.F., & Apt, J., 2010. Compensating for wind variability using co-located natural gas generation and energy storage. Energy Systems, 1(4), 417-439.
Holttinen, H., 2004, The Impact of Large Scale Wind Power Production on the Nordic Electricity System. VTT technical research centre of finland, Julkaisija, VTT Publications 554.
Kant, K., Shukla, A., Sharma, A., & Biwole, P.H., 2016. Heat transfer studies of photovoltaic panel coupled with phase change material. Solar Energy, 140, 151-161.
Ketterer, J.C., 2014. The impact of wind power generation on the electricity price in Germany. Energy Economics, 44, 270-280.
Klein, A., Held, A., Ragwitz, M., Resch, G., & Faber, T. (2008). Evaluation of different feed-in tariff design options: Best practice paper for the International Feed-feed-in Cooperation. Energy Economics Group & Fraunhofer Institute Systems and Innovation Research, Germany.
Markou, H., & Larsen, T.J., 2009. Control strategies for operation of pitch regulated turbines above cut-out wind speeds. In 2009 European Wind Energy Conference and Exhibition, Marseille.
Mauritzen, J., 2012. What happens when it's windy in Denmark? An empirical analysis of wind power on price variability in the Nordic electricity market. Unpublished working paper. Research Institute of Industrial Economics (IFN), Stockholm.
Möst, D., & Genoese, M., 2009. Market power in the German wholesale electricity market. The Journal of Energy Markets, 2(2), 47.
26 Mulder, M., & Scholtens, B., 2016. A plant-level analysis of the spill-over effects of the German Energiewende. Applied Energy, 183, 1259-1271.
Müller, U., Greis, S., & Rothstein, B., 2007. Impacts on water temperatures of selected German rivers and on electricity production of thermal power plants due to climate change. In 8. Forum dkkv/cedim: Disaster Reduction in a Changing Climate, Karlsruhe, Germany.
Newey, W.K., and West D.K., 1987, A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix, Econometrica 55, 703–708
Newey, W. K., & West, K. D. (1994). Automatic lag selection in covariance matrix estimation. The Review of Economic Studies, 61(4), 631-653.
Paul, A.C., Myers, E.C., & Palmer, K.L., 2009. A partial adjustment model of US electricity demand by region, season, and sector. RFF Discussion Paper No. 08-50.
RWE, 2017, retrieved from:
https://www.rwe.com/web/cms/mediablob/de/59784/data/2877620/136/transparenz- offensive/zentraleuropa/kapazitaet-betriebsinformationen/Power-plants-data-RWE-Q2-2017-20170401.pdf; last visited on 10 May 2017.
Sensfuß, F., Ragwitz, M., & Genoese, M. (2008). The merit-order effect: A detailed analysis of the price effect of renewable electricity generation on spot market prices in
Germany. Energy policy, 36(8), 3086-3094.
Sheffrin A., 2002. Predicting market power using the residual supply index, Presented to FERC Market Monitoring Workshop; December 3-4 2002.
Vattenfall, 2016. Retrieved from:
27 Weber, C., & Woll, O., 2007. Merit-Order-Effekte von erneuerbaren Energien-zu schön um wahr zu sein? Unpublished working paper. University of Duisburg-Essen.
Weigt, H., 2009. Germany’s wind energy: The potential for fossil capacity replacement and cost saving. Applied Energy, 86(10), 1857-1863.
Wwindea, 2016. Retrieved from: http://www.wwindea.org/wwea-half-year-report-worldwind-wind-capacity-reached-456-gw/ last visited on 25April 2017
Würzburg, K., Labandeira, X., & Linares, P., 2013. Renewable generation and electricity prices: Taking stock and new evidence for Germany and Austria. Energy Economics, 40, 159-171.
Website for data:
Entsoe actual wind output:
https://transparency.entsoe.eu/generation/r2/actualGenerationPerProductionType/show; last visited on 22 May 2017
Entsoe actual load:
https://transparency.entsoe.eu/load-domain/r2/totalLoadR2/show; last visited on 22 May 2017 Wind and solar information:
ftp://ftp-cdc.dwd.de/pub/CDC/observations_germany; last visited on 22 May 2017 https://www.esrl.noaa.gov/gmd/grad/solcalc/sunrise.html; last visited on 28 May 2017 River water temperature:
http://www.gkd.bayern.de/fluesse/wassertemperatur/karten/index.php?gknr=0&thema=gkd&l ubrik=fluesse&produkt=wassertemperatur; last visited on 22 May 2017
