Econometric analysis of 15-minute intraday electricity prices
Introduction
Trading in the intraday electricity markets increased rapidly since the opening of the market. This may be driven by the need of photovoltaic and wind power operators to balance their production forecast errors, i.e. deviations between forecasted and actual production. Evidence for this is a jump in the volume of intraday trading as the direct marketing of renewable energy was introduced. Furthermore, there may be a generally increased interest in intraday trading activities due to proprietary trading. Our main goal is to identify explanatory variables, specific to the electricity intraday market, that influence the bidding behavior in the 15-minute intraday market at the European Power Exchange (EPEX).
Along the basic timeline of electricity trading activities, see Fig. 1, the intraday activities relate mostly to further adjustments of positions after the closure of the day-ahead market.
While day-ahead trading offers the possibility to correct the long-term production schedule (build on the forward markets) in terms of hourly production schedule of power plants (Delta Hedging) and to adjust for the residual load profiles on an hourly basis, the increasing share of renewable energy sources (wind, solar) in electricity markets requires a finer adjustment.
According to the Equalization Mechanism Ordinance (ger.: Verordnung zur Weiterentwicklung des bundesweiten Ausgleichsmechanismus, abbr.: AuglMechV) all electricity generated by renewable sources has to be traded day-ahead. This is usually done by the transmission system operator (TSO) with the plant operator receiving a legally guaranteed feed-in-tariff. From 2012 on, the inclusion of a market premium led direct marketers within the feed-in premium support scheme to enter the market as well. Trading of electricity from a renewable energy source is based on forecasts which may have a horizon of up to 36 h (taking some data-handling into account). To correct errors in forecasts the AusglMechV requires the marketers of renewable energy to use the intraday market to balance differences in actual and updated forecasts. Intraday trading starts at 3 pm and takes place continuously until up to 30 min before the start of the traded quarter-hour. As forecasts change regularly, marketers may sell and buy the same contract at different times during the trading period.
After the closure of the intraday market balancing energy has to be used to close differences between available and forecasted electricity. As a smaller number of power plants are used for balancing energy the merit-order curve is steeper than that in the intraday market. Thus, on average larger prices are paid and marketers aim at minimizing this difference, see Graeber (2014). In addition, TSOs may impose sanctions on marketers who frequently require balancing energy.
Balancing energy is supplied by generators with the necessary flexibility to balance the market. In case generation is below demand, positive balancing energy is used, otherwise negative balancing energy is used. Graeber and Kleine, 2013, Just and Weber, 2012 contain a detailed description of the integration of renewable energy in electricity markets and the regulatory requirements and we refer the reader to these sources for further information.
The day-ahead market (spot market) and the balancing markets have been investigated extensively. For example, Paraschiv et al. (2014) show that the day-ahead price formation process at EPEX depends on the interaction/substitution effect between the traditional production capacity (coal, gas, oil) with the fluctuant renewable energies (wind and photovoltaic (PV)). Further empirical studies on intraday/balancing markets include Karakatsani and Bunn, 2008, Klaeboe et al., 2013. Also, Mller et al. (2011) linking day-ahead and balancing markets.
An investigation in the merit-order effect is given by Cludius et al. (2014), who find that electricity generation by wind and PV has reduced spot market prices considerably by 6 €/MWh in 2010 rising to 10 €/MWh in 2012. They also show that merit order effects are projected to reach14–16 €/MWh in 2016.
Recent studies of the intraday high-frequency electricity prices at EPEX are Hagemann, 2013, Hagemann and Weber, 2013 who look at liquidity effects and forecast determinants on a hourly basis. Also, Garnier and Madlener (2014) considers trading strategies to minimize costs from imbalances for both PV and wind, but generates price changes in terms of a reduced-form model (using a stochastic process). The focus lies in developing a trading strategy for a given setting, and not on explaining the relevant price process. Several studies have discussed the effects of prognosis errors for wind generation (see Ketterer, 2014, Nicolosi, 2010) . As Fig. 2 suggests, a PV production introduces quarter-hour ramps quite naturally. In addition, changes in forecasts of renewable energy production require a timely correction of day-ahead positions. However, photovoltaic has not been investigated so far.
Hagemann, 2013, Hagemann and Weber, 2013 used the ex-post published wind infeed data to explain ex-ante their impact on the day-ahead market. These are publicly available data from the Transparency Platform EPEX. However, the actual infeed is only known ex-post and therefore it cannot be used directly to explain the price formation on the intraday market. In fact, the intraday market participants have access to updated forecasts of wind. In our study, we will extend the existing literature by taking into account the intraday updated forecasts for wind and PV, which have been supplied by EWE Trading GmbH.
Each day, hourly day-ahead electricity prices are revealed around 2 pm at EPEX (see Paraschiv et al., 2015 ). At the same time, market participants have access to forecasts for wind and PV published by each Transmission System Operator (TSO) in 15-minute intervals for the next day. However, wind and PV forecasts are updated frequently during the trading period. Thus, at the time when market participants place their bids for a particular intraday delivery period (hour, quarter of hour), updated information about the forecasting errors of renewables becomes available. In consequence, also deviations between the intraday prices and the day-ahead price for a specific hour are expected to occur. Our main research question is, thus, to which extent do market participants change their bidding behavior when new information on wind and PV forecasts becomes available. We will employ a unique data set of the latest forecasts of wind and PV available at the time of the bid.
Our analysis is twofold: Firstly, we analyze the difference between the last price bid for a certain quarter of hour and the day-ahead price for that hour. We distinguish between summer/winter, peak/off-peak hours. We test for asymmetric behavior of prices to forecasting errors of renewable energy dependent on the demand quote regime and investigate further the typical zigzag pattern of intraday prices. Thus, we identify a seasonality shape that provides traders important information about the time of the day when they can bid, dependent on their demand/supply profiles. Furthermore, the effect of volume of trades/market liquidity is investigated. Secondly, we are interested in the bidding behavior of market participants in the continuous intraday electricity market. We thus analyze the continuous trades and disentangle the effect of explanatory variables dependent on the time of the day. The econometric analysis is replicated for several traded hourly quarters, at different time of the day. In particular, we are interested to see how delta bid prices change when new information becomes available in the intraday renewable forecasts for wind and PV. We look at the trade-off between autoregressive terms and the market-related exogenous variables impacting the intraday price formation process.
Our contribution to the existing literature is twofold: we use ex-ante forecasts of wind and photovoltaic and employ high-frequency intraday prices for specific quarter hours. Overall, our paper aims at understanding historically the continuous bidding in the intraday market, and proposes a one-period reduced-form forecasting model based on exogenous variables which are observed by market participants at the time of the bid. We show that estimation results are stable over time, but it is highly relevant to re-estimate the econometric model separately for summer/winter, peak/off-peak periods. We used as benchmark an autoregressive model and show that the price formation process is rather driven by market-specific explanatory variables, especially for mid-day delivery periods. The list of explanatory variables includes expected demand, an aggregate index for the power plant availability including traditional capacity planned day-ahead, the volume of trades, control area balances, and intraday updated forecasting errors of wind and photovoltaic. This is the first study which includes ex-ante updates in forecasting errors of renewable energies. This study proves that intraday updated forecasts of wind and PV impact the bidding behavior: we show that market participants access updated forecasts in renewables to have more private information and thus to bid more accurately.
The rest of the paper is organized as follows: In Section 2, we explain the modeling assumptions. 3 Input variables: definition and data sources, 4 Asymmetric econometric model for intraday prices show the data used as input and a theoretical representation of our concept. Section 5 proceeds with the formulation of our reduced-form econometric analysis. Results and their interpretation are given in 6 Estimation results and interpretation, 7 Conclusion concludes.
Section snippets
Theoretical considerations
Our main assumption is that the electricity intraday price formation process depends on how much traditional capacity has been allocated in the day-ahead market and in which proportion it covers the forecasted demand. Let us consider two possible market regimes:
- 1.
The traditional capacity planned for the day-ahead satisfies the expected demand for a certain hour;
- 2.
There is a certain demand quote uncovered by the planned capacity.
Thus, in scenario 2, negative forecasting errors of wind and PV will
Input variables: definition and data sources
As motivated in Section 2, for the analysis we employed historical day-ahead and intraday electricity prices for 15-minute products in the continuous trading system between 01/01/2014 and 30 /06/2014. As explanatory variables selected in this study, we refer to demand forecast, power plant availability, intraday updated forecasts for wind and photovoltaic, volume of trades in the continuous trading, and the control area balance. The latter represents the corresponding use of balancing power in
Threshold model specification
The technical specification of our model follows Paraschiv (2013) and reads: where ωi is the threshold variable used to split the sample into two regimes. The random variable εi is a regression error.
Our observed sample is , where yi represent the dependent variable and xi is an m-vector of independent variables. The threshold variableωi may be an element of xi and is assumed to have a continuous distribution. To write the model in a single
Analysis of intraday prices
We examine whether intraday prices in the continuous bidding system are caused by market-specific variables. As already mentioned earlier in this study, marketers of renewable energy use the intraday market to balance out differences between actual/updated forecasts of wind and PV. Indeed, discussions with energy traders revealed that at the time of the bid market participants have private access to the freshest weather forecasts for a certain quarter of an hour (delivery period) and use this
Analysis of the deviations of last prices from the day-ahead price
Eq. (7) has been estimated for the historical differences between the last prices and the day-ahead prices separately for winter and summer and we further distinguished between peak (8 am and 8 pm) and off-peak hours. This approach is justified by the different price levels in summer compared to the winter time and by the different demand profiles during peak and off-peak hours (see Paraschiv et al., 2015 for an extensive discussion on the seasonality of electricity prices).
As a preliminary
Conclusion
In this study, we investigate the bidding behavior in the intraday electricity market, in the context of a reduced-form econometric analysis. In particular, we shed light on the impact of updated forecasting errors of wind and photovoltaic (PV) on the 15-minute electricity price changes in the continuous bidding. We employ a unique data set of the latest forecasts of wind and PV available to traders prior to the placements of their price bids intraday. To our knowledge, this is the first study
Outlook
Our analysis sheds light on the bidding behavior historically speaking and offers a solid basis for one-period forecast of last intraday prices and continuous bids. Since all variables used as input can be computed based on the information available at the time of the bid (demand quote, updated forecasts in renewables), the econometric model can be used for forecasting the (next) continuous bid. We prove the superiority of this econometric model specification over the classical AR model
Acknowledgments
The authors thank Hendrik Brockmeyer for his valuable input in the data collection step. We further thank Claus Liebenberger and Karl Frauendorfer from the Institute for Operations Research and Computational Finance (IOR /CF), University of St.Gallen for further support in the data collection step and for the interesting discussions in the key steps of the paper. We thank in addition Reik Börger for very useful discussions about the intraday markets and Jonas Adam for technical support with
References (21)
- et al.
The merit order effect of wind and photovoltaic electricity generation in Germany 2008–2016: estimation and distributional implications
Energy Econ.
(2014) - et al.
On the market impact of wind energy forecasts
Energy Econ.
(2010) - et al.
Intra-day and regime-switching dynamics in electricity price formation
Energy Econ.
(2008) The impact of wind power generation on the electricity price in Germany
Energy Econ.
(2014)Wind power generation and power system flexibility — an empirical analysis of extreme events in Germany under the new negative price regime
Energy Policy
(2010)- et al.
The impact of renewable energies on EEX day-ahead electricity prices
Energy Policy
(2014) - et al.
A spot-forward model for electricity prices with regime shifts
Energy Econ.
(2015) - et al.
Balancing Forecast Errors in Continuous-Trade Intraday Markets. FCN WP 2/2014
(2014) Volatility and volume effects in European electricity spot markets
Econ. Notes
(2010)Handel Mit Strom Aus Erneuerbaren Energien
(2014)
Cited by (0)
Part of the work was done while the author was visiting Center of Advanced Study, Norwegian Academy of Sciences and Letters, Oslo, as a member of the group: Stochastics for Environmental and Financial Economics.