According to Weron, models built from fundamentals upwards form a prominent percentage of day-ahead price analytics. . Fundamental models are in line with the standard economic theory that prices in competitive markets result from the equilibrium of demand and supply. The day-ahead market price then equals the marginal costs of the last operating power plant that is needed to satisfy demand.
Scarcities of electricity in day-ahead and intraday markets are reflected in the publicly observable prices on power exchanges. The prices of unobserved bilateral trades will not deviate systematically from exchange prices because traders have the option to trade either on the exchange or bilaterally. They will not trade in one market if trading in the other yields a higher profit. To support the decision making of producers, consumers and traders, various modelling and forecasting approaches for electricity spot prices have been developed. These include fundametals, along with those that incorporate game theory, technical analysis, econometric-stochastic and artificial intelligence.
For the German intraday market, no results from a fundamentals approach have been published so far, and only a few explicit intraday price models exist compared to the day-ahead pricing literature. Hagemann (2015) uses a multiple linear regression model to investigate the influence of intraday price determinants in Germany and explains that intraday supply side shocks may have different price effects. Selasinsky (2014) visualises the causal relationship between German intraday prices and unforeseen intraday changes of the residual load in a contour diagram and Furió analyses the price convergence between the Spanish day-ahead and intraday market from 2000 to 2005 and find significant price differences between both markets.
Further intraday literature covers trading strategies of agents who balance wind power forecast errors and research on market design and liquidity. The major strength of a fundamentals modelling approachs are the ability to account for non-linearities in the supply stack and the ability to combine time-varying information such as fuel and CO2 prices or renewable feed-in consistently. Furthermore fundamental approaches are flexible in the sense that existing models can be integrated for certain fundamental factors (e g. wind forecasting tools). By applying Monte Carlo simulations, insights into future spot price dynamics under various future demand and supply scenarios are possible.
Fundamental modelling including inter-temporal restrictions of plant operation and using optimisation techniques is done by Ellersdorfer et al. (2008) for the German market and by Borenstein et al. (2002) for the Californian market. More recently, Graf and Wozabal (2013) published a work with a similar approach for the German day-ahead market. Publications basing their fundamental modelling on the representation of power plants’ variable costs are used to investigate strength, competitiveness or strategic behaviour in power markets. Usually their focus is not on electricity price forecasting but, instead, on identifying differences between prices predicted by the marginal cost theory and actual prices. Müsgens (2005) uses an advanced fundamental model for more than one region but only with marginal costs based on a monthly time resolution. Schwarz and Lang (2006) apply a similar modelling approach, focusing on Germany and assessing the hourly marginal costs to investigate price mark-ups as well as fly-ups.
The German power market – A brief overview
Because electricity is not economically storable in relevant quantities, a balance of supply and demand is required at all times to maintain system stability. In Germany, this continuous equilibrium is ensured through a sequence of interdependent wholesale markets. In the day-ahead market, participants may trade electricity with physical delivery on the next day either anonymously on a trading platform of the European Power Exchange (EPEX Spot) or bilaterally over the counter (OTC). In terms of trading volumes, the day-ahead auction of the EPEX Spot is the most important spot market for Germany with a total trading volume of approximately 245 TWh for years 2012 and 2013. Here, supply and demand bids are matched in a uniform pricing auction to obtain a market clearing price and quantity for each of the 24 hours of the next day. The day-ahead market is of great importance for the integration of variable renewable energy sources (RES) because market participants forecast the expected production profile (e. g., for wind and solar power plants) for the next day and sell those expected quantities in the day-ahead market.
After the gate closure of the day-ahead market at 12 pm on the day before delivery, trading continues from 3 pm on the day before delivery until 45 min before physical delivery on the electronic intraday platform of the EPEX Spot. Here, market participants are encouraged to self-balance unforeseen deviations from their day-ahead schedules. The German intraday market is a continuous and order driven market. Incoming limit buy and sell orders are stored in the limit order book and executed as soon as the buy price exceeds the sell price or vice versa. In contrast to limit orders, market orders are executed immediately at the best available market price. The German intraday trading volumes have been constantly increasing from 1.4 TWh in 2007 to 16 TWh in 2012 and to 20 TWh in 2013. Trading volumes are much higher in the day-ahead market than in the intraday market. Hence the intraday market is of minor importance for trading and hedging but of higher importance for system security.
After the gate closure of the intraday market, the four TSOs in Germany – 50 Hertz Transmission GmbH, Amprion GmbH, TenneT TSO GmbH and TransnetBW GmbH – use the previously contracted control energy to level out demand and supply in real time and maintain the grid frequency at 50 hertz. In Germany, control energy is classified into primary, secondary and tertiary reserves. The three types of control energies have different activation times and durations of operation. Individual market participants are charged ex-post for imbalances they cause. Therefore, they will generally try to avoid the use of control energy for two reasons. The first reason is that, in Germany, using balancing services is always more expensive than self-balancing on the intraday market. The second reason is that the TSOs may impose sanctions on market participants that cause many imbalances by abrogating their balancing contract.
A fundamental modeling approach for electricity spot prices
In competitive markets, prices are expected to correspond to the intercept of demand and supply. The supply curve in electricity markets is represented by all available power plant capacities sorted in ascending order according to their short run marginal costs. In a perfectly competitive market, the marginal costs of the last running plant for a certain delivery period set the electricity price. In the absence of market power, this bidding strategy ensures maximum profits of suppliers in one-shot auctions with marginal pricing.
Either intraday prices can be modelled directly as the equilibrium price at the intercept of intraday supply and demand or the deviation of intraday prices from day-ahead prices can be modelled as done by Hagemann (2015). The deviation between the intraday and the day-ahead residual load (forecast error) determines whether up-ramping or down-ramping capacity in the intraday market is needed. In addition to the marginal costs of operating plants, other factors may influence the observed electricity prices. Therefore, multiple linear regression models are typically used to test the explanatory power of other spot price determinants beyond those included in the fundamental model. Apart from the fundamental price, regression models capture the influence of (i) (avoided) start-up costs, (ii) market state variables and (iii) trading behavior.
Whilst and assessment of the fundamentals of supply and demand can by highly illuminating, it is important to factor in the human element; traders tend to use the present or past price information to forecast future prices. In the day-ahead market, the so called similar day-approach states that the future prices can be calculated using previous prices which are corrected by changes of fundamental influences (Weron, 2006). For models with hourly resolution, a similar hour-approach is suitable. Weron and Misiorek, 2008 and Kristiansen, 2012 find significant influences of the day-ahead price of the same delivery hour of the previous two days and the previous week on the present price. In electricity systems with a high penetration of wind and solar power feed-in, changing weather conditions have a significant influence on the wind and solar power production and, thus, on the day-ahead prices (Sensfuß et al., 2008). Quickly changing day-ahead prices may reduce the explanatory power of past days. Therefore, only the auto-regressive term for the same hour of the previous day tend to be included in day-ahead regression models. Intraday traders are likely to include the price information of the previous hour in their trading decisions because intraday trading happens continuously. Therefore, the price of the preceding hour is included in intraday regression models.
Are fundamentals enough?
This draws up the logical question when considering modelling day-ahead and intra-day markets; are fundamentals enough to model and trade a market? Pape, Hagemann and Weber (2016) found that a simple fundamental model with carefully selected input data and appropriate calibration is found to explain approximately 75% of the observed intraday price variance. They also found that fundamental models yield significantly better price estimates when market peculiarities such as must-run obligations are considered, e.g., due to CHP constraints or intraday market peculiarities such as shorter lead times of power plants. Furthermore, their results indicate that it is beneficial to use the day-ahead price information to improve the intraday forecasts.
Remaining differences between the prices predicted by the fundamentals model and observed prices may be explained to some extent by (avoided) startup-costs, market states and trading behaviour. Startup-costs of conventional power plants cause day-ahead and intraday prices to rise above the prices predicted by the fundamental model. On the contrary, must-run obligations and power plant inflexibilities set an incentive for power plant owners to keep power plants in operation in the intraday market, which may lead to sales below the short run marginal costs, thus decreasing intraday prices. In the day-ahead market, avoided start-up costs are already appropriately considered in the fundamental model via must-run restrictions. Furthermore, market states have significant price impacts. In times of excess supply, power plant owners are willing to bid at prices below their marginal costs to be appointed for delivery. On the contrary, capacity scarcity leads to significant price increases but only in intraday prices. Moreover, the results indicate that traders use past price information to predict prices in the day-ahead and intraday markets and to trade accordingly.If you’ve found this blog helpful and would like other topics covered, please feel free to drop me an email with suggestions and subscribe to get the latest posts straight to your inbox