Backtest

How does the model compare to reality when we look at historic power prices

Backtesting is a process of running our model in the past to evaluate its performance against what we see historically. This allows us to validate that the model is capturing the fundamental behaviour of power prices, including features such as example top-bottom (TB) price spreads.

Building a Backtest

We use real historical data wherever possible in our backtest. This allows us to calibrate elements of the grid that are more uncertain (for example plant bidding behaviour) as accurately as possible. Historical inputs used in the backtest include:

• Daily commodities prices: gas, coal, carbon credits, FX rates
• Renewable (wind + solar) generation load factors - this removes weather as a source of uncertainty in the backtest
• Plant availability
• Interconnector availability
• Metered demand data from the TSOs (see Demand)

Key metrics

We look at the following headline metrics over a two year period from the start of 2023 to the end of 2024. This is also the range used to produce the backtest plots shown below. Note TB2 is the two-hour top-bottom price spread.

We also pay close attention to the diagnostics:

1. Raw Price Time Series
2. Monthly Price Averages
3. Price Distribution
4. Price Shape

In addition to these price related phenomena we also validate our backtests with

1. Time of Day Generation profiles from different technology types
2. Interconnector flow behaviour - see Interconnection

Raw price time series

Typical price time series vary considerably between different European nations. Note that Spanish and Portuguese wholesale prices tend to converge the majority of the time due to the sizeable interconnection capacity between them.

Monthly price averages

Across nations prices have come down somewhat since early 2023 as gas prices subsided from their elevated levels during the European energy crisis.

Price distribution

Some markets such as GB and especially Germany experience significant scarcity events filtering through to wholesale electricity prices. Spain and Portugal on the other hand, thanks in part to sizeable hydro fleets rarely experience prices over 200 Euros per MWh.

Price shape

Here we plot the intraday shape of prices by subtracting the overall average price from the average price by time of day (TOD). The resultant duck-curve shape consists of three key time periods:

1. Morning peak: As people wake up demand for electricity grows causing more expensive thermal generation to ramp up and prices to rise.
2. Solar Low: Significant solar build out over the last decade has lead to an abundance of cheap solar electricity during the peak solar hours around the middle of the day
3. Evening Peak: Demand typically increases in the evening as people return from work and switch on their household appliances. This, combined with a drop off in solar production, necessitates a significant ramp up in thermal generation leading to typically the highest prices of the day.