Presentation at the 7th International Conference on Smart Energy Systems, September 21-22, 2021, Copenhagen
With the increasing expansion of renewable energies, volatility in the energy system is rising, increasing the demand for flexibility. Flexibility can be provided, for example, via flexible consumers, such as load management in industry or the controlled charging of electric vehicles. As investing in flexibility is an economic decision, profitable business models are needed today and in the future to tap into flexibilities. Frequently, modeled electricity prices from long-term power system models are used to evaluate business models for future years. However, these modeled electricity prices usually exhibit much lower volatility than real electricity prices. Consequently, modeled electricity prices result in lower revenue potentials than time series of historical real prices.
In a first step, this paper quantified the deviations between model results and historical electricity price time series in order to address this problem for Germany. The mean daily standard deviation of electricity prices was used as an indicator to assess revenue potentials. In a second step, identified influencing factors were assigned to superordinate categories (cf. the following figure).
From the category “model simplifications”, many of the influencing factors can be addressed for historical years by including real-world historical data (e.g. historical power plant availabilities, historical imports/exports, and historical pumped storage usage) rather than average values. Based on the historical calculation, the influence of the respective aspect was quantified, and a relevant influence of the average power plant availabilities was identified for historical years. However, the influence for Germany decreases when forecasting future years since the use of coal and nuclear power plants (power plants with high full load hours) deviates in particular when using mean availabilities. For coal, the average availabilities reduce the maximum generation in the first 2000-3000 h/a with a similar development in the further hours.
For nuclear power plants, baseload operation reduces output throughout the year. However, the observed effects become less significant due to the planned coal and nuclear phase-out in the next few years. Due to low full-load hours, gas-fired power plants are only slightly affected by mean availability. The effect of mean NTCs, on the other hand, can be assessed as relevant both today and in the future. However, analyses have shown that NTCs are strongly situation-dependent on many aspects, and general statements about the development of future NTCs are impossible.
The simplified representation of technical restrictions results from the linearization of the model, which is indispensable for reasons of computability. Therefore, a quantification of this influencing factor was also not directly possible. Meanwhile, a relevant influence can be identified for the missing representation of the balancing power market and thus the lower availability of plants for the electricity market. An adjustment of the model regarding the balancing power market is possible via the availability factor. However, an estimation of the attractiveness of the balancing power market for future years is essential, but complicated due to regulatory adjustments and strongly changing prices.
The missing, unconsidered aspects cannot be addressed since they cannot or should not be included in the model. For example, gaming should not be mapped in an overall economic optimization by default. However, these aspects can have a significant influence on electricity prices. For the category “model errors”, consideration is possible through continuous further development. However, this does not completely eliminate the influencing factor.
Overall, most of the identified influencing factors are challenging to address for future years. Therefore, an alternative set of solutions is sought: Identifying indirect influencing factors, which are also available for future years.
Here, the residual load was identified as a possible indirect influencing factor of the deviations between modeled and real electricity prices. For the year 2018, a correlation between the electricity price deviations between modeled and historical data and the residual load can be identified, although the large deviations in the range of negative electricity prices with high RE feed-in are not representative of it (see the following figure). Overall, prices tend to be underestimated at a higher residual load, and prices tend to be overestimated at a low residual load.
The linear regression developed as a downstream correction function can significantly improve the modeled electricity prices for 2018 to evaluate revenue potentials. However, the real revenues of flexibilities still cannot be modeled. For future years (2030 or 2050), however, using this downstream correction function is not possible. Reasons include the expected fundamentally different characteristics of electricity prices compared to 2018, and possible changes in market design. In future years, many points in time with high electricity prices can be expected due to the higher marginal costs of conventional power plants resulting from higher fuel and CO2 prices. In addition, due to substantial RE expansion, generation surpluses from renewables and associated electricity prices of 0 €/MWh occur. This also ensures a downward shift of the residual load. This would reduce the overall average electricity price if the current correction function is used. Adjustments in regulation are expected based on signals given by the new German government, such as adjustments to the EEG or the 4-hour rule, which influence the electricity price characteristics. Overall, the above points imply that using the correction function based on historical electricity prices for future years from 2030 onwards is not likely to produce realistic results.
Overall, this work has shown a wide variety of factors influencing the electricity price deviations between modeled and historical data, in turn affecting the flexibilities’ revenue potential. Influencing factors resulting from model simplifications, such as average power plant availabilities, were analyzed in detail to determine their influence on the future power system, and a possible adjustment of the influencing factor was discussed. Even though many aspects cannot be addressed for future years, the systematic exploration of the influencing factors is helpful for later analyses. A downstream correction function based on the residual load was developed in parallel. This provides good short-term results for the next few years, even if it cannot be used for long-term analyses due to the strongly changing energy system.
The work is part of the Kopernikus SynErgie project.