Forecasting for Pumped Storage in Germany

While the potential role for pumped storage is great, no study had yet been conducted that so thoroughly demonstrated its role in Germany’s overall energy mix. To correct this, researchers developed a new methodology for correcting past inaccuracies.

By Klaus Kruger, Andreas Maaz, Albert Moser and Niklas Rotering

Balancing generation required to meet load with providing reliable available capacity is a major challenge in electrical systems that feature intermittent renewable energy sources- includinging conventional hydropower – as a significant source of power. The systematic use of cost-effective and proven pumped-storage technology can provide significant contributions to meet this challenge.

Using a scenario developed by German power sector think tank Agora Energiewende, this article presents a model to simulate the influence of errors in forecasting renewable energy sources and the demand for a highly-flexible control reserve system. The aim is to identify costs resulting from these forecast errors and to evaluate how pumped storage can reduce these costs in an energy system scenario with a high renewables share.

In contrast to other studies, however, the model used in this research did not rely on the assumption of perfect foresight. Rather, a step-wise approach to simulate the European power market’s uncertainties resulting from renewable forecasting errors was integrated, while different reserve qualities were modeled in detail so that the provision of highly-flexible control reserves could be assessed adequately.


The transition toward a renewable energy economy in Germany poses many technical and economic challenges. Major technical challenges include the balancing of generation and demand at all times, the provision of reliable available capacities and renewable grid integration.

Previously, energy was stored in fossil or nuclear form, and that power generation was flexibly adapted to fit demand. In the case of wind and solar generation, no comparable storage technology exists. Hence the power generation from these sources is volatile, leading to an increasing demand for flexibility from other sources.

The general demand for flexibility can be divided into four categories depending on the required energy, with each category usually addressed in a separate market:

  1. Provision of reliable available capacity (capacity markets);
  2. Balancing residual load (day-ahead markets);
  3. Balancing forecast errors (intraday markets); and
  4. Compensation of outages and short-term fluctuations (frequency control reserves).
Pictured is the control room in the Herdecke pumped-storage plant in North Rhine- Westphalia, which replaced the nearby Koepchenwerk pumped-storage plant after it suffered a mechanical breakdown in the 1980s. (Photo courtesy Voith Hydro)
Pictured is the control room in the Herdecke pumped-storage plant in North Rhine- Westphalia, which replaced the nearby Koepchenwerk pumped-storage plant after it suffered a mechanical breakdown in the 1980s. (Photo courtesy Voith Hydro)

Problems with forecasting

A large degree of uncertainty exists regarding the extent of any required storage expansion in Germany, with some studies questioning whether an expansion is even necessary or economical in the near and mid-term future.

Regardless of the conclusion, however, many projections share a shortcoming in that they do not model and optimize the balancing of forecast errors, as well as the provision of frequency control reserves in detail.

Instead, these studies use an aggregate model that lumps all control reserve requirements and forecast error balancing. These reserve requirements are relaxed, perfect foresight is assumed and the resolution in time is modeled to be one hour. This is mainly because a more detailed model would be more complex and require forecast error data, which is not easily available. Additionally, only one study modelled a possible provision of reliable available capacity by storages, meaning the economic value of storage expansion was underestimated.

To address these shortcomings, the authors developed a new method that allowed for more adequate power plant modeling while considering forecast errors and detailed reserve requirements.

These tools were then used to study the economic value of a pumped-storage expansion in Germany, while simulating the coupled electricity markets in all relevant countries, based on expansion scenarios of 3 GW and 6 GW. The flexibility options of the dispatch simulation included fossil fuel generation, electricity storage and demand-side management.

Running the simulation

The investigation consisted of three different simulation runs. The first simulated the European power markets without additional pumped storage, while two others were performed including additional pumped-storage capacities of 3 GW and 6 GW.

In each of the two pumped-storage scenarios, the ratio between generation capacity and storage capacity was set to nine hours of generation at maximum power. This enabled the pumped storage to cover the peak load for several hours during long periods with low generation from other renewables. In this way, there is a guarantee that generation capacity from storage is reliably available.

All simulations consisted of two parts. In the first part, a deterministic simulation of the full area was performed for one year in an hourly time pattern. The main results of the first part are the imports and exports between the focus area and the bordering markets. This reflects the market situation in Europe, where a market-coupling of most European market areas is state of the art regarding the day-ahead market. For the second part of the simulations these import and export time series were then used to simulate the focus area, taking into account uncertainties and forecast deviations. This reflects the fact, that the cross-border trading at the European intraday markets is limited only to a few countries. In the simulation it was assumed, that all market areas bordering Germany enable cross-border intraday trading.

This graph shows the magnitude of absolute renewable energy source forecast error assuming a 60% renewables share scenario in 2033.
This graph shows the magnitude of absolute renewable energy source forecast error assuming a 60% renewables share scenario in 2033.

First results of interim calculations

An analysis of the computed forecasts (see Figure 1) shows that the forecast error for renewable energy generation with a forecast horizon of three hours before delivery in Germany has rather high peaks. When looking at the 5 % highest values, which occur in more than 438 hours per year, deviations of more than 9 GW occur in Germany alone. When assuming perfect connection to the neighboring markets, the deviations of the total focus area rise to more than 12 GW in the most extreme 438 hours per year.

The forecast error resulting from the prognosis one hour prior to delivery is assumed to be compensated for by means of frequency control reserves. Deviations with a longer prognosis horizon are covered by short-term energy trading, or within the responsible portfolio (e.g. intraday trading).

Such forecast errors cannot be compensated for by a cold reserve of thermal power plants, instead requiring short-term activation of energy within a few hours from pumped storage, hot thermal power plant reserves, gas turbines or combined cycle plants.

2033 base scenario without pumped-storage expansion

The base scenario is simulated deterministically with aggregated reserve requirements. Afterward, a detailed step-wise simulation is performed for the focus area. In this simulation, the reserve provision of the different reserve requirements for primary, secondary and tertiary control reserves are included. The results of the simulation are used as a basis for evaluating pumped-storage expansion later on.

The base scenario is not considering additional energy storage installations in 2033. Using Germany as an example (see Figure 2), a significant share of energy is generated by renewables – particularly solar and wind. A significant portion of energy in the scenario is also still coming from nuclear and coal power. The production from pumped-storage plants in Germany, with about 5 TWh annually, is at a moderate level compared to the full theoretical potential.

Primarily due to nuclear power and its low variable costs, France has a positive energy trade balance, allowing it to export electricity to meet demand in other countries. Additionally, the figure illustrates a predominant use of natural gas-fired plants in Italy, and hydropower as a dominant source in Austria and Switzerland.

Another interesting result can be obtained by comparing the variable system costs of both simulations. The cost difference can be interpreted as the additional costs for the electricity system resulting from forecast deviations and from the provision of highly flexible control reserves. These additional costs sum up to €291 million (US$321 million) annually, reflecting about 3% of the total variable costs in the focus area.

This graph compares the scenario’s projections for annual generation with data compiled by the European Network of Transmission System Operators for Electricity (ENTSOE) in 2014.
This graph compares the scenario’s projections for annual generation with data compiled by the European Network of Transmission System Operators for Electricity (ENTSOE) in 2014.

Consequences of pumped-storage expansion in Germany

Two variants considering pumped-storage expansions were investigated in comparison to the base scenario.

The first variant considered the construction of a further 6 GW of pumped storage, with 54 GWh of storage capacity, and the second included a more moderate expansion of 3 GW, with 27 GWh of storage capacity.

As might be expected by the additional construction of pumped storage, the annual generation of electricity using hydro turbines and the consumption of electricity by pumps increases significantly in both variants. The additional annual production by hydraulic turbines in Germany increases by 4.6 TWh in Variant 1 and by 2.7 TWh in Variant 2. What can also be recognized is reduced generation from natural gas-fired plants in Germany (-2.1 TWh and -1.5 TWh in Variant 1 and 2, respectively), which is compensated for by the cheaper generation options of lignite-fired plants in Germany and nuclear plants in France.

In situations with high residual loads, the additional highly-flexible hydraulic turbines take over the role of gas-fired plants. The additional available pump capacity is used to avoid renewable source curtailment in situations with negative residual loads and also to increase the number of full-load hours for European baseload thermal plants.

It is quite clear that an increased storage capacity enhances the overall generation system flexibility and reduces renewable source curtailments driven by negative market prices. For instance, in Variant 1, the additional constructed pumped-storage fleet can additionally integrate 0.3 TWh of renewable generation, which would be curtailed in the base scenario without pumped-storage expansion.

The turbine and pump utilization are shown in Figure 3 as a horizontal continuous line over a one-year period. In Variant 1, the full load hours in turbine mode are 965 hours annually, whereas in Variant 2 1,066 hours annually can be achieved. As a rule of thumb, 1,000 hours at full load per year in turbine mode is usually assumed as a minimum utilization constraint for an economically viable pumped-storage project. This is a typical calculation value that does not correspond to the actual hours of operation in turbine mode. The turbine and pump operating hours are significantly higher because each pumped-storage plant has to deliver varying output within its control load band. Additional utilization hours come on top by providing additional ancillary services to grid operators, which were not considered in the dispatch simulation.

Without the use of pumped storage, excess production from renewables regularly occurs, which cannot be compensated for even if fossil fueled electricity production is almost completely shut off. In this situation, the pumped-storage fleet receives the excess production from renewables and returns it to the grid several hours later. This avoids curtailing wind and solar installations, creating a win-win situation for the pumped-storage fleet and renewables.

Another win-win situation is evident between pumped-storage and thermal plants because the consistent use of pumped storage as short-term storage (daily cycling) levels off the capacity curve of fossil fuel plants, as well as reducing their peak load. This is advantageous for thermal plants in that it reduces the number of startup and shutdown processes, thus significantly reducing wear.

The load utilization curve for the two pumped storage capacity additions is shown in turbine mode on the left and in pump mode on the right.
The load utilization curve for the two pumped storage capacity additions is shown in turbine mode on the left and in pump mode on the right.


The pumped storage extension leads to a higher energy supply by coal-fired plants. Because of the higher specific CO2 emissions of coal in comparison to natural gas this is accompanied by higher CO2 emissions. This increase is more than compensated for by the additional integration of renewable generation using the additional pumped-storage capacity. Therefore, the overall CO2 expenses can be reduced.

In addition to savings in production costs, a greater integration of renewable generation means that the national targets for the share increase of renewables can be achieved with lower installed renewable capacity. Renewable generation that otherwise would be curtailed can be successfully integrated. This leads to a lower necessary installed capacity from renewables in order to fulfill the nationally defined objectives.

A comparison of the investment and operational costs of the storage to the various benefits and cost savings showed that both expansion variants are macro-economically beneficial to the electricity system. While for the 3 GW expansion variant a clear resulting benefit of €92.4 million (US$102 million) annually is calculated, the effect of a 6 GW expansion is still positive at up to €20.4 million (US$22.5 million) annually. This is to be expected because the possible benefits show a saturation in the case of higher pumped-storage expansion capacity. Therefore, an expansion of up to 6 GW of pumped storage in Germany makes economic sense in the investigated scenario.

This study will now be used as a scientific basis for discussions and presentations with non-governmental organizations, federal and state energy institutions, and others.

Editor’s Note: This article was adapted from a HydroVision International 2016 paper, “Renewable Generation Forecast Error in Europe – Is Additional Pumped Storage a Solution?” To download the entire paper, visit

Klaus Kruger is research and development head for basic development at Voith Hydro Holding. Andreas Maaz is a researcher at Germany’s Institute for Power Systems and Power Economics (IAEW). Albert Moser is the head of IAEW. Niklas Rotering is the owner of modEnergco and Rotering Energy.

Previous articleWorld News
Next articleApplying Vacuum Generator Circuit Breakers at Hydropower Plants

No posts to display