Share with your friends

Computer models in the energy sector

Computer models in the energy sector

Part 2: Marketing the intrinsic value of power generating plants


Related content

FTM Bildwelt: Eisbär

In the last part of our newsletter on the energy sector, we presented fundamental models for the long-term modeling of the energy production landscape and the resulting price developments. These forecasts are often used to evaluate long-term investments. But how about the short-term marketing of generating plants? Fundamental models do not work very well for this aspect because their forecasts are not granular enough and in any case, the futures market provides another more reliable data source. However, the available futures market prices are sometimes not sufficient for many types of plants because their use has to be calculated on an hourly basis in order to maximize marketing results. This also requires hourly prices, which must be calculated with models. We would like to provide a more in-depth look at these models and the operational planning based on these. 

Power plants are very often marketed years before any actual power is generated, primarily for planning reasons. But how do we know how much power a power plant will be producing in a specific hour two years in advance? The deployment of the power plant depends very much on the electricity price that can be achieved in that hour.  The different classes of generation plants have to be regarded differently.  Volatile power plants, such as wind farms or photovoltaic plants are generally not managed as they produce electricity depending on environmental influences. On the other hand, there are manageable power plants such as most conventional power plants, but also biogas plants and power storage facilities. Their planned use not only depends on the price of electricity but also many other factors. How quickly can a power plant be started up or shut down? What costs are involved? Are these costs static or do they depend on the market? The answers to these questions depend very much on the investment classes. A lignite-fired power plant, being an inflexible baseload power plant that is difficult to adjust, therefore will have a very different deployment planning than a highly flexible gas-fired power plant. A gas-fired power plant can react to price signals within 15-minute intervals. As a result, the average price listed on the futures market will not be sufficient to forecast its deployment planning.

The forecasts need to be timely and exact, and the inherent fluctuations are a necessity. For this, so-called hourly price forward curves (HPFCs) have to be calculated. Usually, these HPFCs are generated as follows: The fluctuations of historic price curves are used and adjusted to the prices quoted for base and peak bands on the futures market. In doing so, the different fluctuations for workdays and weekends, specifically also bank holidays, such as Easter or Christmas, are considered. Sometimes average values are used for these so-called typological days, which is generally not recommended as this changes the variance. A peak-load power plant which is only in the money with very high prices, would be used only infrequently in this procedure and would therefore be undervalued. A better approach is therefore to analyze the irregularities and calculate them by a stochastic process. To do this, all periodic fluctuations are deducted from the historical data, as well as predictable fluctuations due to banking holidays, vacation periods or even large sports events such as the Football World Cup. What is left is a curve whose statistical characteristics are modeled with a suitable stochastic process. All previously deducted fluctuations are then added again to this modeled time series. In a last step, the prices averaged over suitable periods are adjusted to the current futures market in order to attain freedom of arbitrage. The result is a statistically solid forecast of hourly rates for the next few years.

Once the HPFC is available, it becomes possible to calculate the deployment. Just as mentioned above, a difference needs to be made between the different types of power stations. Concerning deployment planning, it is especially flexibly manageable power stations, such as coal or gas-fired plants, that are interesting. Coal-fired power plants can be operated with minimum or maximum load, depending on whether they are in the money or not. In this way, the loss due to production in unprofitable hours can be minimized. Gas-fired power plants in their most flexible form as gas engine power plants are able to switch between maximum load zero load within a quarter of an hour. This allows them to exactly trace the price signals of the intraday market. Although the number of start ups will affect maintenance negatively, these costs are not highly variable.

Deployment planning becomes more difficult with power stations that produce not only electricity. These days, where various sectors are connected, this is the case with an increasing number of plants. In addition to conventional power stations for district heating, small to medium-sized biogas plants used to heat individual buildings or neighborhoods should also be mentioned here. Beyond this, refrigeration is also increasingly being provided by electricity-generating plants. This combination of different loads causes further restraints in operational planning. During heat production, a heat accumulator is usually connected in between to ensure maximum flexibility. In such a case, every hour, the question is whether the forecast thermal load should be used from the heat accumulator (i.e. the power station does not produce it) or whether the power plant should produce electricity and heat, thus replenishing the heat accumulator. Among other things, this can lead to a scenario where a power plant produces electricity even though it is not in the money, for instance if the subsequent hours show even lower electricity prices. This connection has to be analyzed across all hours because the optimum can be determined only when looking at the entire observation periods. This requires linear programming algorithms such as the simplex method. Put simply, these try to find the most likely combinations of storing versus drawdown, and using many iterations through systematic variations try to close in on the economic optimum.

An optimal operational planning only represents the inherent value of a manageable power plant. In view of the possibility that the power station can be shut down if prices are below short-term marginal costs, this power station is a true option. In this sense, it also holds an intrinsic value in addition to a time value, which increases incrementally the closer the price of electricity is to short-term marginal costs. The time value results clearly from the asymmetrical disbursement function. If the future electricity price is lower than the forecast price, which is below the short-term marginal costs, the plant remains down and the loss is limited to the long-term marginal costs, which are incurred anyway. If the future electricity price is higher, however, there will be additional profits from higher sales proceeds. The expected value across all possible future scenarios is thus higher than for the forecast underlying the intrinsic value. The next part of our series will deal with the complex presentation of future scenarios and their implementation in the form of delta hedging.

Source: KPMG Corporate Treasury News, Edition 87, December 2018 


© 2020 KPMG AG Wirtschaftsprüfungsgesellschaft, eine Aktiengesellschaft nach deutschem Recht und ein Mitglied der globalen KPMG-Organisation unabhängiger Mitgliedsfirmen, die KPMG International Limited, einer Private English Company Limited by Guarantee, angeschlossen sind. Alle Rechte vorbehalten. Für weitere Einzelheiten über die Struktur der globalen Organisation von KPMG besuchen Sie bitte

Connect with us


Want to do business with KPMG?


loading image Request for proposal