especially those related to the general capacity expansion under environmental constraints in the next decades. Rather, the model from Paper F was primarily built for over a single year to analyse the price developments in detail.
In most cases a model with for example 6 seasons and 2 possible realisations of inflow per season, would give many of the benefits of a more complex stochastic program with relative few drawbacks. Most importantly, it will only be few times larger than the current Balmorel model (the stochastic formulation includes 25 = 32 seasons vs. 12 in the current deterministic model). This would allow reasonably fast computations compared with the model in Paper F and the results would show the capacity needed to deal with a dry year in the region as well as indicate the possible price span during that year depending on the different realisations of the inflow. Also no special solution algorithm as those presented in Paper G is needed. A normal linear programming solver, as those used for solving the Balmorel model, is all that is required.
However, such a model would start with the same reservoir level every year and thus the same price for the first season. This is reasonable for the current year, but for future years modelled, the price span for the first seasons will be too narrow. Also, the price seen over the years will be less volatile than seen historically due to the longer seasons (cf. the implication on the results illustrated in Paper C).
expect to encounter. Hence, a lot of time was spend on data collecting and validating to ensure that few would reject the model due to incorrect data values for various parameters.
A related issue here is data uncertainty. Here, the issue is not that some parameters are uncertain due to their stochastic nature but rather that only estimates of the actual values can be obtained. Model results may be highly sensitive in relation to the data input. Thus, if an estimate of a parameter is adjusted with 1%, the effects on various results may be much higher. Sensitivity analyses should thus be made for all main parameters in order to see if this is the case.
Finally, the set of data to be produced needed to be one that potential users of all the represented countries generally could agree on. To succeed in this, model and data-advisors in the former east-block countries were added to be project. They have helped by collecting national data that was not otherwise available. Some of this did not exist, and had to rely on the advisors’ best estimates. However, with the help of the local advisors for data collecting, data validation, and overall model validation, the model got a dataset, that to a higher degree than otherwise possible, should satisfy potential users throughout the region.
Looking at the data required by the model, it can be divided into the following groups:
• Technological data
• Fuel data
• Temporal data
• Geographical data
• National data
The data is in general described in the Balmorel reports; see Ravn et al. (2001-I) and Ravn et al. (2001-II). Below some additional comments have been made.
Technological data: This represents data that describes the production system. The liberalisation has reduced the amount of unclassified information for instance relating to efficiencies or costs of production. In the future, more and more of this information will have to rely on estimates. To ensure consistency between different technologies in terms of costs, efficiencies, and other parameters a spreadsheet was developed as described in Chapter 5 of Ravn et al. (2001-I). This enabled factoring in the extra costs for example of having a de-sulphuring unit at a power plant so that these costs proportionally would be the same for all technologies. Especially for designing the expected future technologies, this proved very helpful.
Fuel data: This group of data relates to the emission factors of different fuels and the price developments of those. Some of the fuels are sold internationally with a unified price in all countries. These include coal and oil. Other fuels are national which differ in availability and price between the countries. Examples of these are oil shale in Estonia and peat in e.g. Finland and Sweden.
Temporal data: The temporal data is basically profiles for all parameters that have a over-the-year time-dependence, whether this is seasonal, diurnal, or both. The most important of these parameters were introduced in Section 2.3.
Geographical data: This type of information relates to the geographical structure of the power system. Most important is the transmission network and the installed capacities of different production technologies in the geographical entities in the start year and the expected rate of decommissioning. Such data is usually available from the national transmission system operators (TSO), as they need this data to ensure a reliable operation of the transmission system. So this type of data will continue to be available at the required level of detail.
National data: This data group includes estimates of electricity and district heating demands, the price elasticities of those, taxes and emission quotas, as well as the annuity factor. The elasticities and how to find them is discussed in more detail in Grohnheit and Klavs (2000) and in the Balmorel appendices, Ravn et al. (2001-II).
The annuity factor is used to describe the annual costs of an investment, i.e. annual instalments of a loan taken to make the investment. Assuming a discount rate of 10%
and a payback time of 10 years, will give an annuity factor in that country of approximate 0.16. This high discount rate and rather low expected economic lifetime of the investment (10 years) implies a high competitive market. Having a 5% discount rate and a 20-year payback time instead reduces the annuity to 0.08, i.e. the annual costs of the loan are halved. However, such conditions are in general only available in regulated markets where the value of the investments in the future is better known.
In conclusion regarding the data collection work, the dataset currently (October 2002) available for the Balmorel model version 2.10 is reasonable in terms of accuracy and consistency. As well as the actual numbers, a documentation of the data has been made. This describes the assumptions, any transformations done, and where to find the data sources in order to make future updates easier. Making the documentation proved to be more time consuming than expected.
As for possible improvements, much information related to CHP and district heating in the former East block countries still rely on guesswork, but this should improve as traditions are created in those countries for collecting that type of data. Also, the estimation of the price elasticities of electricity in the countries must remain uncertain as in general only little experiences with price vs. demand behaviour of electricity have been obtained in the various countries up till now. Again, this should improve in the future. Finally, sensitivity analyses should be made of more parameters than it has been done till now to check how important the quality of the estimates are for different parameters, such as those related to elasticities.