In the adequacy assessment model, net generating capacity is divided into the following categories: large conventional power plants, hydro power plants, wind power plants and other generating capacity (including small-scale CHP and solar PV generating capacity). Adequacy indicators are computed using an analytical method. Depending on the power plant categories, two-state models, multi-state models and aggregated firm capacity are used for the calculation of COPT.

Large conventional power plants. Large power plants with a nominal generating capacity of above 50 MW are modelled per unit using a two-state model. In the case of existing units, FOR and maintenance requirements were estimated on the basis of past plant availability data (period of 1990-2008) and retrospective system balances [12]. For prospective analysis, assumptions on the plant availability data of projected new units have to be made, as well.

Depending on the technology, the temperature sensitivity of the output power and the effects of the heating period were also considered. (In Hungary, heating period is usually between 15 October and 15 April.) De-rated capacities were assumed for a number of CHP CCGT units in the summer months, since these can be operated with auxiliary cooling only when heat demand is low. Beyond that, the available generating capacity of non-CHP units is also influenced by ambient air temperature due to changes in gas turbine inlet air temperatures, cooling water temperatures, etc. In order to reflect these seasonal effects, nominal net generating capacities were multiplied by factors expressing the ratio of available and nominal capacity. These factors were estimated on a monthly basis based on past plant operation data.

Hydro power plants. For the assessment of the relatively small total generating capacity of run-of river hydro power plants, average monthly capacity factors can be used. These capacity factors can be calculated from past monthly production data [11]. Table 2 contains average monthly values for period 2002-2009, for wet, average and dry years.

Wind power plants. In order to assess the weather-dependent production patterns of wind power plants, a multi-state model was elaborated based on the approximate method proposed by D’Annunzio and Santoso [7], [8]. This method assumes that wind production can be properly modelled by a large number of discrete output states (partial capacity outage states) for adequacy assessment purposes. Instead of using synthetic data, the method is based on power output data.

The 15-minute resolution time series of total wind production in Hungary were collected as input from the years 2011 and 2012. Because of the intra-annual increase of the total installed wind generating capacity, time series were normalised and load factors were calculated. As an approximation, the load factor time series can be used for projected higher net generating capacities, but limitations (e.g. smoothing effect, influence of geographical distribution) should be kept in mind. In addition, aggregated on-line measurement data are also subject to errors, effects of maintenance, outages, trial runs under commissioning period, etc.

The multi-state representation of wind (Table 3) uses a load-factor resolution of 0.02, considering 51 capacity outage states. (This resolution is equivalent to 6.48 MW in case of the net installed capacity at the end of 2012.)

Solar power plants. Due to the negligible present total net generating capacity, there are no aggregated online measurement data available for analysis. As a simplified approach, hourly production time series were estimated and taken into account on the basis of 2011 irradiation data. This approach can be justified by the moderate development targets (see Table 4).

COPT values express the cumulative probability P(X ≥ x) of having a system capacity outage greater than or equal to x. Values of COPT can be computed using a recursive algorithm [8]. When adding the new two-state unit i +1 with FOR of FOR_i+1 and net generating capacity of C_i+1 to the system containing already i units, the following formula can be used:

In the case of multi-state modelling applied for wind power plants, considering m partial capacity outage states with individual probabilities of p_j and capacity outage states C_j, the capacity addition formula is as follows:

Since capacity outage states are given by load factors in Table 3, they need to be up-scaled according to the total net wind generating capacity to be considered.

In the model, capacity outage probability tables were calculated for large power plants, whose availability is described by a two-state model, and for wind power plants (considered as an aggregated multi-state unit) on a monthly basis. Unavailable capacity of hydro power plants, solar power plants and other small-scale units were taken into account as a fixed input.

For the calculations presented here, past hourly system load time series were used, scaled up according to the forecasted annual demand. As well known, system load patterns exhibit annual, daily and weekly seasonality, and in addition to that, both long (e.g. economic framework) and short term effects (e.g. weathering conditions, calendar effects) have a strong influence on system load characteristics. For temperature sensitivity analysis purposes, also a statistical, multiple regression based load model is available.

The applicability of statistical models adjusted to present system load patterns will be very limited in case of a very long time frame; therefore new approach has to be elaborated in order to take into account the effects of demand side management, heat pumps, electric vehicles, etc., properly.

Concerning RES generating capacity, both Scenario A and B assess the expected power balances adjusted to the National Renewable Action Plans submitted by Hungary to the European Commission, as required by Article 4 of Directive 2009/28/EC (Table 4). This document is indicative only for Hungary, since it is subject to revision according to Government Decree 1491/2012. (XI.13.).

For the prospective analysis of thermal and nuclear generating capacity, the existing power plant portfolio, the expected decommissioning of power plants and possible new capacity additions should be assessed. In the SO&AF scenarios, the existing power plant portfolio at the end of 2012 was taken as a basis. The decommissioning of large power plant units was assessed individually, based on the technical lifetime of power generating assets indicated in the licences issued by the regulatory authority. For smaller power plants, a decommissioning rate was estimated. For large-scale new capacity additions, the information on new projects included in the 2012 generation adequacy report prepared by the transmission system operator was used.

Table 5 contains the total net generating capacity and number of two-state units assumed for each category. Smaller thermal power plants were modelled separately in an aggregated way, as described in the methodology; therefore their net generating capacity is not covered in Table 5. Units already mothballed were also not taken into account in the table. (As a result of unfavourable market conditions, several units with a lower thermal efficiency were mothballed recently, reaching a total unavailable capacity of 1.8 GW by the end of 2012.)

As described earlier, Scenarios A and B are identical for 2015, since these forecasts look only one year ahead that is negligible compared to the longer lead time of large thermal and nuclear investments. However, the thermal generating assumptions for the 2020 scenarios differ significantly in gas fired net generating capacity.

Scenario A reflects a pessimistic approach where already existing units that are expected to remain in operation until 2020 are only considered. Scenario B as a best estimate projection takes into account three further CCGT units with a total of 1,339 MW of net generating capacity. Various new gas fired CCGT projects were developed in the recent years including both green-field investment and CCGT-upgrade of existing conventional units, but final investment decisions are delayed due to unfavourable market conditions. It was assumed that three of these projects can be realised until 2020 in order to ensure a satisfactory level of national generation adequacy.

The assumptions for nuclear generating capacity are the same, since the existing units of Paks nuclear power station are expected to be decommissioned after 2030. Two units of the lignite fired Mátra power station can be retired before 2020 (reflected in both Scenario A and B), while no new lignite based project is considered at the moment.

Probabilistic generation adequacy analysis was done for the years 2015 (mid-term) and 2020 (long-term), both for Scenario A and B. A one-hour resolution model was used, containing hourly system load, average production of hydro power plants, and hourly aggregated production profile of solar PV generators, small-scale RES and non-RES power plants. For wind production and large power plant units, the COPT values were calculated on a monthly basis, considering maintenance and partially unavailable capacity due to seasonal effects, as well.

Table 6 compares the COPT calculation results for 2015 (Scenarios A/B) and 2020 (Scenarios A and B). Since the calculation was done on a monthly basis, only the probability values computed for December are shown in the table. As described earlier in the methodology, COPT values express the cumulative probability P(X ≥x) of having a system capacity outage greater than or equal to x.

The probabilities of having a certain level of system capacity outage are determined by the total net generating capacity and its composition (share of variable generation, individual forced outage rates of the units). As shown in Figure 2, Scenario B, year 2020 has the highest net generating capacity, while Scenario A, year 2020 has the lowest installed capacity. The level of variable generation differs slightly: 580 MW (Scenario A/B 2015) and 750 MW (Scenario A/B 2020). As a result of all these influencing factors, the COPT values computed for Scenario B 2020 (largest net generating capacity) are the highest, while the probabilities for Scenario A/B 2015 (lower level of wind generating capacity) are the lowest.

Based on the COPT data, annual LOLE (h) values and LOLP indices at the annual peak load were calculated, as shown in Table 7. (The LOLP indices are given for the hour of the annual peak load).

The reliability indices can help to show to what extent the power system is supposed to rely on imported electricity in the three cases that were examined. Based on the adequacy indicators, a slight shortage of domestic generating capacity is expected by 2015, since the LOLP at the annual peak load (without import) exceeds the reference value of 1%. However, more serious electricity import dependency is foreseen for 2020 if no domestic investments will take place. In case of the conservative capacity forecast Scenario A, both LOLE (h) and LOLP values for the year 2020 are extremely high. The LOLP value of near 100% suggests that the load is very likely to exceed the domestically available generating capacity at the annual peak time. Depending on the future evolution of demand and supply side, a close monitoring of the availability of imported electricity might be necessary.