An important aspect of the simulation of complex systems is to treat each equipment at a similar level of detail, as long as the simplifications involved do not affect the result in a decisive way.

Computer programs for treating energy supply systems available on the market generally tend to treat the system units in much detail and the development tendencies are to increase this even more. The underlying idea is that the model becomes more realistic in this way. However, in general the uncertainties affecting the calculations are not accounted for. They may have an important impact on the results. Uncertainties stem from stochastic variables, e. g. the weather or component lifetimes, and insufficient knowledge,

e. g. the lifetimes of fuel cells.

For this reason the following uncertain parameters will be treated in the simulation:

■ Component failures and repair times

■ Sunshine duration and intensity

■ Electricity demand

The basic computer model used is TRYNSYS® , which was embedded in an environment for treating data uncertainties. TRYNSIS offers standard components for most of the necessary equipment, i. e. photovoltaic arrays, wind turbines, energy converters, fuel cells, electrolysers, storage systems and controllers. Further components were added to TRYNSIS in the context of the Hydrogems® project [2]. They are described in detail in [3]. The program IsIIBat is used as graphic user interface (cf. Figure 1).

Modifications were made to the code, especially for enabling one to account for component failures and repair times.

Component failure probabilities are treated, as usual, using the exponential distribution,

i. e. failures are considered as random events

P {ri < t}= F (t) = 1 — e_v t > 0 (1)

In eq. (2) pi is the mean value of the logarithms of Ai and si2 the corresponding variance. The Monte-Carlo method [4,5] is used for the calculation. It is based on generating realizations of the variables Ti and Ai on the basis of eq.(1) and (2). These are produced from random numbers, Zn, i uniformly distributed on [0,1] by the following transformations

where

Z. = expQ-2 • ln Z21 • cos(2^ • ln Z3i) • st + д.) (4)

The TRNSYS model is evaluated with the above input variables and its result is recorded. The procedure is repeated a total of N times with N=40000, a number deemed sufficient for obtaining a reliable result.

Five years are taken as time horizon. As long as all components work, the system functions properly. In case of component failure it is checked whether the system still works or breaks down as a consequence. If it breaks down this status will prevail until the repair time of the component has elapsed and the component is returned to its functioning state.