Probabilistic maximization of time-dependent capacities in a gas network

Abstract

The determination of free technical capacities belongs to the core tasks of a gas network owner. Since gas loads are uncertain by nature, it makes sense to understand this as a probabilistic problem as far as stochastic modeling of available historical data is possible. Future clients, however, don't have a history or they do not behave in a random way, as is the case, for instance, in gas reservoir management. Therefore, capacity maximization turns into an optimization problem with uncertainty-related constrained which are partially of probabilistic and partially of robust (worst case) type. While previous attempts to solve this problem had be devoted to models with static (time-independent) gas flow, we aim at considering here transient gas flow subordinate to a PDE (Euler equations). The basic challenge here is two-fold: first, a proper way of joining probabilistic constraints to the differential equations has to be found. This will be realized on the basis of the so-called spherical-radial decomposition of Gaussian random vectors. Second, a suitable characterization of the worst-case load behaviour of future customers has to be figured out. It will be shown, that this is possible for quasi-static flow and can be transferred to the transient case. The complexity of the problem forces us to constrain ourselves in this first analysis to simple pipes or to a V-like structure of the network. Numerical solutions are presented and show that the differences between quasi-static and transient solutions are small, at least in these elementary examples.