Abstract
Wind investors in network-constrained electricity markets must make siting and capacity decisions under uncertain demand and wind availability, while market prices and congestion respond to these decisions. This paper examines how alternative representations of these uncertainties affect optimal wind investment and system operation. A stochastic bilevel model is used, with upper-level discrete investment decisions and lower-level lossless DC market clearing. The model is reformulated as a Mathematical Program with Equilibrium Constraints (MPEC) and subsequently as a Mixed-Integer Linear Programming (MILP) problem. Two published scenario sets based on 2007 Iberian Peninsula demand and wind data are compared: 36 duration-curve operating conditions formed from four load blocks and combinations of three demand and three wind levels, and 48 correlated K-means centroids representing joint observations from three demand zones and two wind zones. A 3-bus verification case and eight IEEE 24-bus RTS cases examine investment budgets and spatial/network variations. In the directly comparable budget-constrained cases, 1000 MW is installed under either representation, but its allocation among four candidate buses changes from 300/300/200/200 MW to 100/300/300/300 MW, while the reported expected annual profit differs by 10.20%. Across the four reported 24-bus comparisons, weighted-average system locational marginal prices differ by 3.60-3.71 EUR/MWh. The scenario representations can therefore lead to different investment and network outcomes, although the comparisons involving externally imposed wind-availability factors must also be interpreted in light of their differing spatial assumptions.