Influence of the number of decision stages on multi-stage renewable generation expansion models

The capacity expansion problem of renewable sources faced by a central planner is essentially a long-term multi-stage decision-making problem under uncertainty. However, the size of the optimization problems describing multi-stage decision-making processes may lead to computational intractability even if a small number of stages is considered. We tackle this problem considering an explicit characterization of the long-term uncertainty and compare the outcomes of four different approaches for such problem, namely: (i) multi-stage stochastic-programming; (ii) linear decision rule (LDR); (iii) two-stage stochastic-programming under a rolling window procedure; (iv) and deterministic. The impact of considering an increasingly accurate representation of the evolution over time of the uncertain parameters is studied by solving the previous models for different planning schemes. The pros and cons of each approach are analyzed quantitatively and qualitatively using a case study based on the IEEE 24-node reliability test system (RTS). Finally, the performance of the stochastic programming and the LDR approaches is assessed by performing out-of-sample analyses.