This doctoral dissertation addresses the development of new tools for the optimal expansion planning of power systems considering uncertainty and operational variability. Several mathematical optimization models are provided using a thesis-by-article format, where the results of the research developed by the candidate are presented in three accepted journal papers, two journal papers under review, and an accepted international conference paper.
The first paper presents a novel aggregation technique based on the maximum dissimilarity algorithm that allows modeling the operational variability in power systems, paying special attention to the representation of extreme conditions such as the peak demand level of loads. The proposed aggregation technique is used to obtain representative days (RDs) and representative weeks. Historical data of the electrical demand and the production of wind- and solar-power units are used as input data of the proposed aggregation technique, which is compared with previous clustering methods using a deterministic static transmission network expansion planning (TNEP) model.
The second paper provides a novel aggregation technique based on chronological time-period clustering that, in addition to modeling long-term dynamics throughout the entire time horizon, assigns different priorities to the input data during the clustering procedure to improve the representation of certain values, namely, local and global maximum and minimum values of the electrical demand and the production of solar- and wind-power units. The proposed clustering method, used to represent the operational variability in power systems, is compared with previous aggregation techniques using a TNEP model, a generation expansion planning (GEP) model, and a generation and transmission network expansion planning (G&TNEP) model, where all of them consider a deterministic static approach.
The third paper proposes a novel two-stage aggregation procedure and a novel exact acceleration technique for the risk-averse two-stage stochastic G&TNEP problem, where the conditional value-at-risk (CVaR) is used as a risk measure. A static approach is considered for the sake of simplicity. The uncertainty in the peak demand level of loads, the capacity of conventional and renewable generating units, and the marginal production cost of conventional generating units is modeled through a finite set of scenarios, while RDs are considered to model the operational variability associated with the electrical demand and renewable-based generation. The proposed two-stage aggregation procedure combines the aggregation techniques presented in the first two papers, reducing the resolution of the RDs and paying attention to extreme conditions. Moreover, the proposed acceleration technique is a novel relaxed version of the constraint generation-based algorithm, in which only certain scenarios are considered since not all of them are required to compute the CVaR.
The fourth paper presents a three-level model for the static two-stage adaptive robust optimization (ARO) TNEP problem considering non-convex operational constraints to model the commitment statuses of conventional generating units and to prevent the simultaneous charging and discharging of storage facilities. A cardinality-constrained uncertainty set is considered to model the uncertain parameters, while RDs are used to model the operational variability. The two-stage ARO TNEP problem is solved using the nested column-and-constraint generation algorithm (NCCGA), which is the only available exact solution procedure for two-stage ARO problems with integer recourse variables.
The fifth paper extends the work described in the fourth paper by presenting two novel exact acceleration techniques applied to the NCCGA. The first exact acceleration technique is based on the use of a heuristic solution to initialize the values of the uncertain parameters at the beginning of the inner loop of the NCCGA. Furthermore, the second exact acceleration technique is based on the iterative solution procedure of relaxed versions of the outer- and inner-loop master problems of the NCCGA. Both exact acceleration techniques can be combined.
Finally, the sixth paper extends the work described in the fifth paper by considering a dynamic approach, i.e., investment decisions can be made at different years rather than building new facilities only at the beginning of the planning horizon. The uncertainty throughout the planning horizon is modeled using a cardinality-constrained uncertainty set, where annual evolution rates are considered for the forecast values and the maximum deviations of the uncertain parameters.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados