Survey and Comparison of Optimization-Based Aggregation Methods for the Determination of the Flexibility Potentials at Vertical System Interconnections

Abstract

The aggregation of operational active and reactive power flexibilities as the feasible operation region (FOR) is a main component of a hierarchical multi-voltage-level grid control as well as the cooperation of transmission and distribution system operators at vertical system interconnections. This article presents a new optimization-based aggregation approach, based on a modified particle swarm optimization (PSO) and compares it to non-linear and linear programming. The approach is to combine the advantages of stochastic and optimization-based methods to achieve an appropriate aggregation of flexibilities while obtaining additional meta information during the iterative solution process. The general principles for sampling an FOR are introduced in a survey of aggregation methods from the literature and the adaptation of the classic optimal power flow problem. The investigations are based on simulations of the Cigré medium voltage test system and are divided into three parts. The improvement of the classic PSO algorithm regarding the determination of the FOR are presented. The most suitable of four sampling strategies from the literature is identified and selected for the comparison of the optimization methods. The analysis of the results reveals a better performance of the modified PSO in sampling the FOR compared to the other optimization methods.

Monetarization of the Feasible Operation Region of Active Distribution Grids Based on a Cost-Optimal Flexibility Disaggregation

“…These are stochastic approaches using random search (RS) and optimization based approaches solving an adapted optimal power flow (OPF) problem (cf. [2], [23]).…”

“…flexibility provision per bus, bus voltages) for each point within the FOR (see Fig. 3a) and the simple consideration of non-convex FPU flexibility polygons [2], [23]. Another advantage is the good performance of the RS approaches for larger systems.…”

“…The significance of the advantages of the OPF-based approaches depends on the specific optimization method for the solution of the OPF problem. Linear or quadratically constrained linear programming approaches (LP) determine the OPF edge in short computation times [8], [9], [23], [30], [33], [34]. The non-linear system behavior is only approximated which can lead to local VOLUME 10, 2022 convergence and by this to an over-or underestimation of the FOR compared to the FOR resulting from other optimization methods [2], [8].…”

“…The performance of LP scales proportionally for larger systems but the quality of the results decrease due to the increasing number of non-linearities [30], [35], [36]. Nonlinear programming (NLP) approaches lead to an appropriate sampling of the FOR in short computation times also identifying non-convexities of the FOR [9]- [11], [23], [25]. The performance and the quality of the results of the NLP depends on the system size and the specific solver (e.g.…”

“…Approaches using a metaheuristic (e.g. particle swarm optimization, genetic algorithm) combine the advantages of RS in generating metadata and the appropriate sampling of the FOR-edge of OPF-based methods [9], [23], [28]. In general, solving an OPF problem by metaheuristics is based on an evaluation of the objective function by a variety of power flow calculation, analogously to RS approaches.…”