Constructing renewable-dominated power systems under decision-dependent uncertainty : expansion planning and resilience enhancement

Abstract

Constructing renewable-dominated power systems faces challenges from several perspectives. First, the enlarged renewable scale aggravates the inverse impacts of renewable generation capacity on renewable uncertainties, indicating the decision-dependent uncertainty (DDU) challenging the expansion planning of renewables and flexibility resources. Second, long-distance transmission between unevenly distributed renewable resources and load expands the interface threatening electric infrastructures by extreme weather events, putting forward higher requirements on enhancing the resilience of renewable-dominated systems. This thesis focuses on the expansion planning and resilience enhancement of renewable-dominated power systems under DDU, and conducts the following research works. First, a sensitivity analysis method for assessing the value of renewable uncertainty is proposed. Different from previous work where the renewable forecasting accuracy is perturbed numerically, the proposed method directly derives the sensitivity formulae with respect to global and nodal renewable forecasting accuracy change leveraging the explicit expression of statistical moments of uncertain renewable power output in the Point Estimate Method (PEM)-based operation model. The proposed method provides an effective tool to assess the value of improving forecasting accuracy and identify critical nodes with uncertain injections that are most influential on system operation. Second, to explore how the expansion planning paradigm would change under the spatial smoothing effect-induced DDU, the DDU is incorporated into the expansion planning of large-scale wind farms and flexibility resources. i) A stochastic wind farm expansion planning model with DDU is established based on PEM. A DDU model is established, quantifying the coupling relationship between expansion decisions and the DDU. Via PEM-based models where decisions and decision-dependent statistical moments of uncertain parameters are explicitly expressed, an in-depth discussion on the structural features of the expansion model with DDU is conducted. An iterative solution method with convergence analysis is proposed. ii) A two-stage stochastic model for coordinated wind power generation planning, battery energy storage (BES) planning, and the retirement scheduling of generators is proposed considering the DDU. The DDU model is further extended to capture the spatial correlation among multiple wind farms using Gaussian Mixture Model. An affine function-based solution method is proposed to map decisions to decision-dependent wind power scenario sets. iii) An analytical model reformulation approach for optimization models with DDU is investigated. Specifically, chance-constrained co-expansion planning for power systems is proposed, where wind farm capacity, BES capacity, and power transmission lines are co-expanded. DDU-based chance constraints are reformulated analytically via the approximated polynomial wind power probability distributions. The DDU-based chance-constrained co-expansion planning model is then converted into a mixed-integer second-order cone program. Last, to enhance the resilience of renewable-dominated power systems against extreme weather events, a joint maintenance and operation model for wind power-penetrated systems after sandstorms is proposed. The specific failure features of overhead transmission lines (OTLs) due to post-sandstorm insulator flashover are explored. Considering the inherent dependency of the uncertain availability of OTLs on maintenance decisions, the multi-period maintenance process of OTLs is modeled as a decision-dependent stochastic process. A unique modeling transformation technique is provided to convert the established decision-dependent stochastic model into a computationally efficient form.