Abstract:Real-time dispatch plays a crucial role in managing the integration of large-scale variable renewable generation (VRG) into the grid. However, it necessitates enhanced requirements for renewable energy uncertainty modeling and risk management. This paper addresses this problem by realtime characterization of renewable uncertainties and building a risk-minimizing realtime dispatch model. The conditional Gaussian mixture model (CGMM) is employed to model VRGs, which can sufficiently capture multidimensional scenario-wise features. Specifically, time is modeled as a two-dimensional trigonometric series to account for the discreteness and periodicity of time features. We characterize the temporal correlation of unit output by modeling the joint distribution of prediction errors over continuous time intervals. Building upon that, we construct a two-stage risk-minimizing model for realtime dispatch, where the objective is to minimize the probability of constraint violation. The sample averaging approximation (SAA) technique is applied to transform the intractable model into a mixed-integer linear program problem. Numerical tests verify that the proposed model contributes to the efficient integration of VRGs while maintaining system stability and security.

Risk-Minimizing Real-Time Dispatch for Power Systems with High Proportion Variable Renewable Generators Based on Conditional GMM