Determination and Cost Allocation for Regulation Reserve With Renewables: A Data-Driven Assisted Approach

Abstract

Proper setting of regulation reserve (RR) requirements is essential for the secure and economic operation of power grids. Currently, the RR requirement is normally determined based on ad-hoc experience or numerical methods that cannot comprehensively consider uncertainties and fluctuation characters. Regarding the increasing penetration of renewables, it becomes an obvious challenge for determining the proper RR requirement and reasonably allocating the RR costs to market participants. To address these issues, a data-driven assisted determination and cost allocation method for RR is proposed. A data-driven method considering fluctuation features is used to predict the RR requirement. To avoid the insufficiency of the RR resulting from prediction error, a compensation strategy based on error decomposition is proposed. Ensemble learning is used to handle the model error. The data error caused by the multiperiod forecast uncertainty of load and renewables is compensated by a computationally tractable chance-constrained optimization problem. To encourage grid-friendly behaviors, a hierarchical method for RR cost allocation is proposed to account for the comprehensive contributions of market participants. Case studies show that compared with existing methods, the proposed RR requirement determination method can achieve a more than 24% improvement in frequency performance without excessive costs. The reasonability of the proposed cost allocation method is also illustrated.