Temporal Decomposition-Based Stochastic Economic Dispatch for Smart Grid Energy Management
Temporal Decomposition-Based Stochastic Economic Dispatch for Smart Grid Energy Management
复制标题
DOI：
10.1109/tsg.2020.2993781
复制
发表时间：
2020-09
影响因子：
9.6
通讯作者：
F. Safdarian;A. Kargarian
中科院分区：
文献类型：
--
作者：
F. Safdarian;A. Kargarian
本文提出了一种时间分解策略，将安全约束经济调度（SCED）在调度水平上进行分解，以减少其计算负担和提高其可扩展性。提出了一组子问题，每个子问题都与需求响应、正常约束和N-1次应急纠正措施有关。所提出的分解处理了源于系统设备间相互依赖的计算复杂性，即发电机的斜坡约束和存储设备的充电状态。引入重叠区间的概念，使SCED子问题可以并行求解。与能量储存相关的跨期连通性也在时间分解的背景下建模。此外，为了考虑风力发电的不确定性，将储备上下需求表述为数据驱动的非参数机会约束。$\ φ -$散度的概念用于将非参数机会约束转换为更保守的参数约束。针对风电预测误差计算降低的风险水平，以保证在不确定性真正实现后，系统约束以置信水平得到满足。将辅助问题原理应用于并行协调SCED子问题。在三个测试系统上的数值结果表明了该算法的有效性。
This paper presents a temporal decomposition strategy to decompose security-constrained economic dispatch (SCED) over the scheduling horizon with the goal of reducing its computational burden and enhancing its scalability. A set of subproblems, each with respect to demand response, normal constraints, and $N-1$ contingency corrective actions at a subhorizon, is formulated. The proposed decomposition deals with computational complexities originated from intertemporal interdependencies of system equipment, i.e., generators’ ramp constraints and state of charge of storage devices. The concept of overlapping intervals is introduced to make SCED subproblems solvable in parallel. Intertemporal connectivity related to energy storage is also modeled in the context of temporal decomposition. Besides, reserve up and down requirements are formulated as data-driven nonparametric chance constraints to account for wind generation uncertainties. The concept of $\phi -$ divergence is used to convert nonparametric chance constraints to more conservative parametric constraints. A reduced risk level is calculated with respect to wind generation prediction errors to ensure the satisfaction of system constraints with a confidence level after the true realization of uncertainty. Auxiliary problem principle is applied to coordinate SCED subproblems in parallel. Numerical results on three test systems show the effectiveness of the proposed algorithm.