分布式多元随机动态场景生成及快速精准场景降维算法

Distributed Multivariate Random Dynamic Scenario Generation and Fast&Accurate Scenario Simplified Algorithm

高比例新能源及多源耦合是电力系统发展的重要特征,这也为系统稳定经济运行提出了新挑战。该文以园区型多能系统为对象,研究了分布式多元随机动态场景分析,从多时空角度有效量化不确定因素给系统造成的影响,可为系统灵活重构、多维度协同运行与决策提供有力模型与场景支撑。首先由预测误差驱动拟合多元功率预测误差概率分布,全面反映随机功率出力信息,提高模型泛化性;以时序相关范围参数为数据驱动关联变量,高效动态控制波动强度;最终场景生成利用逆变换映射思想保证置信度。然后针对典型场景提取,提出一种综合递归聚类思想的多段嵌套削减算法,结合改进Wasserstein距离指标,兼具准确、时效、稳定方面的优势。最后由对比实验论证该方法的前沿有效性。

High penetration of renewable energy and multi-energy coupling,the prominent characteristics of the development of power system,have also brought new challenges to the safe and economic operation of the system.This paper takes the community integrated energy system as the object to discuss a distributed multivariate random dynamic scenario analysis,which effectively quantifies the impact of uncertainties on the power system from a multi-temporal and spatial perspectives.This will provide power system with strong modeling and scenario support for applications as flexible reconfiguration,multi-dimensional cooperative operation and decision-making.First,the multi-energy prediction error probability distribution fitting is performed driven by the forecast errors,which can fully reflect the random power output information,thus improving the generalization performance of the model.Using time-series related range parameters as data-driven correlation variables,the fluctuation intensity can be controlled efficiently and dynamically.The final scenario generation adopts the mapping idea of inverse transformation to ensure the confidence.Second,as for the optimal scenario reduction,a recursion&cluster multi-stage nested reduction algorithm is proposed,which combines the improved Wasserstein distance index,possessing comprehensive advantages in accuracy,timeliness and stability aspects.Finally,the frontier and effectiveness of the method are verified through comparative experiments.