基于数据驱动场景建模的综合能源系统调度方法

Data-Driven Scenario Generation-Based Optimal Scheduling Method for Integrated Energy Systems

摘要: 在“双碳”目标的大背景下，耦合了多种异质能源的综合能源系统(Integrated Energy Systems, IES)在可再生能源消纳、能源梯级利用等方面具拥有巨大优势。然而，可再生能源–负荷需求的双侧不确定性对IES的优化调度方法提出了巨大的挑战。针对该问题，本文提出了一种基于数据驱动场景建模的IES日前优化调度方法。在海量功率数据的基础上，建立了基于变分自编码器(Variational Auto-Encoder, VAE)的IES源–荷场景生成模型，并通过改进的K-means聚类算法进行场景削减，得到精确的典型运行场景。随后，建立了以运行成本期望最低为目标函数的IES日前场景随机优化调度模型，有效提升了IES的运行经济性，为不确定性条件下的IES调度提供了理论指导。

Abstract: Integrated Energy Systems (IES) coupled with diverse energy sources offer significant advantages in both renewable energy consumption and energy gradient utilization within the context of the “double carbon” target. Nevertheless, the dual-sided uncertainty in renewable energy-load demand significantly affects the optimization methodology of IES scheduling. To tackle this issue, the paper introduces a data-driven optimal scheduling method for IES, based on scenario generation. Utilizing extensive power data, an IES source-load scenario generation model using Variational Auto-Encoder (VAE) is developed, followed by scenario reduction through an enhanced K-means clustering algorithm, enabling the extraction of precise typical operating scenarios. Consequently, a stochastic optimal scheduling model is formulated for IES day-ahead scenarios, aiming to minimize operating costs, thereby enhancing the operational economy of IES, and offering theoretical guidance for scheduling under uncertain conditions.