综合能源系统通过电、热、冷、气等多种能源广泛互联并紧密互动实现了更可靠、更经济、更灵活的能源供应。综合能量管理是该系统实现安全与经济运行的关键技术。近年来，综合能源系统呈现出从园区级向城市级发展的趋势，使得综合能量管理面临一系列新的技术挑战。在此背景下，本文针对大规模、多主体的综合能源系统，开展了综合能量管理的理论与应用研究。首先，提出了面向综合能源系统统一、高效建模的能路理论。该理论基于电路比拟的思想，将电网建模的方法论应用于气网建模和热网建模，实现了“电路”向“能路”的推广，其核心在于通过傅里叶变换和二端口等值实现能路模型在物理上从时域到频域、从分布参数到集总参数的变换，在数学上从偏微分方程到常微分方程、再到代数方程的化简。相比时间和空间维度上的双重离散差分实现模型代数化，能路模型具有数学形式统一与计算性能高效的显著优势。其次，基于能路模型，提出了综合能源系统的时频混合能流计算模型及其分解并行算法。其中，前者为电网的时域模型和气网、热网的频域模型搭建了耦合接口，以更少的变量与方程完成了联合能流建模；后者通过三层迭代实现了能流计算在不同网络之间的分解、在不同时段或频率分量之间的并行，且具有收敛性保证。时频混合模型与分解并行算法大幅提升了大规模综合能源系统的能流计算效率。然后，基于能路模型，提出了综合能源系统的时频混合优化调度模型及其高效压缩方法。其中，前者用频域中的能路网络约束替代了时域中的偏微分网络约束，以避免大量中间变量与差分约束的引入；后者通过隐变量空间投影消去海量监控变量、通过约束生成算法消去海量冗余安全约束。上述变量与约束规模的削减在不影响模型最优性的情况下大幅提升了大规模综合能源系统的优化调度效率。最后，在多主体综合能源系统中考虑个体理性，提出了转移支付策略及其可信计算方法。其中，前者通过收益再分配达成了总体收益最优前提下的个体互利目标；后者在不完全信息条件下将个体收益表达为边界变量的投影函数，通过偏置漂移实现收益数据的隐私保护、通过连续性校验实现数据篡改行为的检测，从而达成了各主体对转移支付计算的互信目标。互利与互信的达成有力促进了多主体综合能源系统的可持续协同。
Through the extensive connection and active interaction of various energy vectors such as electricity, heat, and natural gas, integrated energy systems (IESs) have made energy supplies more reliable, economical, and flexible. Integrated energy management (IEM) is the key technology to ensure the safe and economical operation of IESs. In recent years, the trend of IES development from park scale to city scale has brought a series of technical challenges to IEM. Under such a circumstance, this paper conducts theoretical and application studies on IEM for large-scale and multi-agent IESs.First, the energy circuit theory that models IESs in a unified and efficient manner is proposed. Based on the idea of circuit anaology, this theory applies the modeling methodology of electricity networks to natural gas networks and heating networks and thereby generalizes the electric circuit to the energy circuit, whose core steps are Fourier transform and two-port equivalence to transform the physical models from the time domain to the frequency domain and from distributed parameters to lump parameters, and simplify the mathematical models from partial differential equations to ordinary differential equations and then to algebraic equations. Compared with discrete difference operation on both temporal and spatial dimensions to realize algebraization, the energy circuit models have the advantages of unified mathematical forms and efficient computational performance.Second, based on the energy circuit models, an energy flow model on the both time and frequency domains for IESs and its decomposed and parallel algorithm are proposed. The proposed model builds a coupling interface for the time-domain model of electricity networks and the frequency-domain model of natural gas networks and heating networks, which uses fewer variables and equations to complete the joint energy flow modeling. The proposed algorithm helps the energy flow calculation achieve decomposition among different energy networks and parallelization among different time periods or frequency components through tri-level iterations, whose convergence is guaranteed. These technologies significantly improve the efficiency of energy flow calculation for large-scale IESs.Third, based on the energy circuit models, an optimal dispatch model on the both time and frequency domains for IESs and its compaction methods are proposed. Compared with the traditional model, the developed model replaces the time-domain network constraint in partial differential equations with the frequency-domain network constraint in algebraic equations to avoid the introduction of numerous intermediate variables and difference constraints. In addition, the variables in this model are compacted as the implicit variable space projection method removes all monitored variables, and the constraints in this model are compacted as the constraint generation algorithm removes the most redundant security constraints. These technologies significantly improve the efficiency of optimal dispatch for large-scale IESs without sacrificing optimality.Finally, considering individual rationality in multi-agent IESs, a strategy of transfer payment and its credible calculation method are proposed. By redistributing the payoffs, the transfer payment achieves the goal of individual mutual benefit under the premise that the total payoff maintains optimality. Involving the incomplete information condition, the projection functions that map boundary variables to individual payoffs are derived. Based on these projection functions, privacy protection of individual payoffs is realized through bias drifting and data tampering for extra profits is detected through continuity check, so that the mutual trust of agents on the transfer payment calculation is ensured. The achievement of mutual benefit and mutual trust strongly promotes the sustainable coordination in multi-IESs.
