基于多准则分区的有源配电网双层式状态估计研究

In order to overcome the drawbacks (computation efficiency, estimation accuracy and analytical capacity of Distributed Generations (DGs)) of the existing state estimation methods utilized in distribution system, a novel state estimation method utilized in active distribution network is firstly proposed based on the combination of multi-criteria partition algorithm and double-layer estimation technique. For the real-time ability, the multi-criteria segment algorithm is employ to divide the subsystems by considering the topology structure, operating characteristics and measurement allocations in the active distribution network. The parallel computation method is proposed to implement the subsystem state estimation based on the different structure and measurement information. For the robustness, the factor graph based on the Bayesian Network which contains the behavior layer and observation layer is built to characterize the correlations among the state variables of different buses. The Recursive Belief Propagation Algorithm (RBPA) is conducted on the factor graph to analyze the integrations of DGs and to complete the state estimation of the active subsystem. For the accuracy, the simplified model of the active distribution network is constructed by implementing the equivalents of all subsystems based on the lower-layer estimation. Furthermore, the synthesized robust estimation model of upper-layer estimation and its corresponding solution algorithm are presented based on the conversions of the different measured data, utilization of the different weighted factors which calculated by the considerations of structure space and observation space. Finally, the real-time ability, robustness and accuracy of the proposed state estimation method in this project are validated based on the simulation model and laboratory platform.

针对现有配电网整体状态估计算法在计算效率、估计精度和分布式电源处理能力等方面的局限性，本项目提出了多准则分区的有源配电网双层式状态估计算法。在算法实时性方面，基于有源配电网的可观性指标和量测设备配置情况，提出多准则的有源配电网分区模型和相应的多决策子系统划分方法；基于差异化子系统的量测数据冗余度，分析异构数据的并行状态估计方法。在算法鲁棒性方面，基于贝叶斯网络结构，建立行为层和观测层相结合的状态估计因子图；基于分布式电源和负荷的状态概率分布模型，提出适合于贝叶斯因子图的递推置信传播状态估计算法；在算法抗差性方面，基于不同量测数据之间的转化方法和子系统下层状态估计的等效模型，构建兼顾结构空间和观测空间的有源配电网综合抗差状态估计模型，通过引入权重因子，实现有源配电网的上层状态估计。在算法验证方面，利用建立的有源配电网仿真模型和测试平台，全面评估多准则分区方法和双层式状态估计算法的性能。

大量分散分布式电源的接入在增大配电网规模的同时，也使得配电网的运行方式更加多样化和复杂化，而现有的配电网状态估计算法不能满足电网监测实时性的要求。本课题针对现有配电网整体状态估计算法在计算效率、估计精度和分布式电源处理能力等方面的局限性进行研究和改进，提出适用于大量分布式电源接入的配电网实时状态估计。.（1）在有源配电网分区方面：基于对有源配电网中实时量测、虚拟量测和伪量测的配置分析，建立了适用于复杂有源配电网状态估计的多准则分区优化模型，该模型可实现对复杂有源配电网的合理分区，同时还可有效减少状态估计问题的系统规模，提高计算效率。.（2）在有源配电网的子系统状态估计方面：引入高斯混合模型模拟分布式电源出力的不确定性，进一步采用因子图表征量测变量和状态变量之间的关联关系，然后通过量测方程中的线性变换和引入置信传播算法求解子系统中各个节点直接的相关性，得到子系统状态估计的结果。每一个子系统可以并行计算，大大减小了有源配电网状态估计的时间；.（3）在算法的抗差性方面：通过多准则分区方法，将有源配电网的各子区域完全解耦，有效减小了系统规模和雅可比矩阵阶数，根据子区域量测数据精度自动选取适合于该子区域的状态估计方法，提出了一种基于融合加权最小二乘法和原对偶内点法的有源配电网自适应抗差状态估计方法。.（4）在上下层的协调状态估计方面：基于分区的子系统状态估计结果，构建了一种协调下层子系统状态估计和上层有源配电网整体状态模型，结合通过Karush Kuhn Tucker方程计算得到的残差，得出配电网整体状态估计的结果。