基于深度神经网络伪量测建模的交直流混合配电网交替迭代法状态估计

当前配电网状态估计面临的一个突出问题是实时量测数量不足,难以实现全网可观性。为了给配电管理系统提供准确的基础数据,提出一种基于深度神经网络伪量测建模的交直流混合配电网交替迭代状态估计方法。首先,建立电压源换流器的稳态模型和混合配电网的实时量测模型;然后,利用历史量测数据对深度神经网络进行离线训练,建立负荷节点注入功率的伪量测模型;最后,对交流区域和直流区域进行交替迭代状态估计,在交替过程中区域间交换VSC支路状态量的估计值,保证了边界状态量的一致性。算例测试结果表明,所提方法能在实时量测覆盖率低的情况下,准确估计混合配电网的状态值。

Alternating-Iteration State Estimation of AC/DC Hybrid Distribution System Based on Pseudo-Measurement Modeling Using Deep Neural Network

At present, a prominent problem in the state estimation of distribution system is the lack of real-time measurements, which makes it difficult to realize the observability of the whole network. In order to provide accurate basic data for distribution management system (DMS), an alternating-iteration state estimation method based on pseudo measurement modeling using deep neural networks (DNN) is proposed. Firstly, the steady-state model of voltage source converter (VSC) and the real-time measurement model of hybrid distribution system are presented. Then the historical measurements are used to train DNN off-line and establish pseudo-measurement models of power injection on load buses. Finally, alternating-iteration state estimation is carried out for AC and DC areas. Only the estimated states of VSC branches need to be exchanged, thus ensuring the consistency of the boundary states. Test results show that the proposed method can accurately estimate the states of hybrid distribution system under low coverage of real-time measurements.