含风电继电保护应用中的电流互感器饱和电流重构方法

Reconstruction Method of Saturation Current of Current Transformer in Relay Protection Application Related to Wind Power

传统继电保护研究往往将抗电流互感器(CT)饱和视作离线判据整定的必要条件,实际上,利用在线快速重构所得电流数据参与保护计算是一种新思路。但在风电接入系统后,流经CT的是风电与同步机的混合故障电流,然而现有的电流重构数值计算方法难以对风电间谐波进行解析建模,数据驱动方法则在计算速度和抗噪能力上有所欠缺。对此,该文提出一种含风电继电保护应用中的电流互感器饱和电流重构方法,主要分为两个阶段:离线阶段,首先结合影响CT饱和的多重因素生成样本数据集,然后训练堆叠式长短期记忆神经网络(Stacked LSTM),继而构建CT二次电流到一次电流的映射模型,再结合贝叶斯优化求取模型最优超参数;在线阶段,首先对故障后CT半周二次电流进行白噪声降噪,再将其输入离线模型重构整周一次电流,最后利用重构电流数据参与保护计算。仿真结果表明,该方法在各种CT工况下具有较强的适应性,对含间谐波和噪声的电流重构精度较高,能为继电保护提供真实有效的数据。

Conventional researches on relay protection regards anti-saturation for current transformer(CT) as a necessary condition for off-line criterion setting. Actually, a new idea is to use the on-line reconstructed current to participate in the protection calculation. However, after the wind power is connected to the system, mixed fault current of wind power and synchronous generators flows through CT. Among existing saturation current reconstruction studies, numerical calculation methods are difficult to model the interharmonics of wind turbines, and data-driven methods are lacking in calculating speed and noise immunity. In this paper, a CT saturation current reconstruction method in relay protection application was proposed, which can be mainly divided into two stages. In the offline stage, firstly,multiple factors affecting CT saturation were considered when generating the dataset. Next, the stacked long short-term memory network(Stacked LSTM) was trained to build the mapping model of CT primary current to secondary current. Then, the optimal hyper parameter was obtained through Bayesian optimization. In the online stage, noise reduction was carried out for the secondary current in the half cycle after fault, then the secondary current after noise reduction was input into the off-line model to reconstruct the primary current, which can be used in protection calculation. The simulation results show that the proposed method has good robustness and high reconstruction accuracy for current with interharmonics and noise, and can provide effective data for relay protection.