Abstract In this paper, a hybrid robust forecasting-aided state estimator (FASE) is proposed that can handle bimodal Gaussian mixture (BGM) distribution of PMU noise, bad data and sudden load changes. It is shown in this paper that the traditional methods will be biased in the presence of BGM noise of PMU measurements. To this end, the generalized-maximum likelihood cubature Kalman filter (GM-CKF) is developed and compared with existing GM-EKF, GM-UKF, CKF, EKF, UKF, and static state estimator (SSE) in different system operating scenarios. It is demonstrated that GM-CKF has better estimation accuracy than all other methods in the presence of BGM errors, and is more stable than GM-UKF and UKF. However, its estimation accuracy is lower than the other state estimators except for CKF in the initial estimation stage and when an unexpected sudden change occurs. This result also means that the GM-CKF is highly sensitive to the anomalies and is condusive for anomaly detection. Finally, a hybrid robust FASE method is proposed that balances well the trade-off between GM-CKF and SSE. Simulation results carried out on several IEEE benchmark systems demonstrate the effectiveness as well as the robustness of the proposed hybrid method.

中文翻译：

---

一种考虑双峰高斯混合测量误差的混合鲁棒预测辅助状态估计器

摘要 在本文中，提出了一种混合鲁棒预测辅助状态估计器 (FASE)，它可以处理 PMU 噪声、不良数据和突然负载变化的双峰高斯混合 (BGM) 分布。本文表明，传统方法在 PMU 测量中存在 BGM 噪声时会产生偏差。为此，开发了广义最大似然容积卡尔曼滤波器 (GM-CKF)，并与现有的 GM-EKF、GM-UKF、CKF、EKF、UKF 和静态估计器 (SSE) 在不同系统运行场景下进行了比较。证明在存在 BGM 误差的情况下，GM-CKF 比所有其他方法具有更好的估计精度，并且比 GM-UKF 和 UKF 更稳定。然而，在初始估计阶段和发生意外突变时，其估计精度低于除 CKF 之外的其他状态估计器。这一结果也意味着 GM-CKF 对异常高度敏感，有利于异常检测。最后，提出了一种混合鲁棒 FASE 方法，可以很好地平衡 GM-CKF 和 SSE 之间的权衡。在几个 IEEE 基准系统上进行的仿真结果证明了所提出的混合方法的有效性和鲁棒性。