To meet the needs of power systems for parameter estimation and tracking maintenance of transmission lines, this paper proposes an intervals estimation and tracking method for transmission line parameters based on quantile regression bidirectional gated recurrent unit (QR-BiGRU) neural network and interval robust augmented state estimation (IRASE). The proposed method is based on a dynamic augmented state estimation model in parameter tracking. Compared with static parameter estimation methods, the proposed method does not need to set up nonlinear measurement equations by multiple measurement scans to improve data redundancy, and has the advantages of easy modeling, high estimation efficiency, and better tracking maintenance; Compared with other parameter tracking methods, the proposed method does not require idealized configuration of branch measurements and only requires the system to be observable to realize the estimation, so it is more practical. The proposed method is based on interval estimation theory. Compared to point estimation, interval estimation can reflect the credibility of estimation values and the possible range of estimation errors and is considered a more secure and reliable estimation method. In addition, the estimated interval of the transmission line parameter can provide boundary information for bad data identification and provide a foundation for interval state estimation considering parameter interval. The contributions of this paper are as follows: 1) Propose an interval estimation and tracking method for transmission line parameters. The proposed method is based on the interval estimation theory and the improvement of the dynamic augmented state estimation model, including: (1) Interval estimation modeling based on measured values and their maximum allowable errors, predicted values and their prediction errors, etc. (2) Interval prediction based on QR-BiGRU neural network. (3) The improvement of robustness based on zero injection power constraints. The proposed method achieves more accurate and robust interval estimation and tracking maintenance for line parameters. 2) Proposed an interval prediction method based on the QR-BiGRU neural network. The proposed method considers the bidirectional features of historical estimated state time series data, achieving more accurate prediction. In addition, the least squares objective function of quantile regression is used as the loss function of the QR-BiGRU neural network, thereby achieving the interval prediction of the state at the next moment. 3) Propose an IRASE solution method based on interval analysis and error propagation theory. The proposed method is based on the law of error propagation and calculates the propagation values of state prediction error and redundant measurement error on parameter estimation error, thereby obtaining the parameter intervals caused by the state prediction interval and measurement interval. Finally, the multiple IEEE cases that integrated time series data from independent system operators in New York, USA, validated the following conclusions: 1) Compared with quantile regression long short term memory neural networks and quantile regression gated recurrent unit neural networks, the QR-BiGRU neural network has better interval prediction performance and accepTab.prediction rates. 2) Interval prediction based on the QR-BiGRU neural network can replace Holt exponential smoothing prediction function for parameter tracking, obtaining more accurate state prediction values and reliable state prediction intervals. 3) Zero injection power constraint could improve the robustness of the model and obtain more accurate parameter estimation results. 4) A relatively reasonable parameter estimated interval indicates the feasibility of the proposed IRASE solution method based on interval analysis and error propagation theory. 5) Compared to static augmented state estimation based on multiple measurement scans, the method proposed in this paper has higher computational efficiency and can effectively achieve long-term tracking and maintenance of transmission line parameters.

为满足现代电网对于输电线路参数估计和跟踪维护的需求，该文提出一种基于分位数回归双向门控循环单元（QR-BiGRU）神经网络与区间抗差增广状态估计的线路参数区间追踪估计方法。所提方法基于区间估计理论与动态增广状态估计模型，首先采用QR-BiGRU神经网络代替传统Holt指数平滑预测函数进行状态区间预测，并获得了更准确、可信的状态预测区间；其次基于量测值及其所允许的最大误差构建量测区间；再次，基于区间分析理论与误差传播定律求解考虑零注入功率约束的区间抗差增广状态估计模型，获取了由状态预测区间、量测区间所导致的参数估计区间，实现了对输电线路电阻、电抗的区间追踪估计；最后，通过融合了美国纽约独立系统运营商时间序列数据的多个IEEE节点测试系统，验证了所提方法的有效性。