Accurate sag source location and precise sag type recognition are both essential to verifying the responsible party for the sag and taking countermeasures to improve power quality. In this paper, an attention-based independently recurrent neural network (IndRNN) for sag source location and sag type recognition in sparsely monitored power system is proposed. Specially, the given inputs are voltage waveforms collected by limited meters in sparsely monitored power system, and the desired outputs simultaneously contain the following information: the located lines where sag occurs; the corresponding sag types, including motor starting, transformer energizing and short circuit; and the fault phase for short circuit. In essence, the responsibility of the proposed method is to automatically establish a nonlinear function that relates the given inputs to the desired outputs with categorization labels as few as possible. A favorable feature of the proposed method is that it can be realized without system parameters or models. The proposed method is validated by IEEE 30-bus system and a real 134-bus system. Experimental results demonstrate that the accuracy of sag source location is higher than 99% for all lines, and the accuracy of sag type recognition is also higher than 99% for various sag sources including motor starting, transformer energizing and 7 different types of short circuits. Furthermore, a comparison among different monitor placements for the proposed method is conducted, which illustrates that the observability of power networks should be ensured to achieve satisfactory performance.

Voltage sag state estimation on the basis of a limited number of installed monitors is essential to dividing the responsibility for the voltage sag and taking corresponding measurements for improvement in voltage quality. Therefore, a deep learning methodology via bidirectional WaveNet for the voltage sag state estimation is proposed in this paper. The presented method can simultaneously estimate voltage sag state at non-monitored buses via limited monitors. Especially, the proposed deep learning method using the bidirectional WaveNet is designed to explore the long-term and long-range temporal dependencies in both the forward and backward directions. In this way, only by using original measured voltages through monitors, high accuracy for voltage sag state estimation can be achieved without restructured or redesign of the raw monitored data. An excellent advantage of the presented algorithm is that it can be implemented without system parameters or operating conditions or any other prior information. The presented methodology was verified by the IEEE 30-bus benchmark system. The experimental results illustrated that the accuracy of the voltage sag state estimation results was over 99.83%. Furthermore, a comparison among different models, including the bidirectional GRU-based model, one-way WaveNet-based model, and bidirectional WaveNet-based model, was also conducted. The results illustrated that the proposed bidirectional WaveNet-based model achieved the highest accuracy and quickest convergence speed.

The switch open-circuit fault signal of the Vienna rectifier possesses non-stationary characteristics and is also vulnerable to external interference factors, such as sensor noise and load variation. This phenomenon reduces the performance of traditional methods, including model-based and signal-based algorithms. In order to improve the accuracy, convergence rate, and robustness of diagnosis models, a hybrid deep learning Transformer–BiTCN optimized via ISGA (Improved Snow Geese Algorithm, ISGA) is proposed in this paper. Firstly, to assess the Vienna rectifier’s open-circuit fault signal, the time-varying and non-stationary characteristics generation mechanism is analyzed. Then, combining the fault signal characteristics of the Vienna rectifier, the hybrid deep learning model using Transformer–BiTCN, along with multi-scale feature fusion, is presented to extract hierarchical features, including both global temporal dependencies and local characteristics to enhance fault diagnosis accuracy and model robustness. Finally, the ISGA optimization algorithm with the Bloch initialization strategy and the Rime search mechanism is further presented to optimize the hyperparameters of the Transformer–BiTCN model so as to improve convergence and improve accuracy. Finally, the effectiveness of our proposed method is tested by simulations and experiments. It has been verified that the Transformer–BiTCN along with ISGA optimization is robust to non-stationary open-circuit fault signals and can achieve high diagnosis accuracy with a fast convergence rate.