This paper proposes a multi-agent deep reinforcement learning-based approach for distribution system voltage regulation with a high level of PV penetration. The designed agents can learn the coordinated control strategies using only local information from historical data through the counter-training of local policy networks and centric critic networks. The learned strategies allow us to deal with PV generation uncertainty and perform online coordinated control. Comparative results with other methods show the enhanced control capability of the proposed method under various conditions.

Effective detection of fault in rolling bearings with a limited amount of data is essential for the safe operation of electric machines. This study proposes a novel meta-learning-enabled method for the detection of fault in rolling bearings of electric machines under varying working conditions with limited data. The fault diagnosis under various working conditions is cast as a few-shot classification problem, which is solved using a model-agnostic meta-learning-based model. Specifically, a meta-learner is first trained using a series of interrelated fault-diagnosis tasks under various working conditions. During this stage, the gradient-by-gradient rule is utilized for parameter optimization to achieve an effective representation of these tasks. Then, the parameters of the meta-learner are refined on a new task. This technique can achieve fast adaptation to new tasks by utilizing only few-shot samples. The proposed method can obtain high fault-detection accuracy under various working conditions when only a limited amount of data is available. Comparative tests among various methods were carried out on the Case Western Reserve University (CWRU) Bearing Dataset and the Paderborn University Rolling Bearing Dataset. The results show that the proposed model performs better than other state-of-the-art methods under various working conditions; our method has stronger generalization ability and faster adaptation ability. The fault diagnosis accuracy for both datasets was at least 99%, which proves that the proposed strategy can be flexibly applied to various scenarios.

The anomalous measurements and inaccurate distribution system physical models cause huge challenges for distribution system optimization. This paper proposes a robust voltage control method that can deal with them by systematically integrating a representation network, the deep reinforcement learning (DRL) method, and the surrogate model. The partial observation of the distribution network is first represented as a graph with tree topology that is processed by a physics-informed global graph attention network (GGAT) and a deep auto-encoder (DAE) to achieve informative and robust representation of the real-time and pseudo-measurements. The extracted features are then fed into the soft actor-critic algorithm, during the training of which a graphical-learning-based power flow surrogate model is developed to provide a reward signal for the DRL algorithm. This allows the proposed method to reduce reliance on accurate distribution system parameters. The embedding of the structural information by the GGAT and the informative features extracted by the DAE further enhances the robustness of the proposed method against anomalous measurements. The proposed method is validated using IEEE 33-node and 119-node systems. Simulation results show the robustness of the proposed method against anomalous measurements.

To improve the conversion efficiency of the dual-active-bridge converter, this article demonstrates a variable-frequency triple-phase-shift (TPS) control strategy with the help of the deep reinforcement learning method. More specifically, the twin delayed deep deterministic policy gradient (TD3) algorithm is adopted to train the agent offline with the aim of minimum power losses, under the TPS modulation with varying switching frequency. Moreover, the zero-voltage-switching performance has been considered during the training of the TD3 algorithm. Based on these, the trained TD3 agent acts as a fast surrogate predictor, which can produce appropriate control strategies in real-time for whole continuous operating conditions with soft switching and maximum conversion efficiency. The effectiveness and correctness of the proposed scheme is validated through the experimental results in a laboratory prototype.

This article proposes a robust topology change-aware distribution system state estimation (DSSE) method based on a physics-informed graph neural network and Bayesian Probability Weighted Averaging (BPWA). A general state estimator is first built utilizing a graph attention network to learn the nonlinear mapping functions under different distribution network topologies. During this stage, the topology information is embedded in the neural network and the attention mechanism is employed to capture collaborative signals and discriminate the importance of neighboring buses. Then, the BPWA method allows assigning proper weights for the state estimation results under different topologies, which finally yields a single consensus solution via the sparse training samples under the new topology. The physics-informed mechanism enables the proposed method to embed the topology knowledge in the neural network while fully exploiting the value of historical data. Robustness to anomalous measurements is achieved through the embedding of physics knowledge. The application of the BPWA method further allows the proposed method to achieve faster adaptation to topology change and quantification of the estimation uncertainties by measurement errors. MATLAB and Python are used to carry out the comparative tests to evaluate the performance of the proposed method.