YDYaping Deng

Yaping Deng

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Yaping Deng has a total of 12 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • A Self-Adaptive Voltage Sag Position Tracing Method: Deep Transfer Learning Under Changed Scene

    by Yaping Deng, Xinghua Liu, Gaoxi Xiao, Huaicheng Yan, Yan Xu, Peng Wang

    For voltage sag position tracing (VSPT) through deep learning methods, model performance deteriorates rapidly under changed scenes. Moreover, time and effort are wasted in retraining numerous models for all different scenes. Therefore, a self-adaptive VSPT method which can response to changed scene is urgently needed. In this article, a deep transfer learning for self-adaptive VSPT under changed scenes is proposed. For accurate VSPT under original scene, a deep learning method via temporal iTransformer is presented, which can enhance local feature extraction capability while retaining the iTransformer’s global perspective. For self-adaptive VSPT under changed scenes, a deep transfer learning based on feature-decoupling is further presented. Here, domain invariant features are calculated via feature-decoupling module, and the difference between source domain features and target domain features is adaptively minimized via feature transference. We test the proposed method via simulation and experimental platform, verifying that the proposed deep transfer learning has satisfactory domain adaptability for self-adaptive VSPT under changed scenes.

  • A Deep Learning Methodology Based on Bidirectional Gated Recurrent Unit for Wind Power Prediction

    by Yaping Deng, Hao Jia, Pengcheng Li, Xiangqian Tong, Xiaodong Qiu, Feng Li

    Wind power predication is very important for dispatching of power grid and the optimal operation of wind power system. However, the inherent uncertainty of wind makes the power fluctuating, leading to the wind power prediction more difficult. In this paper, a deep learning framework based on bidirectional gated recurrent unit for wind power prediction is proposed to improve the accuracy by making full use of the information provided by multiple data sources of numerical weather forecast. The feasibility of deep learning algorithm to deal with wind power prediction is firstly analyzed, and then, the model based on gated recurrent unit is constructed which can describe the relationship between input information and output information through a gating mechanism. Therefore, it can automatically establish the simplified relationship between wind speed, wind direction and wind power. Then, on the above basis, the model can further describe more complicated and difficult to be quantified from the explicit relationship which is simplified and easy to be quantified to ensure the accuracy of prediction results. Finally, the practical data collected from a wind farm in China has been used to verify the conclusions. Results show that the proposed methodology can predicate the wind power accurately


  • A Sequence-to-Sequence Deep Learning Architecture Based on Bidirectional GRU for Type Recognition and Time Location of Combined Power Quality Disturbance

    by Yaping Deng, Lu Wang, Hao Jia, Xiangqian Tong, Feng Li

    In this paper, a sequence-to-sequence deep learning architecture based on bidirectional Gated Recurrent Unit (Bi-GRU) for type recognition and time location of combined power quality disturbance is proposed. Specially, the proposed methodology can determine the type of each element in input sequence, which is different from existing sequence-to-sequence model employing encoder-decoder network. Firstly, the input sequence is normalized and batched. Secondly, deep features are extracted from input sequence by constructing Bi-GRU recurrent neural network, where multiple Bi-GRU layers are stacked together in both forward direction and backward direction. Thirdly, according to above extracted features, fully connected layer and Softmax are employed to calculate the corresponding probability indicating the category that each element in input sequence is classified to. Fourthly, Argmax or Top_K operation is further integrated to determine the type of each element in input sequence by selecting the maximal probability. Lastly, the type is recognized, and meanwhile, starting-ending times of disturbances are also located just at the moment when the type is changed. The proposed model is further validated and tested by synthetic signals and practical field signals, respectively. Experimental results demonstrate that the accuracy of type recognition is over 98% for 96 kinds of disturbances including single and combined disturbances with signal-to-noise ration being 20dB. Besides, the stating-ending times are also located with the absolute error less than 6 sampling points when sampling frequency is 256 points per cycle with noisy environment.