Jing Hong

Jing Hong, Yue Li, Yiran Xu et al.

In Energy Management System (EMS) applications and many other off-line planning and study tools, one-line diagram (OLND) of the whole system and stations is a straightforward view for planners and operators to design, monitor, analyze, and control the power system. Large-scale power system OLND is usually manually developed and maintained. The work is tedious, time-consuming and ease to make mistake. Meanwhile, the manually created diagrams are hard to be shared among the on-line and off-line systems. To save the time and efforts to draw and maintain OLNDs, and provide the capability to share the OLNDs, a tool to automatically develop substation based upon Common Information Model (CIM) standard is needed. Currently, there is no standard rule to draw the substation OLND. Besides, the substation layouts can be altered from the typical formats in textbooks based on factors of economy, efficiency, engineering practice, etc. This paper presents a tool on substation OLND automatic generation and visualization. This tool takes the substation CIM/E model as input, then automatically computes the coordinates of all components and generates the substation OLND based on its components attributes and connectivity relations. Evaluation of the proposed approach is presented using a real provincial power system. Over 95\% of substation OLNDs are decently presented and the rest are corner cases, needing extra effort to do specific reconfiguration.

Baoyu Jing, Yansen Wang, Guoxin Sui et al.

In recent years, Contrastive Learning (CL) has become a predominant representation learning paradigm for time series. Most existing methods manually build specific CL Strategies (CLS) by human heuristics for certain datasets and tasks. However, manually developing CLS usually requires excessive prior knowledge about the data, and massive experiments to determine the detailed CL configurations. In this paper, we present an Automated Machine Learning (AutoML) practice at Microsoft, which automatically learns CLS for time series datasets and tasks, namely Automated Contrastive Learning (AutoCL). We first construct a principled search space of size over $3\times10^{12}$, covering data augmentation, embedding transformation, contrastive pair construction, and contrastive losses. Further, we introduce an efficient reinforcement learning algorithm, which optimizes CLS from the performance on the validation tasks, to obtain effective CLS within the space. Experimental results on various real-world datasets demonstrate that AutoCL could automatically find the suitable CLS for the given dataset and task. From the candidate CLS found by AutoCL on several public datasets/tasks, we compose a transferable Generally Good Strategy (GGS), which has a strong performance for other datasets. We also provide empirical analysis as a guide for the future design of CLS.