Department of Electrical Engineering, Center for Big Data and Artificial Intelligence, State Energy Smart Grid Research and Development Center, Shanghai Jiaotong University, Shanghai 200240, China
Auto-generated profile, based on publicly available information
Fan Yang
Department of Electrical Engineering, Center for Big Data and Artificial Intelligence, State Energy Smart Grid Research and Development Center, Shanghai Jiaotong University, Shanghai 200240, China
Real-time situational awareness and event analysis are crucial to the security of the modern power grid, which is a complicated nonlinear system and hard to be completely modeled. Massive data is collected but the information hasn't been sufficiently leveraged. To effectively extract the event features, this paper proposes a framework for event detection, localization, and classification in power grids based on semi-supervised learning. Specifically, event detection is realized by invertible neural network (INN), hence to learn complex distributions of real-world measurements in a flexible way. Abundant normal measurements are learned by INN and explicit log-likelihoods then serve as the indicator to distinguish events with adequate sensitivity. Moreover, risks induced by events are assessed and spatial locations are determined. Since the majority of power system events are recorded without labels in practice, a pseudo label (PL) technique is leveraged to classify events with limited labels. The PL-based approach has an enhanced separating capability for events and outperforms other approaches under a low labeling rate. Case studies with simulated data in the IEEE 39-bus system and real-world measurements verify the effectiveness of the proposed framework.
14 May 2021 in IEEE Transactions on Power Delivery
Numerical modeling of the high impedance arc fault (HIAF) is essential to supplementing the deficiency of real-world fault cases and extending the fault scenarios. Setting parameters of the model correctly is the basis to reproduce practical fault waveforms. According to the literature, the parameters of HIAF models are mostly determined manually and subjectively. The existing automatic methods of parameter determination are mostly used in the modeling of other types of arcs (such as DC arcs and circuit breaker arcs), which do not consider the connections with power networks. It will be time-consuming if directly applying these methods to the HIAF model with the connection of a power network. In this paper, after introducing the unique requirements in simulating the nonlinearity of HIAFs, a high-efficiency parameter determination method is proposed for HIAF models. The method is realized by interfacing the optimizer programmed in Python with the HIAF model and power network established in PSCAD. In this process, a combination of pulse controllers is designed and certain configurations are proposed, so that the parameter determination procedure can be significantly simplified. Case studies show that, by combining the proposed high-efficiency method with our previously proposed DIST-C HIAF model, fitting accuracy and efficiency can be guaranteed at the same time. The time for parameter determination of a HIAF model is only within minutes, decreasing hundreds of times and independent of the length of desired waveforms.
11 November 2020 in IEEE Transactions on Smart Grid
Real-time detection and analysis of cascading events are crucial to avoiding large blackouts in power systems. Based on spectral distribution analysis (SDA) of online monitoring data, this paper proposes an approach for spatial-temporal localization of cascading events in a modern grid. The anomaly detection problem is built upon hypothesis testing, where a test statistic is designed in advance. The empirical spectral distribution (ESD) of the test statistic is compared with its theoretical counterpart, i.e., the asymptotic spectral distribution (ASD) obtained by Brown Measure. The proposed approach is sensitive and capable of temporally locating the occurrence time of each subevent (including severe disturbances, failures, and some typical physical attacks) in a cascading event. Simultaneously, the spatial information of each subevent is given. It is experimentally justified that the proposed approach is robust to normal fluctuations, oscillations, and bad data. Besides, it can be applied in both large-scale and small-scale systems utilizing the tensor product method. Case studies with simulated data in an ACTIVSg500 System and real-world online monitoring data verify the effectiveness of the proposed approach.
Multiple event detection and analysis in real time is a challenge for a modern grid as its features are usually non-identifiable. This paper, based on high-dimensional factor models, proposes a data-driven approach to gain insight into the constituent components of a multiple event via the high-resolution phasor measurement unit (PMU) data, such that proper actions can be taken before any sporadic fault escalates to cascading blackouts. Under the framework of random matrix theory, the proposed approach maps the raw data into a high-dimensional space with two parts: (1) factors (spikes, mapping faults); (2) residuals (a bulk, mapping white/non-Gaussian noises or normal fluctuations). As for the factors, we employ their number as a spatial indicator to estimate the number of constituent components in a multiple event. Simultaneously, the autoregressive rate of the noises is utilized to measure the variation of the temporal correlation of the residuals for tracking the system movement. Taking the spatial-temporal correlation into account, this approach allows for detection, decomposition and temporal localization of multiple events. Case studies based on simulated data and real 34-PMU data verify the effectiveness of the proposed approach.
Aiming at the coupling relationship between the pressure regulation of thermal and gas systems and the reactive power balance of power grid in electric-heat-gas multi-energy systems, a reactive power optimization model for multi-energy systems is proposed based on the reactive power demand characteristics in the energy transport process of thermo-gas systems. This paper studies the reactive power and output characteristics and their uncertainty of various power sources, and the reactive power demand characteristics of various energy transport processes, and establishes the reactive power balance and its uncertainty model of distribution network for a multi-energy system consisting of electricity, heat and gas with large-scale renewable energy supply. On this basis, the reactive power adjustable capacity quantization model based on the pressure adjustable limit parameters of heat, cold, hydrogen and gas energy medium transport pipeline and the corresponding reactive power demand capacity is studied, and the reactive power demand model under the coupling relationship between power grid and gas network is established to reduce voltage fluctuation. Establish the economic cost model of reactive power compensation device, reactive power supply and pump station; The uncertainty of reactive power balance and the constraint of source network load storage operation are further considered. In order to minimize the voltage fluctuation and the lowest cost, the reactive voltage control method of multi-time scale coordination and optimization is adopted to realize the stable reactive voltage control of distribution network through the reactive power optimization control of multi-energy regulation.