Search

This work presents a new algorithm for detecting and classifying data anomalies in operational measurements using statistical, clustering, and outlier-based approaches. Base detectors explored in this work includes density-based spatial clustering of applications with noise, K-Means, local outlier factor, feature bagging, and robust random cut forests using real distribution system datasets. An ensemble approach is proposed to achieve high detection accuracy and precision compared with any of the base detector and with less dependency on hyperparameter tuning. Also, developed ensemble architecture can integrate additional base detectors. In addition, a simplistic anomaly classification approach is developed, utilizing the clustering concept, while considering the physics of the power distribution systems. The developed schemes are rigorously tested and validated using data from multiple distribution phasor measurement unit devices in the Bronzeville community microgrid, with a diverse set of events and distributed energy resources at dispersed locations. Performance analysis using three test cases are provided to showcase superiority of the proposed approaches.

In contrast with traditional utility monitoring and operational tools, advanced distribution management systems (ADMS) provide modern operational features to monitor, secure and operate the distribution system in an integrated manner. Large-scale adoption of ADMS at utilities is in the early stages since advanced applications within ADMS environments are still evolving. The topology estimation module is one of the important and challenging ADMS applications with enhanced automation. This article integrates the optimization-based topology estimation module with the load/photovoltaics (PV) estimation module. Validation of the feeder connectivity information is the major challenge faced during the integration process. The statistical analysis of the error of the load/PV estimation module shows that the error does not follow the Gaussian distribution. To accurately capture the uncertainty in the error of the load/PV estimation module, the error is modeled with the Gaussian mixture model (GMM). The study using the actual utility feeder data and the ADMS platform shows that in an integrated operation of the load/PV estimation module and topology estimation module, the accuracy of the topology estimation module is improved when the estimation error of the load/PV estimation module is modeled with GMM. The methodology outlined in this article serves as the proof-of-concept and framework for integration and validation of advanced applications such as load/PV estimation and topology estimation modules in the ADMS environment.

Situational awareness (SA) is critical to properly operating active power distribution systems during normal and outage conditions. Appropriate SA tools should provide an accurate estimate of system voltage and current variables, the operational network topology, and the power injections from distributed energy sources (DERs), including behind-the-meter (BTM) photovoltaics (PVs). Obtaining an accurate SA, especially by estimating network topology and gross load demand, is increasingly challenging with the proliferation of BTM DERs with intermittent power generation. Moreover, the SA required for distribution system restoration is even more challenging after a medium to a prolonged outage due to inaccurate estimates of cold load pickup (CLPU) and switch statuses. This paper proposes an integrated real-time model update (RTMU) module for SA enhancement to help distribution system operators (DSOs) understand the power system conditions in dynamic and DER-rich environments. The proposed RTMU consists of several modules to obtain the required level of SA for operational decision-making. It includes estimators for a) BTM PV power, b) network topology, and c) CLPU. The proposed approaches leverage multiple data resources, deep learning approaches, and domain knowledge of the power system to provide the required level of SA to DSOs. The dependencies among these modules are actively leveraged to enhance SA under normal conditions and during power outages. We demonstrate and analyze the performance of the proposed RTMU on a modified IEEE 123-node feeder and a utility distribution system from the Western United States.

Event detection, classification, and localization (ED-C-L) in an active distribution network are challenging given system dynamics but critical for taking the control actions to mitigate the impact on the system operation. In this work, a new set of algorithms are developed for ED-C-L including, a) prony based dynamic window and event detection, b) data-driven system identification (DD-SI) using Koopman Mode analysis (KMA), physics-rule, and weighted-vote based assembling technique for event classification, and c) weighted graph and KMA based localization using distribution phasor measurement unit (D-PMU) data. DD-SI is utilized to construct a transient energy matrix (TEM) and capture the system dynamics using the temporal and spatial signature of events without using a labeled dataset. The proposed technique considers system dynamics for ED-C-L to achieve enhanced accuracy, compared to most of the existing data-driven techniques. A lab-scale real-time testbed is developed to generate D-PMU data utilizing a 24 bus distribution system simulated in the OPAL-RT/Hypersim for evaluating the performance of the proposed algorithm, and measured using selected metrics. Case scenarios consider utility-scale solar photovoltaic (PV) and battery energy storage system (BESS) integration at dispersed locations to validate proposed algorithms under dynamically changing operational scenarios. Further, the developed approach is applied on DPMU data from real distribution systems with DERs to showcase the superiority of the proposed approach to field systems.

Integrating sustainable energy resources transforms the distribution grid into an active system with higher variations observed in load and generation. Estimating distributed generation, gross load, and cold load pick-up (CLPU) become more challenging with behind-the-meter (BTM) distributed energy resources (DERs), especially in case of outages caused by extreme events. This work proposes a resilience-driven restoration scheme using the most updated information from an integrated and enhanced situational awareness tool (ESAT) using kernelized Bayesian state-space inference (KBSI) with Markov Chain Monte Carlo (MCMC) and multiple optimization algorithms. ESAT consists of the BTM load/ DER estimation and disaggregation, CLPU estimation, and network topology estimation with de-energized islands. The proposed work provides solutions to establish informed restoration schemes considering resilience criteria for a quick recovery of high-priority loads. A resilience metric is utilized after outages to measure the effectiveness of ESAT-driven restoration for the supposed threats. The performance of developed ESAT is demonstrated using actual field datasets and validated using the emulated real scenarios on a benchmark model.