Final Scientific/Technical Report Advanced Intelligent Sensor Development and Demonstration for Future Distribution Systems with High Penetration DERs WORK PERFORMED UNDER AGREEMENT DE-OE0000873 University of Texas at Austin 10100 Burnet Rd., Bldg 133 Austin, TX, 78758 Award Period of Performance: 10/01/2017 to 7/31/2025 Submitted: October 2025 PRINCIPAL INVESTIGATOR Robert Hebner TEAM MEMBERS University of Texas at Austin NI Verivolt Eaton Wood County Electric Cooperative Guadalupe Valley Electric Cooperative Capstan Technologies Argonne National Laboratory SPONSORING PROGRAM OFFICE U. S. Department of Energy Office of Electricity Delivery and Energy Reliability Via the National Energy Technology Laboratory TABLE OF CONTENTS LIST OF TABLES ........................................................................................................... 4 LIST OF FIGURES.......................................................................................................... 5 LIST OF ACRONYMS AND ABBREVIATIONS ............................................................. 6 I. Executive Summary ............................................................................................... 7 II. Objectives ............................................................................................................... 8 III. Technical Approach ............................................................................................... 9 IV. Accomplishments and Conclusions .................................................................. 33 APPENDIX A: Product or Technology Production ................................................... 34 Acknowledgement and Disclaimer Acknowledgment: "This material is based upon work supported by the Department of Energy under Award Number DE-OE0000873." Disclaimer: "This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof." 3 LIST OF TABLES Table 1. Event summary………………………………………………………………………28 4 LIST OF FIGURES Figure 1 Block diagram of the smart power line sensor………………………………………. 10 Figure 2 The arrangement of the primary components of the sensor package when mounted on a power line………………………………………………………………………………10 Figure 3 Sensor dimensions…………………………………………………………………………..11 Figure 4 Internal view of the sensor package showing the arrangement of the key components……………………………………………………………………………………………….11 Figure 5 Installation of a sensor on a 12 kV powerline………………………………………….12 Figure 6 Basic concept of the power harvester………………..………………………………..13 Figure 7 Weight limit from the point of view of the powerline cable………………………..14 Figure 8 Left: Core’s cross section & weight per power. Right: Power threshold for a given core………………………………………………………………,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,…….15 Figure 9 Left: FE simulations (Altair Flux). Right: Physical modeling using MATLAB/Simscape simulations…………………………………….………………………………..15 Figure 10 Fault waveforms captured by the DAQ system and SEL relay….….……………17 Figure 11 A sub-cycle transient event with a 3-phase voltage step change….…………..18 Figure 12 Traveling wave fault location method……………………………………….…………20 Figure 13 Simulation circuit diagram and key parameters……………………,……………..20 Figure 14 Line-to-ground voltages at substation and feeder end……………….………….22 Figure 15 Sensor system CHIL test environment………………………………….……………25 Figure 16 CEM test facility electrical schematic…………………………………….…………27 Figure 17 Grounded pneumatic lift…………………………………………………….…………..27 Figure 18 Three examples of disturbances detected by the broadband sensor that were not detected by the legacy sensor system…………………………………….……29 Figure 19 Sensor A, 07:02:10 am. The circled sections show the initial lightning Strike…………………………………………………………………………………………………….…31 Figure 20 Sensor B 07:02:10 am……………..…………………………………………………….32 5 LIST OF ACRONYMS AND ABBREVIATIONS A – Ampere ac – Alternating current ACSR – Aluminum conductor steel reinforced C - Celsius CHIL – Control hardware in the loop CPU – Central processing unit DAQ – Data acquisition DMA – Direct memory access DOE – Department of Energy DRAM – Dynamic random-access memory FFT – Fast Fourier transform FIFO – First in, first out FPGA – Field programmable gate array Ft - Feet GB - Gigabit GDP – Gross domestic product GPS – Global positioning system GVEC – Guadelupe Valley Electric Cooperative H - Hours HiZ – High impedance Hz – Hertz, a unit of frequency kA - Kiloamp kHz – Kilohertz km – Kilometers kV - Kilovolt kVar – Kilovolts reactive m – Meters mm - Millimeters mH - Millihenry Mhz - Megahertz ms – Millisecond ns – Nanosecond pC - Picocoulomb ps – Picoseconds rms – Root mean square SEL - Schweitzer Engineering Laboratories STD - Standard U.S. - United States UT – University of Texas at Austin V – Volt W - Watt WCEC – Wood County Electric Cooperative 6 I. Executive Summary The nation’s electric grid is evolving to supply the electricity needed for economic growth, health, and quality of life. As the DOE has pointed out “Old Tools Won’t Solve New Problems” This project conceived, designed, built, and demonstrated a new tool that promises to make the emerging grid dependable at a lower cost. The tool is a smart sensor with the sensitivity and the bandwidth to measure actionable information about grid health that was unmeasurable. The traditional approach has been to combine routine preventative maintenance with sensors that can promptly turn off the appropriate parts of the grid as failure occurs. This innovation is a bridge technology. It measures the same quantities as the traditional sensors and much more. Due to its significantly improved bandwidth and sensitivity, it provides enhanced capability for fault identification and location, i.e., what went wrong and where. It also can detect and identify information that the legacy sensors cannot, including incipient fault information that enhances distribution grid reliability. The sensor integrates a computer, electric measurement instruments, and a communication system into a package that clamps on to an energized power line in the grid. The package can be installed by a lineman in less than ten minutes and it has the necessary components to scavenge the power it needs from the line on which it is mounted. This advance is also unique in that the current practice is to have a specialized sensor for each event that is to be measured. This is a sentinel observing a wide range of information on the grid and alerting the operators if action may be needed. This function is possible because of the small, rugged computing capability that is available today. The system has also demonstrated capabilities that were previously unavailable. Preliminary indications are that these include: • Detecting the moment of break of an energized power line. • Detecting the interaction between an energized line and vegetation. Recording the differences, I that interaction prior to any smoke being produced, after smoke but before flames and after flames are ignited. • Cracked or polluted insulators. More unique examples are likely, but require more study for definitive identification. 7 II. Objectives The objective is to advance technology which enhances the reliability of the electrical distribution grid via a smart sensor. The cost of outages is a well-studied topic. For example, a one-day outage in the Chicago area1 was estimated to reduce the three-month GDP by $1.8 billion. Similarly, outages in Texas are estimated2 to cost the economy almost $36 million for every megawatt hour of unserved load. So, improving reliability provides a financial return as well as enhancing quality of life. A technical challenge to improving reliability is the high level that has been achieved3. The U.S. Department of Energy’s (DOE) report baseline is that reliability is achieved if the total amount of energy that cannot be supplied to customers does not exceed 0.002% of the total energy demanded each year. The approach that enabled meeting today’s reliability goals has been to employ sensors in substations that detected dangerous events and removed power in high-risk situations. As was emphasized in the DOE report in footnote 3, “Old Tools Won’t Solve New Problems”. Fortunately, technology is making possible a novel approach to sensing the grid that promises to be a key tool for increasing grid reliability. Progress is possible because laboratory research shows that precursors to serious failures are present on the grid, but at too high a frequency and too low of an amplitude to be detected with legacy substation instrumentation. These precursors are high impedance (HiZ) faults. HiZ fault detection has been the subject of extensive research, with hundreds of published papers proposing various methods. Most of these approaches focus on signal processing techniques applied to the steady-state behavior of HiZ faults. Although these methods show promise, none have become universally accepted or standardized methodology for HiZ fault detection. The field remains a topic of investigation due to the risks posed by HiZ faults, which can lead to fires or evolve into short circuits that cause system outages. The sensor developed in this project can serve as a platform for the various artificial intelligence, machine learning and signal processing techniques that have been proposed by researchers around the world. This project accelerates the adoption of intelligent powerline sensors for fault identification, fault location, and islanding detection. The focus is on distribution systems serving larger geographic areas and accommodating distributed energy resources. The project designed, 1 Wing, I.S., Larsen, P.H., Carvallo, J.P. et al. A Method to estimate the economy-wide consequences of widespread, long duration electric power interruptions. Nat Commun 16, 3335 (2025). 2/https://interchange.puc.texas.gov/Documents/55837_12_1421762.PDF 3https://www.energy.gov/sites/default/files/2025- 07/DOE%20Final%20EO%20Report%20%28FINAL%20JULY%207%29_0.pdf 8 built, and installed first-of-a-kind sensors operating on the powerlines of electric power distribution systems. It exploits modern computing and communications for initial data analysis at the grid edge before sending information to a control room for further analysis and/or action. By identifying and locating high impedance faults early, widespread outages, including those leading to fires, can be mitigated. The sensor is designed to both enhance routine reliability and reduce catastrophic failures by providing unprecedented situational awareness. This measurement capability also promises to recognize incipient fault behavior to enhance scheduled maintenance. The system augments routine maintenance such as tree trimming and equipment inspection. Failure is a weak link, not an average event. Consequently, failure does not occur when the average tree limb grows back or the average piece of equipment ages. The fastest growing and the fastest aging risks are the most important. Sensors can detect these risks. The sensor system is installed using a bucket truck. It is self-powered from the line on which it is mounted. The naturally occurring signals on the powerline provide information on the state of the distribution system. These signals are monitored using voltage and current sensors. The sensor data is examined for any relevant anomalies. Anomalous behavior is saved for immediate analysis in the sensor system and/or for transmission to a central computer system. A user interface was developed for the central computer to facilitate situational awareness. The development was coordinated by researchers at the University of Texas at Austin, but success would not have been achieved without key contributions from Verivolt, LLC, NI, Capstan Technologies, Inc., Wood County Electrical Cooperative, Guadalupe Valley Electric Cooperative, and Argonne National Laboratory. III. Technical Approach III.1 Sensor construction. The initial activity was to design and evaluate an appropriate sensor. A functional block diagram of the sensor is shown in Figure 1. 9 Figure 1 Block diagram of the smart power line sensor The physical arrangement of the sensor mounted on a power line is shown in Figure 2, and the dimensions of the components are indicated in Figure 3. Figure 2 The arrangement of the primary components of the sensor package when mounted on a power line. 10 Figure 3 Sensor dimensions Figure 4 shows an internal view of the sensor package to indicate the arrangement of the key components. All components are designed so that the sensor can be clamped around an operating power line. This permits installation or removal with no power interruption. Figure 4 Internal view of the sensor package showing the arrangement of the key components Figure 5 shows the installation of a sensor on an energized 12 kV power line. Once the bucket trucks are in place, the installation requires five to ten minutes. The initial concept envisioned installation using a hot stick rather than a bucket truck. Discussion with cooperatives, however, highlighted a clear preference for installation using bucket trucks for greater personnel safety. 11 Figure 5 Installation of a sensor on a 12 kV powerline III.1.a Power harvester The power harvester, the sensor’s primary power source, extracts power from the power line through magnetic induction. Batteries provide temporary backup power in cases of low current or power outage. Our primary design goals were to provide sufficient power at lower currents, minimize noise & vibrations, and achieve adequate environmental protection while providing adequate power for reliable operation. The design is modular because the distribution circuit current is smaller at locations further from a substation. Consequently, a smaller transformer assembly is required to placements nearer a substation III.1.a.1 Principle of operation The power harvester operation is based on well-known transformer technology. Specifically, an ac current flows in the power line that produces an alternating magnetic flux. This flux in the ferromagnetic core induces an alternating voltage in the coil output terminals. This voltage is rectified to power the sensor. Figure 6 summarizes this concept The power harvester is a current transformer with two parasitic air gaps resulting from the need to clamp the sensor to an operating power line. 12 Figure 6 Basic concept of the power harvester III.1.a.2 Requirements The powerline-current range in an operating distribution system can vary, e.g., between 1-50 A rms. It was decided that designing for operation at locations where the current is typically at least 15 A rms provides a good compromise among utility, size, weight, and cost. If lower current operation is needed, more modules can be added. The minimum load power at the battery charging threshold is 7 W. The sensors are designed for continuous operation in all weather conditions and weeks of operation in a distribution grid have been demonstrated. This is a design constraint. The modules were designed so that the number needed in practical applications would not exceed these limits Mechanical constraints include a max weight of 7 kg and length of 400 mm. The sensor must clamp on to both 1/0 ACSR (0.4”) or #4 ACSR (0.26”) power cables to accommodate the diversity in the field. The system must be capable of being installed and by a single lineman wearing insulated gloves using a bucket truck. The installation should require less than ten minutes. Additionally, the sensor is designed to operate over an ambient temperature range of -10C to +40C while accommodating the higher internal temperature due to solar radiation. It should operate at relative humidity levels up to 90%. The weight of the entire sensor is limited by the breaking strength of the cable. The National Electric Safety Code sets this limit at 60% of the cable’s rated breaking strength. Ergonomic considerations. Figure 7 summarized weight constraints. 13 Figure 7 Weight limit from the point of view of the powerline cable III.1.a.3 Electromagnetic analysis The harvester design parameters include inner and outer radius, core length, magnetic material, and the number of turns in the coil. Preliminary analytical assessment provided important guidance. One is that the while the core cross section should be maximized, its inner diameter is limited by the wire diameter and the maximum outer dimeter is a compromise because the magnetic field produced by the current in the powerline decreases as the reciprocal of the distance from the wire but the weight increases as the square of that distance. In addition, the core material should have high permeability, high saturation, and be corrosion resistant. These requirements suggest a grain-oriented core material, which dictates a tape wound core. The analysis also highlighted that there is an optimum number of turns for which the power is maximized. I also showed the significant negative influence of the two air gaps. That influence is minimized by excellent surface quality and good tightening of the two halves. The derivation of the power threshold for a given core is based on power flow analysis. The analysis also provided guidance on the core's cross-sectional area and weight per unit of power. Figure 8 summarizes the analysis. The left side of the figure shows the nonlinear relationship between the line current and the required core means. It highlights conductors with really low currents would require very large cores. The right side of the figure shows the load per current in our prototype. The results of the extracted power exceed the expectations of the core manufacturer. 14 Figure 8 Left: Core’s cross section & weight per power. Right: Power threshold for a given core III.1.a.4 Detailed Design The detailed design requires finite element magnetic and electrostatic modeling, as well as physical modeling. Figure 9 summarizes the analysis results. The bottom figures show that for each load value, there is an optimal number of turns in the coils that maximizes the load power. This held true in both the simulations and the tests. Figure 9: Left: FE simulations (Altair Flux). Right: Physical modeling using MATLAB/Simscape simulations 15 III.1.a.4 Fabrication and Test This, being a first-of-a-kind device, required innovation, not only in design, but also in manufacturing. The key steps were coil winding, varnishing, and assembly, with twenty cores in total for five sensors. The most significant problem was caused by quality control issues with the purchased cores A purpose-built test jig supported assessing the cores, including investigating the optimal number of turns. Success in this testing was followed by testing at UT’s nominally 12 kV overhead line test cell to verify that there is no electrical breakdown, and system level performance. The achieved power density outperformed expectations by 150-200%. A set of four cores provided 7 W power at 9 A rms powerline current. The four-core system weighed 6.5 kg. III.2 Software Applications The value of this sensor is that it can identify signals on the grid so that appropriate action can be taken. To provide documented progress, the research contract specified three applications. The field tests and demonstrations identified an additional set of early applications. III.2.a Baseline Applications The baseline applications are fault-type identification, fault location, and islanding detection. Each of these has been developed and published in the technical literature4,5,6 . IIII.2.a.1 Fault Identification Transient events in distribution systems such as animal contacts, vegetation contacts, equipment failures, and switching operations produce characteristic voltage and current waveforms with unique frequency content. These signals, captured by line-mounted sensors or substation data acquisition (DAQ) units, are typically recorded in COMTRADE format with multiple analog channels sampled at high rates. Manual analysis of such datasets is time- consuming, error-prone, and difficult to scale. To address this, the system developed in this investigation integrates field data collection with an automated MATLAB-based processing framework. Fault signatures were first recorded 4Feng X, Hebner R, Strank S. Design consideration for power line sensors in power distribution systems. In2020 IEEE Energy Conversion Congress and Exposition (ECCE) 2020 Oct 11 (pp. 456-462). IEEE. 5 Fan W, Kang N, Hebner R, Feng X. Islanding detection in rural distribution systems. Energies. 2020 Oct 20;13(20):5503. 6 Feng X, Strank SM, Paz P, Hebner RE. Enhancing fault identification in rural power distribution systems. Electric Power Systems Research. 2025 Apr 1;241:111331. 16 using the custom DAQ system and validated against legacy SEL relay records provided by the cooperating utility. The fault identification was based on utility operator outage report. Figure 10 Fault waveforms captured by the DAQ system and SEL relay. Figure 10 shows a single-phase short-circuit fault measured by both the DAQ system and the SEL relay, with consistent waveforms. For another event, Figure 11 shows the data captured during a three-phase voltage step change. The sensor revealed that similar transients only occurred during such three-phase step changes. The FFT representation in Figures 11(b) and 11(c) shows the spectrum of phase A signals, with significant energy from hundreds of hertz to a few kilohertz. This spectral content is key to distinguishing this event type from others. 17 Figure 11 A sub-cycle transient event with a 3-phase voltage step change. A set of interconnected MATLAB functions was developed to automate the end-to-end processing of COMTRADE files: • Batch organization: Group .cfg and .dat pairs into event folders. • Event isolation: Identify transient windows and split data into pre-event, event, and post- event segments. 18 • Spectral analysis: Compute FFTs of voltage and current channels to obtain frequency- domain signatures. • Fault classification: Match new events to a fault signature library using a least-squares comparison of spectra. • Export and archiving: Generate Excel workbooks, time-domain/FFT plots, and structured result folders for reproducibility. The framework includes four key components: I. process_comtrade_files.m (driver script), II. classify_fault_fft.m (FFT-based classifier), III. export_comtrade_to_excel.m (Excel exporter), and IV. build_fft_library.m (library builder). This pipeline ensures that new fault events can be automatically categorized and archived for future analysis. With utility input and nine months of field deployment, seven distinct event types were identified. Each exhibits unique characteristics in both the time and frequency domains, enabling reliable separation via FFT-based classification. Type 1 - Three-Phase Voltage Step Change Rapid RMS voltage shifts (100–200 V) across all phases, occurring a few times per week. These events are often linked to voltage regulation and do not necessarily indicate system failures. Transients are modest, with FFT showing energy up to a few kilohertz (Figures 11(b) and 11(c)). Type 2 - Spiking (Rain/Lightning-Induced Transients) High-frequency bursts caused by lightning or partial discharges during precipitation. Most are three-phase events, with spectral energy above 1 kHz. Typically, clear within 1–2 AC cycles. Type 3 - One-Cycle Overcurrent Brief (1–2 cycle) high-current faults that do not trip feeder-head reclosers. Impedance remains constant during the fault. FFT reveals pronounced high-order harmonics due to low impedance. Type 4 - Sub-Cycle Transient Failure Transformer or equipment failures generating sub-cycle transients. High fault currents cause protective devices to clear these events almost immediately. Spectra show broad high- frequency content. Type 5 - Tree Touch Single-phase events caused by vegetation contact. Initially low-level currents persist for seconds before evolving into a short-circuit. The initial transient is a key early indicator. FFT reveals high-frequency content before significant current rise. To confirm the accuracy of the developed sensor, signatures measured by others using other means applied to the sensor using an Omicron CMC256 plus. The UT sensor accurately captured 19 the signals, even with current levels below 1 amp (broken conductor). Over a one-second sliding window, the error in harmonic content was less than 5%. Type 6 - Load Pick-Up (Cold Load Reconnection) Inrush currents following grid outages. The DAQ system captured both the initiating overcurrent and the subsequent load pick-up. FFT shows strong 2nd and 3rd harmonics decaying over 10–20 cycles, often with accompanying voltage dips. Type 7 - Animal Contact Short, high-current faults from animal contact. These may blow fuses at service transformers, causing localized outages. FFT spectra show strong high-order harmonic peaks in both voltage and current. Finally, the MATLAB framework produces: • A consolidated Excel workbook with all events (each on a separate sheet), • PNG plots of time-domain waveforms and FFT spectra, • A summary CSV file listing event names, types, and error metrics, and • A structured folder hierarchy for reproducibility. • The fault signature library (built with build_fft_library.m) can be iteratively updated with new labeled events, ensuring continuous improvement of classification accuracy. III.2.a.2 Fault Location The sensor not only identifies anomalous behavior but includes software to locate events. The traveling wave fault location method is explained in figure 12. Once a fault occurs, the transient signal propagates in both directions on a line. The speed of the traveling wave is estimated using the line parameters. If two sensors capture the fault waveforms, the time arrival difference of pulse signals can be used to estimate the location of the fault. This method is extended to multiple laterals and loop configuration by integrating a sensor- network with GPS synchronization. Multiple sensors with sub-microsecond level synchronized waveforms could improve the fault location accuracy. The key challenge is that the sensor needs high bandwidth. Local intelligence on sensor controller can detect A detailed transient simulation study is performed with realistic line model from utility to specify a suitable frequency bandwidth for voltage sensors. The simulation circuit is shown in figure 13. All the parameters are from an operational rural distribution feeder with overhead lines. Based on the line parameters, the wave speed is around 1.84•108 m/s. The total length of this feeder in the simulation is 15 km. The circuit was simulated in Matlab with 250 ns time step. Two faults are studied. Fault 1 is 10 km from the substation. It is a high impedance fault. Fault 2 happens in one cell of a capacitor. Fault 1 happens at 5 ms. The line-to-ground fault 20 resistance is 150 Ω. The fault is cleared at the end of the first ac cycle. Fault 2 happens at 40 ms. The fault behaviors are measured at the substation bus and the feeder end in the study. Figure 12 Traveling wave fault location method. Figure 13 Simulation circuit diagram and key parameters. Show All Figure 14(a) shows the voltage signals at substation and feeder end during the two faults. Figure 14(b) shows that the voltage signals at substation and feeder end have quite different transient behavior. This is because fault 1 is the upstream of the capacitor bank. When the fault wave passes through the capacitor, this device performs as a low pass filter reducing the high frequency components. Figure 14(c) shows the voltage waveforms during fault 2. GPS synchronization accuracy is ±100 ns. Without considering sensor bandwidth and accuracy limitations, the best accuracy for fault location using traveling wave is ±36.8 m. Sensor and data acquisition system limitations and other field noise decrease the fault location accuracy. When fault 1 happens the line current increases from 75 A to 150 A. This increase may not be sufficient to trigger an over-current relay at the substation. This type of event is hard to detect using the current signal only with a conventional current transformer, especially when the fault happens at a time of low instantaneous voltage during a cycle. For the capacitor insulation breakdown, only a small oscillation online current is detected. This difficulty arises because the large line inductance filters the high frequency content from the line current signal. Compared to underground cables, overhead lines have smaller distributed capacitance. This allows the 21 high frequency voltage signal during transient faults to travel longer distances. Thus, voltage signals are more desirable to detect high impedance and transient faults in a distribution feeder with overhead lines. To investigate the impact of voltage sensor frequency bandwidth, the cutoff frequency of voltage sensor is varied from 10 kHz to 1 MHz. As expected, as the cutoff frequency decreases, the high frequency components are attenuated. To achieve sufficient fidelity to identify transient faults, it suggests that the cutoff frequency should not be less than 100 kHz. Otherwise, signals could not be accurately captured. Fast Fourier Transforms (FFT) are effective tools to provide additional insight for the sensor frequency bandwidth selection. 22 Figure 14 Line-to-ground voltages at substation and feeder end. ow All 23 The field testing was augmented with CHIL testing. The CHIL simulation system for sensor application test is shown in figure 15. The simulated model in Opal-RT simulator is developed based on an operating circuit in a rural distribution system. Two different simulation approaches are adopted. For the islanding detection test, circuit model is simulated using electromagnetic transient simulation approach in a CPU-based simulator with 200 ps time step. In the traveling wave fault location test, FPGA-based simulator is used to model detailed traveling wave behavior during fault with less than 1 ps simulation time step. A similar circuit was implemented in the FPGA simulator to evaluate the traveling wave fault location method. The lengths of lines 1, 2, and 3 are reduced to 4 km, 1 km, and 1 km, respectively. All the other parameters are maintained the same. The reduced line length makes FPGA real-time simulation more efficient with less than 1 ps simulation time step. The controllers are compact RIO 9039 (1.91 GHz Quad-Core CPU, 2 GB DRAM, and Kintex-7 325T FPGA) and 9043 (1.30 GHz Dual-Core CPU, 2 GB DRAM, Kintex-7 160T FPGA). GPS module is available for controllers. In the lab test, the GPS antenna receives signals from 11 satellites. Two groups of analog input channels can be configured for different applications. The first group channels have 50 kHz sampling rate to support regular data acquisition or functions (such as fault event record and islanding detection) and the second group channels have 1 MHz sampling rate to support traveling wave fault location. The data acquisition system and sensor local intelligence are implemented on the sensor controller. Various basic fault detection criteria are implemented including over-current, under-voltage, and under-frequency to detect permanent or short-circuit faults. FFT algorithm is implemented on FPGA to extract high frequency content from measured signals with 1 MHz rate. The calculated frequency spectrum of a pre-selected time window of data by FPGA is transferred to the processor through direct memory access (DMA) FIFO in operational real-time to detect sub-cycle transient faults which usually include significant amount of high frequency content (1 kHz - 100 kHz). Once a fault is detected, a fault waveform including ten-cycle before and after the fault initiation is captured with 50 kHz and 1 MHz rates. The fault event record function is evaluated in the CHIL system and the monitored peak CPU consumption is less than 60% and the averaged value is less than 25%. To detect the fast pulse signal for traveling wave fault location, a second-order high-pass Butterworth filter is implemented on FPGA with 50 kHz frequency cut-off. This is an effective way to remove low frequency content in the fault signal and a constant threshold (such as 100 V) is applied as a triggering criterion. In the test, two sensors are placed at feeder head and end, respectively, with 6 km separation. A GPS module is used to synchronize fault waveforms and the captured waveforms are transferred through a communication network to a central controller for fault location identification. The tests are repeated 30 times and the estimation error is always less than 344 m. Figure 12(c) shows the captured waveform when fault 2 occurs. Thirty tests show that the fault location error is less than 416 m. 24 The error percentage of the test results is less than 7%. The length of the studied feeder. 6 km, this contributes to the error percentage. The accuracy of the traveling wave fault location method is sensitive to the line length. As the line length increases, the error is expected to decrease correspondingly. The line model in the FPGA simulator introduces errors as compared with a distributed line model. It is expected that the fault location error could be further reduced if the sampling rate is increased or a more accurate distributed line model is used. The key for field application is noise mitigation and accurate sensing at high frequency. III.2.a.3 Islanding Detection Local measurement-based islanding detection method has been implemented on the sensor controller. Local measured voltage and frequency are used to detect an islanding event. A 5% deviation from nominal value is selected as an islanding criterion, and an islanding notification is sent to the central controller once an island is detected. In the real-time simulator, the test system is a rural distribution feeder. The simulated system includes over 30 aggregated loads, over 30 aggregated line segments, and capacitor banks. When an island happens, the local generation and load mismatch is larger. Thus, large frequency and voltage deviation can be quickly detected. The total islanding detection time in 30 tests is always less than 50 ms on the local sensor controller. The event is transmitted to the central controller through a local area network in a second. 25 Figure 15 Sensor system CHIL test environment. Preliminary synchrophasor-based islanding detection is also prototyped on the intelligent sensor. Each sensor controller acquires local voltage and current signals with 50 kHz sampling rate. Standard P class synchrophasor measurement approach is prototyped using single phase voltage and current signals. All the synchrophasor computation is performed on the FPGA with an update rate of 10-60 reports per second. Each report includes synchronized time stamp, phase angle, power flow, ROCOF, etc. Each data packet is transferred to the processor through DMA FIFO. When an island occurs, the phasor on the island quickly deviates from phasors in the rest of the grid. Using the same rural test system, the synchrophasor-based approach can detect island in less than 1 second for 30 tests using a local area network. 26 III.2.b Vegetation Interaction The interaction between power lines and vegetation can and has led to major fires. This has stimulated layers of defense against such events. This sensor has shown it can be a valuable addition to that defense. It can provide earlier warning than other observation systems. Measurements in the grid of the Wood County Electric Cooperative showed that the sensor did detect the same contact that was measured by the substation equipment. Measurements taken in cooperation with Eaton, however, documented that, when a line fell to the ground at a point where the local ground had insufficient conductivity to trip breakers, the sensor detected both the break and the continuing interaction between the line and the ground. To generate supporting data, the project developed a test facility at UT. The implementation followed the design guidance outlined in IEEE Std 51-1983 “IEEE Recommended Practices for Safety in High-Voltage and High-Power Testing”. CEM’s facility is approximately 20’ by 20’. It is under a covered awning approximately 12’ high and is enclosed by a gated 8’ chain link fence. The gate to the area features an interlock switch and warning strobe. The fenced MV facility is inside a second 8’ chain link fenced yard which surrounds the rear side of CEM and has limited accessibility. For additional safety, the facility is run remotely from an operator console that is inside the building. A video camera is used to monitor the experiments An electrical circuit of the test facility is shown below in figure 16. Primary power is provided via one of CEM’s 480 V ac 3 phase buss ducts. A 400 A circuit breaker provides power to a fused remote operated contactor. Since the line transformer requires 277 V ac input, a Delta-Wye transformer was used for conversion. A single 277 V ac phase powers the low voltage winding of a single phase 167 kVA line transformer to step-up to 12.47 kV. An overhead loop connects the test line to a step-down transformer. The step-down transformer powers a load resistance, which represents the nominal load condition. A common ground reference is connected to the cases of all components. An integral feature of the facility is the grounded pneumatic lift, figure 17. The lift is used to raise different potential fault-creating objects, such as a tree branch, into the overhead MV loop. The lift is controlled remotely using air to make and break contact with the overhead line. Use of plastic pneumatic tubing insulates the operator from the test voltage. A trough is attached to the top of the lift to facilitate attaching test objects and to connect to electrical ground. 27 Figure 16 CEM test facility electrical schematic Figure 17 Grounded pneumatic lift Outside in the test facility, Instrumentation and control cables are in overhead, grounded, metal cable trays which were positioned to avoid potential damage resulting from planned arc fault events. Data acquisition and remote control are performed from an operator’s console outside of the test space. The operators’ console features a large monitor on which four video camera images are displayed and a laptop computer which provides control and data acquisition. A dSpace DS1103 prototype control hardware platform runs a real-time operating system at 10,000 samples/s. DSpace Control Desk was used to generate a graphical user interface 28 III.3 Technical performance extensions An important characteristic of operating the sensor on a power line is the amount of potential information available to a sensor with better sensitivity and bandwidth than has been used traditionally. Table I provides an example. Table 1. Event summary The data produced in November showed the system was operating as anticipated. Some of the events it detected were also detected by the legacy equipment operated by the cooperative. On December 1, based on the accumulated data, the event threshold was raised. The reduction in sensitivity was not due to the knowledge that fewer notable events were thereby ignored. A decision was made that since this new sensor data had been ignored by the utilities for decades, the limited project duration made it prudent to initially focus on larger events. The data not detected by legacy sensors is a key benefit of the sensor. Technically important data was detected by the broadband sensor but not by the legacy system. This underscores the improvement promised by the new sensor technology, particularly in enhancing situational awareness and providing guidance on preventative maintenance. The challenge is in identifying the cause of these signals that have gone undetected by the utility. Additional field testing augmented by hardware in the loop laboratory testing is needed. Laboratory testing is important because field data occurs infrequently. Figure 18 shows information with a research hypothesis as to the source of the disturbance. 29 Figure 18 Three examples of disturbances detected by the broadband sensor that were not detected by the legacy sensor system. Dataset C has a signature like what would be expected from a nearby cracked insulator. The data cannot be attributed definitively to that cause without further research. This underscores a fundamental opportunity. The novel high bandwidth requires follow up research to unambiguously identify the cause of the signals and to be able to ignore any that are benign. The sensor has also provided indication of the multiple events that can occur in a gird due to lightning. For background, lightning is a special case in power system design and analysis. Experience has shown that lightning strikes on a grid are inevitable. But the location, the size, and the distributed effect throughout the grid are less well understood. This situation has led to: 30 • Standard lightning test waveforms for grid equipment. The highest test voltage equipment received to qualify for grid installation is during a standard lightning test. It is called “standard” because it is a 1.2 x 50 µ s unipolar pulse specified in standards promulgated by the IEEE and the International Electrotechnical Commission. • Computer codes optimized for the prediction of the propagation of transients, e.g., lighting impulses through the grid. • Development of enhanced techniques, including supervised learning and machine learning to predict an imminent lightning strike in a particular location to enhance protection. A missing part in this global effort to reduce lightning-induced outages has been reliable measurements of the signals with appropriate bandwidth to record salient behavior. Data provided by this novel sensor deserves special attention because it shows promise in helping to fill this information gap. During a thunderstorm on April 20th, 2025, between 5 and 8 a.m., overvoltage, arcing, and overcurrent signatures were recorded by two sensors on the same phase, Sensor A and Sensor B. Sensor A was 1 mile from the nearest substation while Sensor B was 2.4 miles from that substation. These signatures, figures 19 and 20 captured the evolution of fault events from initial lightning-induced overvoltage to subsequent short circuits that led to multiple outages. 31 Figure 19 Sensor A, 07:02:10 am 32 Figure 20 Sensor B 07:02:10 am Within the marked circles in Figure 19, over voltages can be observed to be potentially the first strike followed by two reflections. During this time (8 cycles) a high harmonic content is observed in the current (green line) mirroring a highly nonlinear load. The waveforms suggest that an arrester actuated during the first strike, but insulation degradation followed the arrester discharge sustaining a high impedance fault that evolved to a lower impedance fault during the reflection of the overvoltage. In Figure 20, no high frequency overvoltage is observed. However, a voltage sag is observed and a cycle overcurrent. The voltage sag is potentially due to an external circuit short circuit. IV. Accomplishments and Conclusions Enhancing reliability in our emerging grid requires using the information the grid provides about its health that was ignorable in the past. Emerging sensor technology is an emerging key part needed in the future. This project showed that the technology is available to support a meaningful improvement in grid measurements. Significantly, the sensor was designed, built, and demonstrated both in laboratory environments and in grids operated by the Wood County Electric Cooperative and by the Guadalupe Valley Electric Cooperative. The laboratory assessment consisted of both conventional experiments and assessment using hardware-in-the- loop and control-hardware-in-the-loop approaches. 33 The project showed: • The advantage of higher sensitivity and broader bandwidth than are used in today’s grid. Those signals contain valuable information about grid health but are unavailable today • The feasibility of using edge computing to identify different events with a single sensor. This included a pathway via assisted learning to exploit artificial intelligence and machine learning to enhance grid reliability. • The technology readiness of a system that senses, analyzes, and communicates in a package appropriate to be mounted on a powerline and powered by that line. • This is bridge technology. It detects the events detected by legacy equipment and provides vital details of the event that the legacy sensors cannot. In addition, it provides valuable information that is not available from legacy sensors. • Sensors can economically be located where needed and not concentrated. Sensing, computation, and communication are possible with a clamp-on sensor that can be installed in ten minutes. This project met all its SOPO requirements. More importantly, it met them in a way that met the larger DOE objectives by demonstrating the feasibility of an improved concept for monitoring the emerging grid. The legacy approach has been to detect failure and repair quickly. This system opens the door to an innovative approach in which naturally occurring precursors to failure are often detected in time for system repair or reconfiguration without an outage. A critical impediment to rapid commercialization is that this sensor’s major value is that it measures indicators of grid performance that have not been measurable previously. Consequently, the utilities have not established procedures to assimilate, process, and act upon this new data. Industry feedback has led to the insight that the commercialization path is likely through a few early applications and upgrading applications over time. This is a cell phone model in which the product is made more valuable as more apps are available. The initial sensor may include apps for fault location, detection of unintentional islanding, vegetation interactions, downed lines, and other operational considerations. The makers of power equipment or third-party vendors can then exploit the platform by developing apps that enhance reliability by detecting indications of incipient failure in time to schedule maintenance so failure is avoided. 34 APPENDIX A: Product or Technology Production Publications: • P. Paz, R. Hebner, S. Strank, E. Jimenez, S.Santoso. Rethinking High-Impedance Faults: Beyond a Low-Current Footprint. 2026 IEEE PES Transmission & Distribution (Submitted) • P. Paz and S. Strank, S. Chandra, S. Talukder and D. Ishchenko High Impedance Fault Detection and Location via Distributed Sensing. 79th Annual Conference for Protective Relay Engineers (Submitted). • Fan W, Kang N, Hebner R, Feng X. Islanding detection in rural distribution systems. Energies. 2020 Oct 20;13(20):5503. • Feng X, Strank SM, Paz P, Hebner RE. Enhancing fault identification in rural power distribution systems. Electric Power Systems Research. 2025 Apr 1;241:111331. • Feng X, Hebner R, Strank S. Design consideration for power line sensors in power distribution systems. In2020 IEEE Energy Conversion Congress and Exposition (ECCE) 2020 Oct 11 (pp. 456-462). IEEE 35