Southern California Edison

Recently, the application of stochastic gradient descent (SGD) with Polyak stepsizes has gained attention and exhibited promising performance for machine learning problems. However, when the interpolation condition (where each loss attains a minimum at a globally optimal solution) is not satisfied, SGD with Polyak stepsizes encounters a primary limitation that impedes its overall effectiveness: a lack of knowledge of the optimal loss. In this study, we introduce a non-diminishing accelerated stochastic Polyak stepsize (AccSPS) with level adjustment coupled with approximate variance-reduced gradient (AVRG) descent to overcome this limitation. Our approach incorporates a decision-based level adjustment method to obtain an accurate estimation of the optimal loss. To significantly reduce memory requirements, we adopt a variance-reduced method that keeps a snapshot of average gradients after specific iterations. Theoretical analyses establish the convergence rate to the exact minimum in the non-interpolated setting. Numerical studies demonstrate the superior performance of AccSPS, showcasing a significantly lower loss when compared to state-of-the-art algorithms like DecSPS, AdaGrad, Adam, and AMSGrad, up to several orders of magnitude. Note to Practitioners-This paper addresses a common challenge in training machine learning models-namely, the difficulty of tuning learning rates when the optimal loss value is unknown. Traditional approaches, such as stochastic gradient descent (SGD) with Polyak stepsizes, work well when every training example achieves its minimum at a known optimal solution. However, in many practical settings this condition does not hold, limiting the effectiveness of these methods. The proposed solution introduces an accelerated variant of SGD, named AccSPS, which dynamically adjusts its stepsize without needing the exact optimal loss value. It does this by using a decision-based level adjustment method that estimates the optimal loss during training, coupled with a technique that reduces the memory burden by averaging gradients at strategic points. As a result, this approach not only simplifies the tuning process but also improves the overall performance of the training algorithm. Numerical experiments show that AccSPS can achieve significantly lower loss values compared to widely used algorithms like DecSPS, AdaGrad, Adam, and AMSGrad. While the method has demonstrated promising performance in theoretical and experimental studies, future work will focus on further refining these techniques for even broader applications in machine learning.

Identifying which distribution feeder carries a ground fault is a hard problem in networks compensated by a Ground Fault Neutralizer (GFN): the compensation that limits fault energy also suppresses the zero-sequence signatures that conventional protection relies on, so a fault on one feeder can de-energize an entire substation. High impedance faults (HIFs) are harder still; their small, distorted currents can escape detection while posing electrocution and wildfire risks. Existing data-driven localization methods are validated almost exclusively on simulation. To the best of our knowledge, this paper presents the first feeder-level ground fault localization framework in GFN-compensated distribution networks developed and validated entirely on utility field data, using 183 staged fault tests recorded by Southern California Edison on two live 12 kV, three-feeder substations under varying load. The task is formulated as multi-class classification over per-cycle phasor features of the zero-sequence voltage and feeder currents: fundamental and higher-order harmonic magnitudes and phases, bus–feeder phase differences, per-feeder admittances, total harmonic distortion (THD), and discrete wavelet transform (DWT) coefficients, organized into six progressively richer feature sets. Each configuration is trained separately for low impedance faults (LIFs) and HIFs and evaluated under an event-held-out protocol with multi-seed averaging and event-level majority voting, which scores each fault event as a unit so that correlated frames from one event never span the train and test split and inflate accuracy; on the strongest feature set for each task, recurrent, feedforward, and tree-ensemble classifiers of matched capacity are compared. Low-order harmonic features localize LIFs at 99.8 ± 0.1% frame-level accuracy with every held-out event correct. For HIFs, frame-level accuracy plateaus near 90% when 20–40 harmonics are retained, but event-level reliability does not follow frame accuracy: a compact 93-feature harmonic-plus-wavelet configuration localizes all 79 HIF events in event-level cross-validation, because its frame errors scatter within events instead of overturning majority votes. Classifier choice is secondary: feedforward, recurrent, and tree-ensemble models of matched capacity perform comparably, so the contributions are the per-fault-type harmonic and relational feature engineering and the event-level evaluation protocol. Feeding multi-frame sequences to the recurrent models does not improve on per-frame classification, so a per-frame feeder label is available one cycle after the analysis window fills. Trained on one substation and tested on another, localization transfers for LIFs but not for HIFs, which remain substation-specific and require local training data. The framework offers a data-driven alternative to fixed-threshold protection for compensated distribution networks.

Transportation electrification is a cornerstone of the transition to clean energy. As households, businesses, and public agencies increasingly replace gasoline-powered vehicles with electric vehicles (EVs), the demand for charging infrastructure accelerates, driving a substantial increase in electricity consumption. To support effective grid planning, utilities must leverage granular geographic data to anticipate when and where EVs will emerge or expand across the service territory. This requires a suite of predictive models, including: 1) Detection models to identify existing EVs that remain unaccounted for in utility databases. 2) Propensity models to estimate which customers are most likely to adopt EVs in the near future. 3) Forecasting models to project future energy demand growth in specific service areas, enabling proactive infrastructure investment. 4) EV load disaggregation model to improve the accuracy in estimating incremental demand attributable to EV charging. This paper presents a bottom-up blueprint that integrates diverse datasets to build targeted predictive models aligned with distinct operational and planning needs. In addition to machine learning techniques, the paper explores innovative applications of graph theory, survival analysis and the Bass diffusion model to enhance predictive accuracy and streamline the modeling process. The framework and data sources presented in this paper are adaptable for a range of modeling applications, including the adoption of solar panels, battery storage, and other green energy technologies.

The increasing demand for electrical energy and the integration of distributed energy resources (DERs) are driving the need for new substation deployments and equipment upgrades to maintain safe and reliable service. Minimizing the physical footprint of distribution substations has become increasingly important and can be achieved through modular design in both primary switchgear and protection and control systems. Modular or centralized protection relays, capable of protecting multiple assets within a single device, represent a key element of this approach.This paper presents a case study on the implementation and testing of Centralized Protection and Control (CPC) relays for a distribution substation. The project was conducted in two phases. The first phase evaluated a CPC relay equipped with nine sets of three-phase CT inputs and one set of VT inputs, configured to protect a 69 kV transformer bank, a medium-voltage busbar, and two 12 kV feeders. The second phase tested relay with process bus capabilities, subscribing to IEC 61850-9-2LE sampled value streams from nine merging units for the same substation configuration.Extensive laboratory testing was performed to assess relay performance under a wide range of abnormal and fault conditions. Relay and Merging Unit equipment from different vendors was included to validate interoperability. Internal relay logic was engineered to comply with well-defined Southern California Edison (SCE) protection standards.Authors discuss a comparative performance assessment between relays with conventional CT/VT inputs and IED receiving IEC 61850-9-2LE SMVs from the Merging Units. The results show that while conventional inputs provide proven reliability, the process bus configuration significantly reduces wiring complexity, improves scalability, and deliver comparable protection performance when proper time synchronization is ensured.

Some 43 years have passed since federal legislation established 1998 as the year for the United States Department of Energy (DOE) to begin taking title and possession of used nuclear fuel from commercial nuclear power plants including the now-retired San Onofre Nuclear Generating Station (SONGS). Southern California Edison (SCE) is decommissioning the retired nuclear plant but cannot restore the land until the used fuel is relocated offsite. While used fuel is safely stored at San Onofre, the material eventually must be permanently isolated from the biosphere in a deep geologic repository. Nuclear utility customers who pre-paid billions of dollars for disposal deserve solutions. Although SCE cannot control outcomes, it is catalyzing action. In 2021, SCE published the Strategic Plan for the Relocation of SONGS Spent Nuclear Fuel to an Offsite Storage Facility or a Repository. 1 At that time, SCE partnered with local governments to form Spent Fuel Solutions 2 —an advocacy coalition that has grown to more than 300 members from 18 states. SCE and Spent Fuel Solutions are working to amend federal legislation and establish a more durable used fuel program that includes near-term storage and long-term disposal. Efforts include raising awareness about the issue and supporting DOE initiatives to collaborate with potential host communities for used fuel facilities. The used fuel challenge is an old one, dating to 1982 when the Nuclear Waste Policy Act (NWPA) of 1982 3 was enacted. Legislative reform is difficult, but the current push for energy abundance is bringing renewed attention to the role of commercial nuclear power. The lack of solutions to the back end of the fuel cycle risks impeding the deployment of advanced reactors in the US. SCE is working to navigate a window of opportunity that may finally make away-from-reactor storage and disposal facilities a reality.