Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “time-series modeling”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 109 records · Page 6
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Explainable AI for Multivariate Time Series Pattern Exploration: Latent Space Visual Analytics With Temporal Fusion Transformer and Variational Autoencoders in Power Grid Event Diagnosis

Detecting and analyzing complex patterns in multivariate time-series data is crucial for decision-making in urban and environmental system operations. However, challenges arise from the high dimensionality, intricate complexity, and interconnected nature of complex patterns, which hinder the understanding of their underlying physical processes. Existing AI methods often face limitations in interpretability, computational efficiency, and scalability, reducing their applicability in real-world scenarios. This paper proposes a novel visual analytics framework that integrates two generative AI models, Temporal Fusion Transformer (TFT) and Variational Autoencoders (VAEs), to reduce complex patterns into lower-dimensional latent spaces and visualize them in 2D using dimensionality reduction techniques such as PCA, t-SNE, and UMAP with DBSCAN. These visualizations, presented through coordinated and interactive views and tailored glyphs, enable intuitive exploration of complex multivariate temporal patterns, identifying patterns’ similarities and uncover their potential correlations for a better interpretability of the AI outputs. The framework is demonstrated through a case study on power grid signal data, where it identifies multi-label grid event signatures, including faults and anomalies with diverse root causes. Additionally, novel metrics and visualizations are introduced to validate the models and assess the performance, efficiency, and consistency of latent maps generated by VAE, which have been utilized in prior studies for latent space cartography and used as a benchmark in this study, and the emerging TFT architecture under various configurations. These analyses provide actionable insights for model parameter tuning and reliability improvements. Comparative results highlight that TFT achieves shorter run times and superior scalability to diverse time-series data shapes compared to VAE. This work advances fault diagnosis in multivariate time series, fostering explainable AI to support critical system operations.

Explainable AI↗
DOE OSTI2025

A Neural Optimizer With Decision-Focused Learning for Optimal Energy Storage Operation

Here, this article introduces a neural optimizer-based framework for optimizing battery energy storage system (BESS) control for grid services, including demand charge and energy cost reduction. By leveraging decision-focused learning (DFL), the proposed framework ensures seamless integration and adaptation, significantly enhancing control performance. A patch time-series transformer is employed for peak load forecasting, incorporating aleatoric uncertainty quantification to account for forecasting uncertainties within the decision-making process. The framework utilizes a solver-in-the-loop approach to generate optimal BESS actions, which are then used to train the neural optimizer-based agent. By co-optimizing both BESS operational modes and output power within the NN, the system achieves improved performance and robustness. After initial training, the forecasting and control models are jointly fine-tuned to account for forecasting errors, further improving decision precision and efficiency through DFL. Case studies are performed to validate the performance of the framework using multiple real-world datasets, demonstrating superior performance in monthly peak load forecasting compared to state-of-the-art models. In addition, the results are compared against existing decision-making approaches. The results demonstrate a reduction in monthly peak forecasting error by approximately 15% across various performance measures and achieve an optimization gap for BESS operation that is about three times smaller compared to existing methods.

Kim, Hyeonjin [Pacific Northwest National Laborato↗
DOE OSTI2024

Deep Multitask Learning Models for Radiation Estimation at High Energy Accelerator Facility

Controlling the dose of radiation exposure in potential radioactive facilities is critical for ensuring the safety of staff and the public. Here, in this paper, we developed machine learning models to estimate radiation exposure efficiently at the Thomas Jefferson National Accelerator Facility (JLab), aiming to enhance safety in both accelerator facilities and public areas. Multiple sensors were deployed around the three experimental halls at JLab. Data on single-beam currents, energy levels, and radiation values at the sensor locations were collected during accelerator operation. We proposed a multi-task learning model for radiation estimation, utilizing either one-dimensional convolutional neural networks (1-D CNNs) or long short-term memory networks (LSTMs) as the backbone. The proposed model was trained to simultaneously estimate radiation levels at the sensor locations. Experimental results demonstrated that the proposed model with LSTM backbone achieved the best estimation performance, with an average R 2 score of 0.7557 for estimation within the same year and 0.7157 for estimation across different years. These results significantly surpassed those of competing models.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

Solving high-dimensional inverse problems using amortized likelihood-free inference with noisy and incomplete data

Here, we present a likelihood-free probabilistic inversion method based on normalizing flows for high-dimensional inverse problems. The proposed method is composed of two complementary networks: a summary network for data compression and an inference network for parameter estimation. The summary network encodes raw observations into a fixed-size vector of summary features, while the inference network generates samples of the approximate posterior distribution of the model parameters based on these summary features. The posterior samples are produced in a deep generative fashion by sampling from a latent Gaussian distribution and passing these samples through an invertible transformation. We construct this invertible transformation by sequentially alternating conditional invertible neural network and conditional neural spline flow layers. The summary and inference networks are trained simultaneously. We apply the proposed method to an inversion problem in groundwater hydrology to estimate the posterior distribution of the log-conductivity field conditioned on spatially sparse time-series observations of the system’s hydraulic head responses. The conductivity field is represented with 706 degrees of freedom in the considered problem. Comparison with the likelihood-based iterative ensemble smoother PEST-IES method demonstrates that the proposed method accurately estimates the parameter posterior distribution and the observations’ predictive posterior distribution at a fraction of the inference time of PEST-IES.

conditional invertible neural network↗
DOE OSTI2024

Progress of Gas Injection EOR Surveillance in the Bakken Unconventional Play—Technical Review and Machine Learning Study

Although considerable laboratory and modeling activities were performed to investigate the enhanced oil recovery (EOR) mechanisms and potential in unconventional reservoirs, only limited research has been reported to investigate actual EOR implementations and their surveillance in fields. Eleven EOR pilot tests that used CO2, rich gas, surfactant, water, etc., have been conducted in the Bakken unconventional play since 2008. Gas injection was involved in eight of these pilots with huff ‘n’ puff, flooding, and injectivity operations. Surveillance data, including daily production/injection rates, bottomhole injection pressure, gas composition, well logs, and tracer testing, were collected from these tests to generate time-series plots or analytics that can inform operators of downhole conditions. A technical review showed that pressure buildup, conformance issues, and timely gas breakthrough detection were some of the main challenges because of the interconnected fractures between injection and offset wells. The latest operation of co-injecting gas, water, and surfactant through the same injection well showed that these challenges could be mitigated by careful EOR design and continuous reservoir monitoring. Reservoir simulation and machine learning were then conducted for operators to rapidly predict EOR performance and take control actions to improve EOR outcomes in unconventional reservoirs.

Energy & Fuels↗
DOE OSTI2025

ComStock Measure Scenario Documentation: Chiller Replacement

Building on a 3-year effort to calibrate and validate the U.S. Department of Energy's ResStock (TM) and ComStock (TM) models, this work produces national datasets that empower analysts working for federal, state, utility, city, and manufacturer stakeholders to answer a broad range of questions regarding their commercial building stock. ComStock is a highly granular, bottom-up model that uses multiple data sources, statistical sampling methods, and advanced building energy simulations to estimate the annual energy consumption (at a subhourly resolution) of the commercial building stock across the United States. The baseline model intends to represent the U.S. commercial building stock as it existed in 2018. The methodology and results of the baseline model are discussed in the final technical report of the End-Use Load Profiles project. The goal of this work is to develop energy efficiency and demand flexibility end-use load shapes that cover high-impact, market-ready (or nearly market-ready) measures. "Measures" refers to various "what-if" scenarios that can be applied to buildings. An end-use savings shape is the difference in energy consumption between a baseline building (or collection of buildings) and a building with an energy efficiency or demand flexibility measure applied. It results in a time-series profile broken down by end use and fuel (electricity or on-site gas, propane, or fuel oil use) at each time step, as well as annual aggregations. This report describes the modeling methodology for a single end-use savings shape measure - chiller replacement - and briefly introduces key results. The full public dataset can be accessed on the ComStock (TM) data lake or via the Data Viewer at comstock.nrel.gov. The public data set enables users to create custom aggregations of results for their use case (e.g., filter to a specific county).

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2025

ComStock Measure Scenario Documentation: Standard Performance Heat Pump Rooftop Unit With New Windows

Building on a 3-year effort to calibrate and validate the U.S. Department of Energy's ResStock (TM) and ComStock (TM) models, this work produces national datasets that empower analysts working for federal, state, utility, city, and manufacturer stakeholders to answer a broad range of questions regarding their commercial building stock. ComStock is a highly granular, bottom-up model that uses multiple data sources, statistical sampling methods, and advanced building energy simulations to estimate the annual energy consumption (at subhourly resolution) of the commercial building stock across the United States. The baseline model intends to represent the U.S. commercial building stock as it existed in 2018. The methodology and results of the baseline model are discussed in the final technical report of the End-Use Load Profiles project. The goal of this work is to develop energy efficiency and demand flexibility end-use load shapes that cover high-impact, market-ready (or nearly market-ready) measures. "Measures" refers to various "what-if" scenarios that can be applied to buildings. An end-use savings shape is the difference in energy consumption between a baseline building (or collection of buildings) and a building with an energy efficiency or demand flexibility measure applied. It results in a time-series profile broken down by end use and fuel (electricity or on-site gas, propane, or fuel oil use) at each time step, as well as annual aggregations.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2025

AI-Enabled Operations at Fermi Complex: Multivariate Time Series Prediction for Outage Prediction and Diagnosis

The Main Control Room of the Fermilab accelerator complex continuously gathers extensive time-series data from thousands of sensors monitoring the beam. However, unplanned events such as trips or voltage fluctuations often result in beam outages, causing operational downtime. This downtime not only consumes operator effort in diagnosing and addressing the issue but also leads to unnecessary energy consumption by idle machines awaiting beam restoration. The current threshold-based alarm system is reactive and faces challenges including frequent false alarms and inconsistent outage-cause labeling. To address these limitations, we propose an AI-enabled framework that leverages predictive analytics and automated labeling. Using data from $2,703$ Linac devices and $80$ operator-labeled outages, we evaluate state-of-the-art deep learning architectures, including recurrent, attention-based, and linear models, for beam outage prediction. Additionally, we assess a Random Forest-based labeling system for providing consistent, confidence-scored outage annotations. Our findings highlight the strengths and weaknesses of these architectures for beam outage prediction and identify critical gaps that must be addressed to fully harness AI for transitioning downtime handling from reactive to predictive, ultimately reducing downtime and improving decision-making in accelerator management.

Jain, Milan [PNL, Richland] (ORCID:000000021676111↗
DOE OSTI2024

Design, Detection, and Countermeasure of Frequency Spectrum Attack and Its Impact on Long Short-Term Memory Load Forecasting and Microgrid Energy Management

This paper introduces a frequency-domain false data injection attack called Frequency Spectrum Attack (FSA) and explores its effects on load forecasting and the energy management system (EMS) in a microgrid. The FSA analyzes time-series signals in the frequency domain to identify patterns in their frequency spectrum. It learns the distribution of dominant frequencies in a dataset of healthy signals. Subsequently, it manipulates the amplitudes of dominant frequencies within this healthy distribution, ensuring a stealthy attack against statistical analysis of the signal spectrum. We evaluated the performance of FSA on LSTM, a state-of-the-art network for load forecasting. The results show that FSA can triple the Mean Absolute Error (MAE) of predictions compared to the normal case and increase it by 70% compared to noise injection attacks. Furthermore, FSA indirectly enhances battery utilization in the EMS by 45%. We then proposed a detection method that combines statistical analysis and machine-learning-based classification techniques with features. The model effectively distinguishes FSA from healthy and noisy signals, achieving an accuracy of 98.7% and an F1-score of 98.1% on a load dataset, covering healthy, FSA, and noisy load data. Finally, a countermeasure was introduced based on the statistical analysis of the frequency spectrum of healthy signals to mitigate the impact of FSA. This countermeasure successfully reduces the MAE of the attacked model from 0.135 to 0.053, validating its effectiveness in mitigating FSA.

Nazeri, Amirhossein↗
DOE OSTI2024

Comprehensive Cradle to Grave Life Cycle Analysis of On-Road Vehicles in the United States based on GREET

To properly compare and contrast the environmental performance of one vehicle technology against another, it is necessary to consider their production, operation, and end-of-life fates. Since 1995, Argonne’s GREET® life cycle analysis model (Greenhouse gases, Regulated Emissions, and Energy use in Technologies) has been annually updated to model and refine the latest developments in fuels and materials production, as well as vehicle operational and composition characteristics. Updated cradle-to-grave life cycle analysis results from the model’s latest release are described for a wide variety of fuel and powertrain options for U.S. light-duty and medium/heavy-duty vehicles. Light-duty vehicles include a passenger car, sports utility vehicle (SUV), and pick-up truck, while medium/heavy-duty vehicles include a Class 6 pickup-and-delivery truck, Class 8 day-cab (regional) truck, and Class 8 sleeper-cab (long-haul) truck. Powertrain coverage includes internal combustion (spark ignition and compression ignition) engines, hybrid electric, plug-in hybrid, full battery electric, and fuel cell vehicles powered by conventional and low carbon energy sources. The results offer insights into the current state of these technologies, as well as a projection of the likely environmental implications of future fuel and vehicle advancements through a time-series evaluation of life cycle greenhouse gas emissions.

Kelly, Jarod C.↗
DOE OSTI2025

Uncovering heterogeneous intercommunity disease transmission from neutral allele frequency time series

The COVID-19 pandemic has underscored the need for accurate epidemic forecasting to predict pathogen spread, evolution, and evaluate intervention strategies. Forecast reliability hinges on detailed knowledge of disease transmission across population segments, which may be inferred from contact surveys or mobility data. However, these indirect approaches make it difficult to estimate rare transmissions between socially or geographically distant communities. We show that the steep ramp-up of genome sequencing surveillance during the pandemic can be leveraged to directly identify transmission patterns between geographically defined communities. Our approach uses a hidden Markov model to infer the fraction of infections a community imports from others based on how rapidly allele frequencies in the focal community converge to those in the donor communities. Applying this method to SARS-CoV-2 sequencing data from England and the United States, we uncover networks of intercommunity transmission that reflect geographical relationships while exposing significant long-range interactions. The scaling of importation rate with distance is consistent across both countries, yet weaker than expected based on mobility data, highlighting limitations of indirect inference. We show that transmission patterns can change between waves of variants of concern and analyze how the inferred heterogeneity in intercommunity transmission impacts evolutionary forecasts. While applied here to geographically defined communities, our approach could be applied to those defined by other traits (e.g., age, socioeconomic status), provided time-series data can be stratified accordingly. Overall, our study highlights population genomic time series data as a crucial record of epidemiological interactions, which can be deciphered using tree-free inference methods.

Okada, Takashi [Department of Physics; University ↗
DOE OSTI2025

Profiles of Radiative Fluxes at ENA

Profiles of radiative fluxes observed at the Atmospheric Radiation Measurement (ARM)’s Eastern North Atlantic (ENA) observatory along with the ancillary measurements are reported. The below-cloud drizzle properties were derived by combining the data from the ceilometer and Ka-band ARM Zenith Radar (KAZR) following the technique explained by Ghate et al. (2021 JAMC). The cloud and drizzle water path values were derived from the brightness temperatures reported by the microwave radiometer following the technique of Cadeddu et al. (2020 AMT). The cloud water path was then scaled to the KAZR-reported radar reflectivity to calculate profiles of liquid water content (LWC). Following the analysis from Ghate et al. (2023 JGR), cloud droplet effective radius was calculated using the number concentration value of 100 cm-3. The cloud properties, along with the thermodynamic properties, served as an input to the Rapid Radiative Transfer Model (RRTM) to yield profiles of radiative fluxes at a 1-minute temporal and 50-m vertical resolution. The fluxes were then averaged to hourly temporal resolution for analysis. In Mitra et al. (2025 JClim), the calculated profiles were compared against those derived from the satellite measurements (SYN1deg). Flux profiles from the SYN1deg and the thermodynamic and cloud properties used for deriving them are also reported here. Both all-sky and clear-sky radiative flux profiles were calculated. Due to the large data volume, the surface and top-of-atmosphere (TOA) radiative fluxes for the six-year period, and the hourly profiles of the radiative fluxes for January 2018, are submitted here. Full profiles of radiative fluxes calculated from the thermodynamic and cloud properties measured at the ENA site at 1-minute temporal and 50-m vertical resolution for a six-year period are available from the authors. Six files here correspond to the following data: 1_ENARAD_CERES_with_cld_amount_timeseries.nc: Time-series of hourly values of RRTM-simulated values of upwelling and downwelling fluxes at the surface and TOA, observed boundary-layer cloud fractions, and upwelling and downwelling fluxes from the SYN1deg from July 2015 to January 2022. 2_CERES_2018_at_CERES_levels.nc: SYN1deg radiative fluxes at six levels for the year 2018. 3_ENARad_2018_at_CERES_levels.nc: RRTM calculated fluxes at the SYN1deg vertical levels for the year 2018. 4_ENARad_rrtminputs_hourly_201801.nc: Thermodynamic and cloud properties used as an input to the RRTM for January 2018. 5_CERES_inputs_hourly_201801.nc: Thermodynamic and cloud properties utilized by SYN1deg algorithm for January 2018. 6_ENARAD_hourly_201801.nc: Full profiles of hourly averaged radiative fluxes from the RRTM simulations for January 2018.

Atmosphere↗
DOE OSTI2024

Results from a multi-laboratory ocean metaproteomic intercomparison: effects of LC-MS acquisition and data analysis procedures

Metaproteomics is an increasingly popular methodology that provides information regarding the metabolic functions of specific microbial taxa and has potential for contributing to ocean ecology and biogeochemical studies. A blinded multi-laboratory intercomparison was conducted to assess comparability and reproducibility of taxonomic and functional results and their sensitivity to methodological variables. Euphotic zone samples from the Bermuda Atlantic Time-series Study (BATS) in the North Atlantic Ocean collected by in situ pumps and the autonomous underwater vehicle (AUV) Clio were distributed with a paired metagenome, and one-dimensional (1D) liquid chromatographic data-dependent acquisition mass spectrometry analysis was stipulated. Analysis of mass spectra from seven laboratories through a common bioinformatic pipeline identified a shared set of 1056 proteins from 1395 shared peptide constituents. Quantitative analyses showed good reproducibility: pairwise regressions of spectral counts between laboratories yielded R 2 values averaged 0.62±0.11, and a Sørensen similarity analysis of the top 1000 proteins revealed 70 %–80 % similarity between laboratory groups. Taxonomic and functional assignments showed good coherence between technical replicates and different laboratories. A bioinformatic intercomparison study, involving 10 laboratories using eight software packages, successfully identified thousands of peptides within the complex metaproteomic datasets, demonstrating the utility of these software tools for ocean metaproteomic research. Lessons learned and potential improvements in methods were described. Future efforts could examine reproducibility in deeper metaproteomes, examine accuracy in targeted absolute quantitation analyses, and develop standards for data output formats to improve data interoperability. Together, these results demonstrate the reproducibility of metaproteomic analyses and their suitability for microbial oceanography research, including integration into global-scale ocean surveys and ocean biogeochemical models.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Attribute-Aware RBFs: Interactive Visualization of Time Series Particle Volumes Using RT Core Range Queries

Smoothed-particle hydrodynamics (SPH) is a mesh-free method used to simulate volumetric media in fluids, astrophysics, and solid mechanics. Visualizing these simulations is problematic because these datasets often contain millions, if not billions of particles carrying physical attributes and moving over time. Radial basis functions (RBFs) are used to model particles, and overlapping particles are interpolated to reconstruct a high-quality volumetric field; however, this interpolation process is expensive and makes interactive visualization difficult. Existing RBF interpolation schemes do not account for color-mapped attributes and are instead constrained to visualizing just the density field. To address these challenges, we exploit ray tracing cores in modern GPU architectures to accelerate scalar field reconstruction. We use a novel RBF interpolation scheme to integrate per-particle colors and densities, and leverage GPU-parallel tree construction and refitting to quickly update the tree as the simulation animates over time or when the user manipulates particle radii. We also propose a Hilbert reordering scheme to cluster particles together at the leaves of the tree to reduce tree memory consumption. Finally, we reduce the noise of volumetric shadows by adopting a spatially temporal blue noise sampling scheme. Our method can provide a more detailed and interactive view of these large, volumetric, time-series particle datasets than traditional methods, leading to new insights into these physics simulations.

Particle Volumes↗
DOE OSTI2024

Bayesian inference of structured latent spaces from neural population activity with the orthogonal stochastic linear mixing model

The brain produces diverse functions, from perceiving sounds to producing arm reaches, through the collective activity of populations of many neurons. Determining if and how the features of these exogenous variables (e.g., sound frequency, reach angle) are reflected in population neural activity is important for understanding how the brain operates. Often, high-dimensional neural population activity is confined to low-dimensional latent spaces. However, many current methods fail to extract latent spaces that are clearly structured by exogenous variables. This has contributed to a debate about whether or not brains should be thought of as dynamical systems or representational systems. Here, we developed a new latent process Bayesian regression framework, the orthogonal stochastic linear mixing model (OSLMM) which introduces an orthogonality constraint amongst time-varying mixture coefficients, and provide Markov chain Monte Carlo inference procedures. We demonstrate superior performance of OSLMM on latent trajectory recovery in synthetic experiments and show superior computational efficiency and prediction performance on several real-world benchmark data sets. We primarily focus on demonstrating the utility of OSLMM in two neural data sets: μ ECoG recordings from rat auditory cortex during presentation of pure tones and multi-single unit recordings form monkey motor cortex during complex arm reaching. We show that OSLMM achieves superior or comparable predictive accuracy of neural data and decoding of external variables (e.g., reach velocity). Most importantly, in both experimental contexts, we demonstrate that OSLMM latent trajectories directly reflect features of the sounds and reaches, demonstrating that neural dynamics are structured by neural representations. Together, these results demonstrate that OSLMM will be useful for the analysis of diverse, large-scale biological time-series datasets.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Lineage frequency time series reveal elevated levels of genetic drift in SARS-CoV-2 transmission in England

Genetic drift in infectious disease transmission results from randomness of transmission and host recovery or death. The strength of genetic drift for SARS-CoV-2 transmission is expected to be high due to high levels of superspreading, and this is expected to substantially impact disease epidemiology and evolution. However, we don’t yet have an understanding of how genetic drift changes over time or across locations. Furthermore, noise that results from data collection can potentially confound estimates of genetic drift. To address this challenge, we develop and validate a method to jointly infer genetic drift and measurement noise from time-series lineage frequency data. Our method is highly scalable to increasingly large genomic datasets, which overcomes a limitation in commonly used phylogenetic methods. We apply this method to over 490,000 SARS-CoV-2 genomic sequences from England collected between March 2020 and December 2021 by the COVID-19 Genomics UK (COG-UK) consortium and separately infer the strength of genetic drift for pre-B.1.177, B.1.177, Alpha, and Delta. We find that even after correcting for measurement noise, the strength of genetic drift is consistently, throughout time, higher than that expected from the observed number of COVID-19 positive individuals in England by 1 to 3 orders of magnitude, which cannot be explained by literature values of superspreading. Our estimates of genetic drift suggest low and time-varying establishment probabilities for new mutations, inform the parametrization of SARS-CoV-2 evolutionary models, and motivate future studies of the potential mechanisms for increased stochasticity in this system.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2024

Availability and Performance Loss Factors for U.S. PV Fleet Systems

In the PV Fleet Performance Data Initiative, we partner with photovoltaic (PV) fleet owners to collect time-series PV production data and publish aggregated, anonymized results. This report is an update of our previous publications, specifically a FY 2021 performance index publication and a FY 2022 fleet degradation analysis. In this analysis, we have increased our data participants and system totals by around 10% to 8.5 GW and 24,000 separate inverter data channels. Four major analysis topics are considered in this report: Performance Index (PI) trends, PV system availability, soiling losses, and PV system degradation. Performance Index and inverter availability are assessed on a larger set of data from our FY 2021 report: 1,128 systems compared with 200 systems from before. The increased number of systems is due to an improved data quality methodology, as well as introducing new systems to the analysis. Overall results are similar to previously published values - overall inverter availability is low in the first six months of system performance before reaching steady-state by the end of the first year. Excluding this six-month startup period, system-level aggregated data shows a median (P50) system availability of 0.99 and a lower 10th percentile (P90) value of 0.95 (Figure ES-1). A dependence on system size is also demonstrated, with worse inverter availability results for larger PV systems. Causes of this effect are under investigation, but may be impacted by inverter size, which also show lower availability for larger inverter sizes. This report also investigates PI, correcting for degradation, soiling, snow, and availability. Following these corrections, the median system PI over its entire lifetime is 0.95. PI values reported here are approximately 3% lower than what we presented in our previous FY 2021 report. Soiling loss is assessed in a comprehensive way for the first time in this report. Results are presented using the COmbined Degradation and Soiling (CODS) method, as implemented in RdTools (v3.0.0a4). Soiling values are presented for 255 systems, which indicated irradiance-weighted soiling loss greater than 1%. The values have been published in an updated NREL soiling map at nrel.gov/pv/soiling.html. Finally, we investigated system degradation using three different data analysis techniques: conventional RdTools (year-on-year (YOY)), CODS, and Performance Loss Rate (PLR) analysis. Overall degradation results are consistent with our previous publications. Rerunning conventional RdTools on our updated fleet shows that some data partners have systematically fallen below the median system degradation rate (change over time) of -0.75 %/year. A comparison with PLR analysis, which looks at change in annual PI over time, shows that median system degradation is consistent with -0.5% to -0.75% per year change. However, at the P90 value, system degradation is substantially faster. These two results are consistent and indicate that resulting degradation statistics depend to a great degree on the population of PV systems making up the analysis cohort and whether soiling impacts the systems. The use of CODS for degradation analysis provides a different method for degradation assessment, which explicitly excludes the impact of recoverable soiling on degradation analysis. Excluding soiling effects yields an annual system degradation around -0.5% per year on average. This indicates that a portion of system performance loss may be attributed to periodic soiling that is not fully recovered. This report provides PV system owners/operators with background and methods to analyze PV system performance, give guidance for expected cohort performance, and performance loss values for use in pro-forma financial models, which guide new-build system design and bankability reports.

14 SOLAR ENERGY↗