Search NASA⌕ Search

Engineering topics

Perr-Sauer, Jordan

Publications and source records attributed to Perr-Sauer, Jordan.

NEB-Tool (Multiple Non Energy Benefits Tool) [SWR-24-08]

A user friendly, graphical, open-source implementation of the multiple benefits framework to non-energy-benefits incorporating additional aspects to make the tool effective for use in the United States. The software may be distributed as web, desktop and/or mobile apps. Development is ongoing at the related JUSTIFI repository, found here: https://github.com/ORNL-AMO/JUSTIFI

Perr-Sauer, Jordan↗

BUTTER-Clarifier [SWR-23-61]

This repository contains a python package of neural network interpretability techniques (interpretability) and a keras callback to easily compute and capture data related to these techniques (values metrics) during training. It was developed to be used with NREL's BUTTER Deep Learning Experimental Framework, but does not depend on this framework and may be useful to projects outside of this framework. The vision for this codebase is to collect algorithms for explainable artificial intelligence (XAI) in a single framework that is easy to use, easy to read, and can be expand upon. Here, we package XAI algorithms into a module called "metrics", which are implemented as python functions. The return type of a metric is typically a dictionary holding data of multiple data types, such as real values and numpy matrices. Callbacks and any other connector code is provided as necessary in a separate module to make these metrics more easily usable. This project depends on Tensorflow's Keras API, although it would be nice to try and support multiple backends one day. Related to: https://github.com/NREL/BUTTER-Empirical-Deep-Learning-Experimental-Framework

Perr-Sauer, Jordan↗

BUTTER-E - Energy Consumption Data for the BUTTER Empirical Deep Learning Dataset

The BUTTER-E - Energy Consumption Data for the BUTTER Empirical Deep Learning Dataset adds node-level energy consumption data from watt-meters to the primary sweep of the BUTTER - Empirical Deep Learning Dataset. This dataset contains energy consumption and performance data from 63,527 individual experimental runs spanning 30,582 distinct configurations: 13 datasets, 20 sizes (number of trainable parameters), 8 network "shapes", and 14 depths on both CPU and GPU hardware collected using node-level watt-meters. This dataset reveals the complex relationship between dataset size, network structure, and energy use, and highlights the impact of cache effects. BUTTER-E is intended to be joined with the BUTTER dataset (see "BUTTER - Empirical Deep Learning Dataset on OEDI" resource below) which characterizes the performance of 483k distinct fully connected neural networks but does not include energy measurements.

Array↗

An independent analysis of bias sources and variability in wind plant pre-construction energy yield estimation methods

The wind resource assessment community has long had the goal of reducing the bias between wind plant pre-construction energy yield assessment (EYA) and the observed annual energy production (AEP). This comparison is typically made between the 50% probability of exceedance (P50) value of the EYA and the long-term corrected operational AEP (hereafter OA P50), and is known as the P50 bias. The industry has critically lacked an independent analysis of bias reduction investigated across multiple consultants to identify the greatest sources of uncertainty and variance in the EYA process and the best opportunities for uncertainty reduction. The present study addresses this gap by benchmarking consultant methodologies against each other and against operational data at a scale not seen before in industry collaborations. We consider data from 10 wind plants and evaluate discrepancies between eight consultancies in the steps taken from estimates of gross to net energy. Consultants tend to overestimate the gross energy produced at the turbines and then compensate by further overestimating downstream losses, leading to a mean P50 bias near zero, still with significant variability among the individual wind plants. Within our data sample, we find that consultant estimates of all loss categories, except environmental losses, tend to reduce the project-to-project variability of the P50 bias. The disagreement between consultants, however, remains flat throughout the addition of losses. Finally, we find that differences in consultants’ estimates of project performance can lead to differences up to $10/MWh in the levelized cost of energy for a wind plant.

Todd, Austin C.↗

JobQueue-PG: A Task Queue for Coordinating Varied Tasks Across Multiple HPC Resources and HPC Jobs

The software allows for queueing and dispatch of tasks of small, varied, or uncertain runtimes across multiple HPC jobs, resources, and other computing systems. The software was designed to allow scientists to enqueue, run, and accumulate results from computational experiments in an efficient, manageable manner. For example, the software can be used to enqueue many small computational experiments and run them using several long-running multi-node HPC jobs that may or may not run simultaneously.

Tripp, Charles↗

Wind Plant Performance Prediction Benchmark Phase 1 (Technical Report)

Financial risk resulting from the uncertainty associated with developing, owning, and operating wind power plants remains a barrier to reducing the levelized cost of energy (LCOE). On average, modern wind power plants in the U.S. underperform their expected annual energy output by 3.5-4.5% , with many underperforming by over 10%. To compensate for this uncertainty, investors require a larger return on investment (ROI) and apply "knock-down" factors that mask much of the underlying sources of uncertainty. Wind energy projects thus have reduced access to low-cost capital. Furthermore, operating wind plants often take a simple approach to estimating operations & maintenance (O&M) costs (e.g. straight-line estimates based on similar plants), which can eat into profits. To overcome these issues, the wind industry must improve the models they use for estimating wind plant performance and operations. An industry consortium (IC) requested that the National Renewable Energy Laboratory (NREL) lead a Department of Energy (DOE) working group to benchmark the accuracy of wind power plant energy predictions against real operational data. The IC was also motivated by DOE and NREL's potential to characterize systematic energy underperformance, identify sources of uncertainty, and explore root causes. The Wind Plant Performance Prediction (WP3) project was created out of this request, and this report represents the successful completion of Phase 1 of the WP3 project. During the project, wind plant owners provided both pre-construction and operational data to NREL. The pre-construction data was provided to wind resource assessment (WRA) consultants so they could conduct energy yield assessments (EYA). NREL took all of the completed EYAs, along with the operational data, and conducted an operational assessment to benchmark the EYA results against actual operational data. Given the large amounts of sensitive data required for this effort, as well as historical opposition to sharing data within industry, successful completion of Phase 1 represents an unprecedented milestone for industry data sharing. To improve the accuracy and confidence of pre-construction EYAs, wind plant owners and investors need better, more certain, energy yield predictions. The WP3 Benchmark Project is an industry-driven response to this reality. For the first time, industry has taken the important step of working together at scale, sharing valuable operational data with DOE and NREL in order to investigate the sources of bias and uncertainty in these energy estimates. This IC provides wind plant preconstruction and operational data to NREL in an organized and documented fashion and provides guidance and feedback as needed. The IC also provides introspection of the design of experiment, key metrics of success, data challenges, analysis best practices, and quality of results.

17 WIND ENERGY↗

Code Artifact for: Clustering Analysis of Commercial Vehicles Using Automatically Extracted Features from Time Series Data [SWR-21-96]

This repository contains data ingestion, feature extraction, and analysis code used in NREL Technical report "Clustering Analysis of Commercial Vehicles Using Automatically Extracted Features from Time Series Data." The code is written in Python. The ETL and feature extraction code must be run in a Spark context. The analysis code can be run without Spark, provided you have pre-computed features in a CSV file. Analysis code related to the NREL Technical Report NREL/TP-2C00-74212. Includes PySpark functions to perform trip segmentation and feature extraction over big time series data in Apache Spark. Includes "domain specific" features such as Aerodynamic Speed (ft/s), Characteristic Acceleration (ft/s2), Percent Below 55 (%), Percent Zero (%), Stops Per Mile, Average Speed (mph), Maximum Speed (mph), and Speed Standard Deviation (mph). Includes Pyspark UDF to compute "domain agnostic" features using the TSFresh library. This software record also includes the analysis notebooks and code to generate the results in the previously mentioned technical report.

Perr-Sauer, Jordan↗

OpenOA: An Open-Source Codebase For Operational Analysis of Wind Farms

OpenOA is an open source framework for operational data analysis of wind energy plants, implemented in the Python programming language. OpenOA provides a common data model, high level analysis workflows, and low-level convenience functions that engineers, analysts, and researchers in the wind energy industry can use to facilitate analytics workflows on operational data sets. OpenOA contains documentation, worked out examples in Jupyter notebooks, and a corresponding example dataset from the Engie Renewable’s La Haute Borne Dataset.

17 WIND ENERGY↗

The Power Curve Working Group's assessment of wind turbine power performance prediction methods

Wind turbine power production deviates from the reference power curve in real-world atmospheric conditions. Correctly predicting turbine power performance requires models to be validated for a wide range of wind turbines using inflow in different locations. The Share-3 exercise is the most recent intelligence-sharing exercise of the Power Curve Working Group, which aims to advance the modeling of turbine performance. The goal of the exercise is to search for modeling methods that reduce error and uncertainty in power prediction when wind shear and turbulence digress from design conditions. Herein, we analyze data from 55 wind turbine power performance tests from nine contributing organizations with statistical tests to quantify the skills of the prediction-correction methods. We assess the accuracy and precision of four proposed trial methods against the baseline method, which uses the conventional definition of a power curve with wind speed and air density at hub height. The trial methods reduce power-production prediction errors compared to the baseline method at high wind speeds, which contribute heavily to power production; however, the trial methods fail to significantly reduce prediction uncertainty in most meteorological conditions. For the meteorological conditions when a wind turbine produces less than the power its reference power curve suggests, using power deviation matrices leads to more accurate power prediction. We also determine that for more than half of the submissions, the data set has a large influence on the effectiveness of a trial method. Overall, this work affirms the value of data-sharing efforts in advancing power curve modeling and establishes the groundwork for future collaborations.

17 WIND ENERGY↗

Clustering Analysis of Commercial Vehicles Using Automatically Extracted Features from Time Series Data

Standard of practice approaches to time series cluster analysis involve careful feature engineering, often utilizing expert input to tune and select features by hand. In many cases, expert input may not be readily available, or there may not yet exist a community consensus on the ideal features for a given application. This paper compares the results of several cluster analysis methods, using both hand selected features and those extracted automatically, when applied to large geospatial time series telematics data from commercial trucking fleets. The impacts of feature selection, dimensionality reduction, and choice of clustering algorithm on the quality of clustering results are explored. Results from this analysis confirm prior results that domain agnostic features are competitive with the hand engineered features with respect to clustering quality metrics. These results also provide new insight into the most successful strategies for identifying structure in large unstructured vehicle telematics data, and suggest that time series clustering using automatic feature extraction can be an effective approach to extract structure from large scale geospatial time series data in cases when hand selected features are not available.

33 ADVANCED PROPULSION SYSTEMS↗

Short-term wind forecasting using statistical models with a fully observable wind flow

The utility of model output data from the Weather Research and Forecasting mesoscale model is explored for very short-term forecasting (5-30 minutes horizon) of wind speed to be used in large scale simulations of an autonomous electric power grid. Using this synthetic data for the development and evaluation of short-term forecasting algorithms offer many unique advantages over observational data, such as the ability to observe the full wind flow field in the surrounding region. Several short-term forecasting algorithms are implemented and evaluated using the synthetic data at several different time horizons and for three different geographic locations. Comparison is made with observational data from one location. We find that short-term forecasts of the synthetic data considering wind flow from the surrounding region perform 26% better than persistence in terms of root mean square error at the 5-minute time horizon. This improvement is comparable to studies of observational data in the literature. These results provide motivation to use synthetic data for short term forecasting in grid simulations, and open the door to future algorithmic improvements.

17 WIND ENERGY↗

The Power Curve Working Group's Assessment of Wind Turbine Power Performance Prediction Methods

A wind turbine power curve is often only strictly valid for a subset of all atmospheric conditions (i.e., the inner range), while wind turbines also operate in other scenarios (i.e., the outer range). Hence, modeling the power output in real-world conditions is a fundamental challenge. For example, the power deviation matrix (PDM) in Figure 1 displays an overprediction of power production using a reference power curve when wind speed (WS) and turbulence intensity (TI) are both low. Many experts use the PDM approach to observe any systematic bias in power curves and correct this in energy yield models. The mission of the Power Curve Working Group (PCWG) is to bring together wind industry stakeholders to help identify, validate, and develop ways to improve modeling of wind turbine performance in all atmospheric conditions.

17 WIND ENERGY↗

Wind Plant Performance Prediction (WP3) Benchmark Introduction

This presentation outlines the results from the Wind Plant Performance Prediction (WP3) Benchmarking group, which is concentrated on predicted versus actual wind plant performance, honing in on the variables that can cause over-estimation of power generation and subsequent 'under performance.'

17 WIND ENERGY↗

The importance of atmospheric turbulence and stability in machine-learning models of wind farm power production

Machine learning is frequently applied in the wind energy industry to build statistical models of wind farm power production using atmospheric data as input. In the field of wind power forecasting, in particular, there has been substantial research into finding the best-performing learning algorithms that improve model predictions. Overlooked in the literature, however, is the influence of atmospheric turbulence and stability measurements in improving model predictions. It has been well-established through observations and physical models that these effects can have considerable influence on wind farm power production; yet consideration of these effects in statistical models is almost entirely absent from the literature. In this work, we examine the impact of atmospheric turbulence and stability inputs on statistical model predictions of wind farm power output. Hourly observations from a wind farm in the Pacific Northwest United States located in very complex terrain are used. Five common learning algorithms and nine atmospheric variables are considered, five of which represent some measure of turbulence or stability. Here, we find a considerable improvement in hourly power predictions when some measure of turbulence or stability is included in the model. In particular, turbulent kinetic energy was found to be the most important variable apart from wind speed and more important than wind direction, pressure, and temperature. By contrast, the choice of learning algorithm is shown to be relatively less important in improving predictions. Based on this work, we recommend that turbulence and stability variables become standard inputs into statistical models of wind farm power production.

17 WIND ENERGY↗

Leveraging Big Data Analysis Techniques for U.S. Vocational Vehicle Drive Cycle Characterization, Segmentation, and Development

Under a collaborative interagency agreement between the U.S. Environmental Protection Agency and the U.S. Department of Energy (DOE), the National Renewable Energy Laboratory (NREL) performed a series of in-depth analyses to characterize on-road driving behavior including distributions of vehicle speed, idle time, accelerations and decelerations, and other driving metrics of medium- and heavy-duty vocational vehicles operating within the United States. As part of this effort, NREL researchers segmented U.S. medium- and heavy-duty vocational vehicle driving characteristics into three distinct operating groups or clusters using real-world drive cycle data collected at 1 Hz and stored in NREL's Fleet DNA database. The Fleet DNA database contains millions of miles of historical drive cycle data captured from medium- and heavy-duty vehicles operating across the United States. The data encompass existing DOE activities as well as contributions from valued industry stakeholder participants. For this project, data captured from 913 unique vehicles comprising 16,250 days of operation were drawn from the Fleet DNA database and examined. The Fleet DNA data used as a source for this analysis has been collected from a total of 30 unique fleets/data providers operating across 22 unique geographic locations spread across the United States. This includes locations with topographies ranging from the foothills of Denver, Colorado, to the flats of Miami, Florida. This paper includes the results of the statistical analysis performed by NREL and a discussion and detailed summary of the development of the vocational drive cycle weights and representative transient drive cycles for testing and simulation. Additional discussion of known limitations and potential future work is also included.

33 ADVANCED PROPULSION SYSTEMS↗

OpenOA™, FKA: Wind Plant Performance Project (WP3) Benchmarking

This library provides a framework for working with large timeseries data from wind plants, such as SCADA. Its development has been motivated by the WP3 Benchmarking (PRUF) project, which aims to provide a reference implementation for plant-level performance assessment.

Perr-Sauer, Jordan↗