Search NASA⌕ Search

SEARCH · Search NASA

Results for “mean squared error”

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 181 records · Page 10

Machine Learning for Solar Technology Portfolio Management

Technology readiness level (TRL) labeling and forecasting today is done entirely manually. A computationally-assisted model of a TRL-like classification for estimation and tracking could enable a research and development (R&D) portfolio manager to more accurately estimate the relative level of development for technologies in that portfolio. This report describes the work to understand readiness level systems, train a machine learning algorithm to recognize textual indicators of particular readiness levels within written material from carefully chosen case studies in renewable technology development, and apply that algorithm to predict the readiness levels of other technology development cases. We account for the results of each step of the work conducted, including the predictive accuracy of the trained model: 54% correct, within a +/-1 readiness level error threshold and a root mean squared error of 1.88. Great potential remains to improve and apply this model for management of individual projects and across portfolios.

14 SOLAR ENERGY↗

Co-Firing Switchgrass and Waste Coal in A Power Plant: A Techno-Economic and Life Cycle Evaluation for The Ohio River Valley (SWITCH) (Final Technical Report for Ohio State/FE0032204)

Abandoned coal mine lands (AMLs) represent one of the most persistent environmental challenges in the United States. Prior to the enactment of the Surface Mining Control and Reclamation Act (SMCRA) in 1977, coal mining operations were not legally required to reclaim disturbed lands, leaving behind approximately 500,000 AML sites nationwide. These sites pose severe environmental and health risks, including acid mine drainage, soil and water contamination, and spontaneous combustion of waste coal piles. Millions of Americans live within one mile of these AMLs, underscoring the urgency of remediation. Traditional reclamation practices, such as planting cool-season grasses, often fail to fully restore ecological function or leverage the economic potential of these lands. This project addressed these challenges by developing integrated strategies for resource recovery, land reclamation, and sustainable energy production. This project evaluated an integrated strategy to convert this liability into an opportunity by recovering waste coal and co-firing it with switchgrass (Panicum virgatum L.) cultivated on reclaimed or marginal AML areas in existing coal-fired power plants. Switchgrass not only provides a renewable feedstock but also aids in land reclamation and carbon sequestration. 1) Remote Sensing and Machine Learning for Waste Coal Identification Using Sentinel-2 satellite imagery and supervised classification, we applied four machine learning models to detect historical waste coal piles. Random Forest achieved the highest accuracy (precision: 86%, recall: 77%). Time-series analysis revealed gradual vegetation recovery since 1986, indicating natural reclamation processes in historical sites, while active mining areas showed ongoing disturbance. This workflow enables scalable monitoring and prioritization of reclamation efforts. 2) UAS-Based Stockpile Volume Estimation To quantify recoverable waste coal, we evaluated Unmanned Aerial Systems (UAS) equipped with Light Detection and Ranging (LiDAR) and multispectral sensors. Structure-from-Motion (SfM) photogrammetry combined with interpolated Digital Terrain Models (DTMs) achieved strong agreement with LiDAR reference volumes (Root Mean Square Error (RMSE) ≈147 m 3 , Mean Absolute Percentage Error (MAPE) ≈2%). Sensitivity analysis confirmed that spatial resolution significantly influences accuracy, emphasizing the need for high-resolution data for precise volume estimation. This approach offers a scalable, cost-effective, and accurate alternative to conventional ground-based surveys. 3) Switchgrass Cultivation for Bioenergy and Water Quality Improvement We assessed the hydrological and water quality impacts of converting AMLs to switchgrass production areas using the Soil and Water Assessment Tool (SWAT). Results showed that converting 10% of the watershed area into the switchgrass production zone reduced streamflow by 3.1%, total suspended solids by 18.1%, total nitrogen by 7.6%, and total phosphorus by 6.2%, while achieving biomass yields of 8.6–9.2 metric tons per hectare. These findings highlight switchgrass as a dual-benefit strategy for land reclamation and bioenergy feedstock production. 4) Integrated Co-Firing and CCS for Carbon-Negative Power Generation We modeled co-firing scenarios using the Power Plant Flexible Model (PPFM) to evaluate plant efficiency, greenhouse gas (GHG) emissions, and levelized cost of electricity (LCOE). Without carbon capture and storage (CCS), increasing switchgrass co-firing ratios reduced LCOE from $\$$150/MWh at 0% biomass to $\$$110/MWh at full substitution. Under CCS, costs remained higher (~$\$$250/MWh at 0% biomass) but decreased to $\$$200/MWh at 100% biomass, while enabling net-zero or carbon-negative electricity due to switchgrass sequestration benefits. Although CCS introduced efficiency penalties, pairing it with biomass co-firing offset these impacts and maximized climate benefits. Overall, optimizing co-firing ratios between 60-100%, supported by reliable logistics and storage strategies, emerged as a practical pathway to balance affordability, sustainability, and net-zero or negative GHG emissions while promoting productive reuse of AMLs.

01 COAL, LIGNITE, AND PEAT↗

Regional evapotranspiration from an image-based implementation of the Surface Temperature Initiated Closure (STIC1.2) model and its validation across an aridity gradient in the conterminous US

Recent studies have highlighted the need for improved characterizations of aerodynamic conductance and temperature (g A and T 0 ) in thermal remote-sensing-based surface energy balance (SEB) models to reduce uncertainties in regional-scale evapotranspiration (ET) mapping. By integrating radiometric surface temperature (TR) into the Penman–Monteith (PM) equation and finding analytical solutions of gA and T0, this need was recently addressed by the Surface Temperature Initiated Closure (STIC) model. However, previous implementations of STIC were confined to the ecosystem-scale using flux tower observations of infrared temperature. This study demonstrates the first regional-scale implementation of the most recent version of the STIC model (STIC1.2) that integrates the Moderate Resolution Imaging Spectroradiometer (MODIS) derived TR and ancillary land surface variables in conjunction with NLDAS (North American Land Data Assimilation System) atmospheric variables into a combined structure of the PM and Shuttleworth–Wallace (SW) framework for estimating ET at 1 km × 1 km spatial resolution. Evaluation of STIC1.2 at 13 core AmeriFlux sites covering a broad spectrum of climates and biomes across an aridity gradient in the conterminous US suggests that STIC1.2 can provide spatially explicit ET maps with reliable accuracies from dry to wet extremes. When observed ET from one wet, one dry, and one normal precipitation year from all sites were combined, STIC1.2 explained 66 % of the variability in observed 8-day cumulative ET with a root mean square error (RMSE) of 7.4 mm/8-day, mean absolute error (MAE) of 5 mm/8-day, and percent bias (PBIAS) of -4 %. These error statistics showed relatively better accuracies than a widely used but previous version of the SEB-based Surface Energy Balance System (SEBS) model, which utilized a simple NDVI-based parameterization of surface roughness (z OM ), and the PM-based MOD16 ET. SEBS was found to overestimate (PBIAS = 28 %) and MOD16 was found to underestimate ET (PBIAS =-26 %). The performance of STIC1.2 was better in forest and grassland ecosystems as compared to cropland (20 % underestimation) and woody savanna (40 % overestimation). Model inter-comparison suggested that ET differences between the models are robustly correlated with g A and associated roughness length estimation uncertainties which are intrinsically connected to T R uncertainties, vapor pressure deficit (D A ), and vegetation cover. A consistent performance of STIC1.2 in a broad range of hydrological and biome categories, as well as the capacity to capture spatio-temporal ET signatures across an aridity gradient, points to the potential for this simplified analytical model for near-real-time ET mapping from regional to continental scales.

Bhattarai, Nishan (ORCID:0000000327493549)↗

A North Sea in Situ Evaluation of the Fitch Wind Farm Parameterization Within the Mellor-Yamada-Nakanishi-Niino and 3D Planetary Boundary Layer Schemes

Wind resource assessments and wind power forecasts that account for wind farm wakes are sensitive to the choice of planetary boundary layer (PBL) scheme. This work compares the one-dimensional Mellor-Yamada-Nakanishi-Niino (MYNN) PBL scheme with a three-dimensional PBL (3DPBL) scheme, evaluating predictions made with both schemes against two sets of North Sea in situ observations of wind farm wakes. The optimal PBL scheme varies based on the observations (FINO1 tower vs. aircraft), the quantity of interest (wind speed vs. turbulence kinetic energy [TKE]), and the error metric (bias, centered root mean square error [cRMSE], R2, and earth mover's distance [EMD]). Whereas 3DPBL wind speeds outperform MYNN wind speeds with respect to the cRMSE at the FINO1 site located at a single point within the turbine rotor layer, 3DPBL TKE bias is larger than MYNN TKE bias when compared to aircraft observations taken 100 m above a wind farm. Wind speeds in the aircraft region are ambiguous with regard to which PBL scheme is optimal. Aircraft MYNN wind speeds outperform 3DPBL wind speeds with respect to R2 and cRMSE but underperform with respect to bias and EMD. Future evaluations across broader temporal and spatial scales may offer further insight into model differences.

17 WIND ENERGY↗

Integration of GOES Data for Solar Resource Assessment of the Contiguous United States

The National Solar Radiation Database (NSRDB), produced by the National Laboratory of the Rockies (NLR), provides high-resolution solar resource data for the contiguous United States (CONUS) using Geostationary Operational Environmental Satellite (GOES) East and West observations. This study evaluates the integration of multi-satellite data within the GOES-East/West overlap regions, where conventional longitude-based selection methods often produce an artificial boundary seam. Our results demonstrate that an advanced blending algorithm, which incorporates sun-satellite scattering angles and satellite viewing zenith angles, improves NSRDB accuracy and creates a spatially continuous dataset. Validation against ground-based irradiance measurements reveals reductions in both percentage error (PE) and normalized Root Mean Square Error (nRMSE), particularly in the central United States. The dynamical integration of multi-satellite data provides a robust foundation for more precise modeling of solar resource and improved spatiotemporal analysis of solar ramp across the CONUS.

14 SOLAR ENERGY↗

High-degree Gravity Models from GRAIL Primary Mission Data

We have analyzed Ka‒band range rate (KBRR) and Deep Space Network (DSN) data from the Gravity Recovery and Interior Laboratory (GRAIL) primary mission (1 March to 29 May 2012) to derive gravity models of the Moon to degree 420, 540, and 660 in spherical harmonics. For these models, GRGM420A, GRGM540A, and GRGM660PRIM, a Kaula constraint was applied only beyond degree 330. Variance‒component estimation (VCE) was used to adjust the a priori weights and obtain a calibrated error covariance. The global root‒mean‒square error in the gravity anomalies computed from the error covariance to 320×320 is 0.77 mGal, compared to 29.0 mGal with the pre‒GRAIL model derived with the SELENE mission data, SGM150J, only to 140×140. The global correlations with the Lunar Orbiter Laser Altimeter‒derived topography are larger than 0.985 between l = 120 and 330. The free‒air gravity anomalies, especially over the lunar farside, display a dramatic increase in detail compared to the pre‒GRAIL models (SGM150J and LP150Q) and, through degree 320, are free of the orbit‒track‒related artifacts present in the earlier models. For GRAIL, we obtain an a posteriori fit to the S‒band DSN data of 0.13 mm/s. The a posteriori fits to the KBRR data range from 0.08 to 1.5 micrometers/s for GRGM420A and from 0.03 to 0.06 micrometers/s for GRGM660PRIM. Using the GRAIL data, we obtain solutions for the degree 2 Love numbers, k20=0.024615+/-0.0000914, k21=0.023915+/-0.0000132, and k22=0.024852+/-0.0000167, and a preliminary solution for the k30 Love number of k30=0.00734+/-0.0015, where the Love number error sigmas are those obtained with VCE.

Primary 1↗

Investigating seasonal ENSO forecast amplitude calibration for GEOS-S2S-2

The GEOS-S2S-2 is a global coupled model and assimilation system, encompassing many aspects of the Earth climate system. This project looked at seasonal (nine-month) ensemble forecasts, over the period from 1982 through the present. We explored and validated a forecast amplitude correction technique used by the North American Multi-Model Ensemble (NMME), of which GEOS is a member, for Niño3.4 sea-surface temperature (SST) anomaly predictions. The method relies on deriving a set of correction factors based on the standard deviations of hindcast and observed SST anomalies. This algorithm was implemented at the Global Modeling and Assimilation Office (GMAO) and applied to the GEOS-S2S-2 Niño 3.4 hindcasts. In general, the correction reduced the forecast amplitude. This improved the quality of the forecasts when measured by the root mean square error (RMSE) of the ensemble mean forecast. RMSE decreased for most initialization months and forecast leads. The correction had a larger impact on seasons with strong El Niño-Southern Oscillation (ENSO) events than neutral seasons. The effects of the correction on the ensemble characteristics were also examined, and a variation of the algorithm was proposed that preserves the ensemble spread. The outcome of this project is a tool that can be used to calibrate GEOS-S2S-2 ENSO forecasts and potentially improve the precision of multi-model El Niño outlooks.

ENSO↗

Hybrid Improved Empirical Mode Decomposition and Artificial Neural Network Model for the Prediction of Critical Heat Flux (CHF)

Three Hybrid artificial neural network (ANN) models namely radial basis function (RBF), generalized regression neural networks (GRNN), and multi-layer perceptron (MLP) combined with empirical mode decomposition (EMD) are developed for CHF predictive modelling using CHF experimental databases.First, the original experimental inputs data series are decomposed into several intrinsic mode functions (IMFs) and one residual by EMD, whose components are divided into high, medium and low components. The performance parameters of the hybrid models indicates that the root mean square error (RMSE) are 0.8831, 0.6522, and 0.4149; the mean absolute error (MAE) are 0.6697, 0.4636, and 0.1935. The values of the R-square of the developed prediction approach utilizing EMD-RBF, EMD-GRNN, and EMD-MLP models are 0.8553, 0.9302, and 0.9818, and the index of agreement are 0.9464, 0.9700, and 0.9894., The value of the R-square and the index of agreement of the proposed models are much higher than those of the simple models .The Pearson's test results show that the association strength between the measured and the predicted values of the proposed model EMD-MLP is the strongest. These results show the following: (a) compared with other related, recent studies, the prediction accuracy of the hybrid model EMD- MLP proposed in this research is the best hybrid model; (b) the proposed hybrid model (EMD-MLP) attains superior performance compared with simple models.

Djeddou, Messaoud↗

A comparative study of optimum and suboptimum direct-detection laser ranging receivers

A summary of previously proposed receiver strategies for direct-detection laser ranging receivers is presented. Computer simulations are used to compare performance of candidate implementation strategies in the 1- to 100-photoelectron region. Under the condition of no background radiation, the maximum-likelihood and minimum mean-square error estimators were found to give the same performance for both bell-shaped and rectangular optical-pulse shapes. For signal energies greater than 100 photoelectrons, the root-mean-square range error is shown to decrease as Q to the -1/2 power for bell-shaped pulses and Q to the -1 power for rectangular pulses, where Q represents the average pulse energy. Of several receiver implementations presented, the matched-filter peak detector was found to be preferable. A similar configuration, using a constant-fraction discriminator, exhibited a signal-level dependent time bias.

Abshire, J. B.↗

Forecasting district-scale energy dynamics through integrating building network and long short-term memory learning algorithm

With the development of data-driven techniques, building energy prediction on a district level has attracted increasing attention in recent years for revealing energy use patterns and reduction potentials. However, data acquisition in large building groups is difficult and adjacent buildings also interact with each other. To reduce data cost and incorporate the inter-building impact with the data-driven building energy model, this study proposes a deep learning predictive approach that fuses the building network model with a long short-term memory learning model for district-scale building energy modeling. The building network was constructed based on correlations between the energy use intensity of buildings, which can significantly reduce the computational complexity of the deep learning models for energy dynamic prediction. Five typical building groups with energy use data from 2015 to 2018 on two institutional campuses were selected to perform the validation experiment with TensorFlow. Based on the prediction error assessments, the results suggest that for total building energy use intensity prediction, the proposed model can achieve a mean absolute percentage error of 6.66% and a root mean square error of 0.36 kWh/m 2 , compared to 12.05% and 0.63 kWh/m 2 of the conventional artificial neural network model and to 11.06% and 0.89 kWh/m 2 for the support vector regression model.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

A Pilot Evaluation of a 4-Dimensional Cone-Beam Computed Tomographic Scheme Based on Simultaneous Motion Estimation and Image Reconstruction

Purpose: To evaluate the performance of a 4-dimensional (4-D) cone-beam computed tomographic (CBCT) reconstruction scheme based on simultaneous motion estimation and image reconstruction (SMEIR) through patient studies. Methods and Materials: The SMEIR algorithm contains 2 alternating steps: (1) motion-compensated CBCT reconstruction using projections from all phases to reconstruct a reference phase 4D-CBCT by explicitly considering the motion models between each different phase and (2) estimation of motion models directly from projections by matching the measured projections to the forward projection of the deformed reference phase 4D-CBCT. Four lung cancer patients were scanned for 4 to 6 minutes to obtain approximately 2000 projections for each patient. To evaluate the performance of the SMEIR algorithm on a conventional 1-minute CBCT scan, the number of projections at each phase was reduced by a factor of 5, 8, or 10 for each patient. Then, 4D-CBCTs were reconstructed from the down-sampled projections using Feldkamp-Davis-Kress, total variation (TV) minimization, prior image constrained compressive sensing (PICCS), and SMEIR. Using the 4D-CBCT reconstructed from the fully sampled projections as a reference, the relative error (RE) of reconstructed images, root mean square error (RMSE), and maximum error (MaxE) of estimated tumor positions were analyzed to quantify the performance of the SMEIR algorithm. Results: The SMEIR algorithm can achieve results consistent with the reference 4D-CBCT reconstructed with many more projections per phase. With an average of 30 to 40 projections per phase, the MaxE in tumor position detection is less than 1 mm in SMEIR for all 4 patients. Conclusion: The results from a limited number of patients show that SMEIR is a promising tool for high-quality 4D-CBCT reconstruction and tumor motion modeling.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

An finite element analysis surrogate model with boundary oriented graph embedding approach for rapid design

Abstract In this work, we present a boundary oriented graph embedding (BOGE) approach for the graph neural network to assist in rapid design and digital prototyping. The cantilever beam problem has been solved as an example to validate its potential of providing physical field results and optimized designs using only 10 ms. Providing shortcuts for both boundary elements and local neighbor elements, the BOGE approach can embed unstructured mesh elements into the graph and performs an efficient regression on large-scale triangular-mesh-based finite element analysis (FEA) results, which cannot be realized by other machine-learning-based surrogate methods. It has the potential to serve as a surrogate model for other boundary value problems. Focusing on the cantilever beam problem, the BOGE approach with 3-layer DeepGCN model achieves the regression with mean square error (MSE) of 0.011 706 (2.41% mean absolute percentage error) for stress field prediction and 0.002 735 MSE (with 1.58% elements having error larger than 0.01) for topological optimization. The overall concept of the BOGE approach paves the way for a general and efficient deep-learning-based FEA simulator that will benefit both industry and Computer Aided Design (CAD) design-related areas.

42 ENGINEERING↗

Thermal modeling of directed energy deposition additive manufacturing using graph theory

Purpose: The purpose of this paper is to develop, apply and validate a mesh-free graph theory–based approach for rapid thermal modeling of the directed energy deposition (DED) additive manufacturing (AM) process. Design/methodology/approach: Here, the authors develop a novel mesh-free graph theory–based approach to predict the thermal history of the DED process. Subsequently, the authors validated the graph theory predicted temperature trends using experimental temperature data for DED of titanium alloy parts (Ti-6Al-4V). Temperature trends were tracked by embedding thermocouples in the substrate. The DED process was simulated using the graph theory approach, and the thermal history predictions were validated based on the data from the thermocouples. Findings: The temperature trends predicted by the graph theory approach have mean absolute percentage error of approximately 11% and root mean square error of 23°C when compared to the experimental data. Moreover, the graph theory simulation was obtained within 4 min using desktop computing resources, which is less than the build time of 25 min. By comparison, a finite element–based model required 136 min to converge to similar level of error. Research limitations/implications: This study uses data from fixed thermocouples when printing thin-wall DED parts. In the future, the authors will incorporate infrared thermal camera data from large parts. Practical implications: The DED process is particularly valuable for near-net shape manufacturing, repair and remanufacturing applications. However, DED parts are often afflicted with flaws, such as cracking and distortion. In DED, flaw formation is largely governed by the intensity and spatial distribution of heat in the part during the process, often referred to as the thermal history. Accordingly, fast and accurate thermal models to predict the thermal history are necessary to understand and preclude flaw formation. Originality/value: This paper presents a new mesh-free computational thermal modeling approach based on graph theory (network science) and applies it to DED. The approach eschews the tedious and computationally demanding meshing aspect of finite element modeling and allows rapid simulation of the thermal history in additive manufacturing. Although the graph theory has been applied to thermal modeling of laser powder bed fusion (LPBF), there are distinct phenomenological differences between DED and LPBF that necessitate substantial modifications to the graph theory approach.

42 ENGINEERING↗

A General Spatiotemporal Imputation Framework for Missing Sensor Data

Many applications from precision agriculture, environmental monitoring and transportation networks rely on data collected across space and time over a large geographic area. Missing data poses a significant challenge for any data-driven inference and control tasks. Data imputation or the estimation of missing data can help fill these gaps by utilizing inherent spatial relationships and temporal patterns. A variety of spatiotemporal imputation models have been developed to address missing data in spatiotemporal datasets. However, these classical methods rely on the assumption that the underlying data follows a smooth trend and fail to provide accurate estimates when there is a large number of missing points in the data. Even though there are machine learning driven tensor completion approaches such as convolutional neural network based tensor completion (CoSTCo) that capture the non-linear relationships in the dataset, the transductive nature makes the algorithm less scalable. Thus, existing approaches for estimating the missing information do not effectively capture all dimensions of the spatiotemporal data structure, resulting in erroneous predictions and poor performance. The main contributions of this paper are: (1) We propose a novel inductive framework (G-LSTM) for missing data imputation that integrates a graph neural network with LSTMs to effectively capture both spatial and temporal dependencies. (2) Experimental results on a traffic dataset demonstrate that the proposed GNN integrated with an LSTM framework achieves improved imputation and maintains steady performance even when there are extreme missing conditions in comparison with the state-of-the-art imputation framework (i.e, CoSTCo). (3) The simulation results on a traffic network show up to 69% reduction in mean absolute error and 61% reduction in root mean square error when compared to CoSTCo.

Tharzeen, Aabila↗

Validation of Multiple Tools for Flat Plate Photovoltaic Modeling Against Measured Data

This report expands upon a previous work by the same authors, published in the 40th IEEE Photovoltaic Specialists conference. In this validation study, comprehensive analysis is performed on nine photovoltaic systems for which NREL could obtain detailed performance data and specifications, including three utility-scale systems and six commercial scale systems. Multiple photovoltaic performance modeling tools were used to model these nine systems, and the error of each tool was analyzed compared to quality-controlled measured performance data. This study shows that, excluding identified outliers, all tools achieve annual errors within +/-8% and hourly root mean squared errors less than 7% for all systems. It is further shown using SAM that module model and irradiance input choices can change the annual error with respect to measured data by as much as 6.6% for these nine systems, although all combinations examined still fall within an annual error range of +/-8.5%. Additionally, a seasonal variation in monthly error is shown for all tools. Finally, the effects of irradiance data uncertainty and the use of default loss assumptions on annual error are explored, and two approaches to reduce the error inherent in photovoltaic modeling are proposed.

14 SOLAR ENERGY↗

Patterns, drivers, and a predictive model of dam removal cost in the United States

Given the burgeoning dam removal movement and the large number of dams approaching obsolescence in the United States, cost estimating data and tools are needed for dam removal prioritization, planning, and execution. We used the list of removed dams compiled by American Rivers to search for publicly available reported costs for dam removal projects. Total cost information could include component costs related to project planning, dam deconstruction, monitoring, and several categories of mitigation activities. We compiled reported costs from 455 unique sources for 668 dams removed in the United States from 1965 to 2020. The dam removals occurred within 571 unique projects involving 1–18 dams. When adjusted for inflation into 2020 USD, cost of these projects totaled $\$1.522$ billion, with per-dam costs ranging from $\$1$ thousand (k) to $\$268.8$ million (M). The median cost for dam removals was $\$157$k, $\$823$k, and $\$6.2$M for dams that were< 5 m, between 5–10 m, and > 10 m in height, respectively. Geographic differences in total costs showed that northern states in general, and the Pacific Northwest in particular, spent the most on dam removal. The Midwest and the Northeast spent proportionally more on removal of dams less than 5 m in height, whereas the Northwest and Southwest spent the most on larger dam removals > 10 m tall. We used stochastic gradient boosting with quantile regression to model dam removal cost against potential predictor variables including dam characteristics (dam height and material), hydrography (average annual discharge and drainage area), project complexity (inferred from construction and sediment management, mitigation, and post-removal cost drivers), and geographic region. Dam height, annual average discharge at the dam site, and project complexity were the predominant drivers of removal cost. The final model had an R 2 of 57% and when applied to a test dataset model predictions had a root mean squared error of $\$5.09$M and a mean absolute error of $\$1.45$M, indicating its potential utility to predict estimated costs of dam removal. We developed a R shiny application for estimating dam removal costs using customized model inputs for exploratory analyses and potential dam removal planning.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

Machine learning models for maintenance cost estimation in delivery trucks using diesel and natural gas fuels

The maintenance costs can represent about 15%–60% of the cost of produced goods depending on the type of goods transported. To comply with stringent emissions regulations, diesel engines are incorporated with complex after-treatment systems that demand increased maintenance. The availability of alternative fuels such as natural gas and propane has fostered the natural gas and propane powertrain systems as well as electrification options for heavy- and medium-duty vehicles. A critical barrier to adopting alternative fuel vehicles has been the lack of knowledge on comparative vehicle maintenance/repair costs with conventional diesel. Moreover, the region of operation, the type of vehicle operation, and seasonal temperature changes also affect the duty cycle which impacts the maintenance and repair costs. This study focuses on estimating the cost-per-mile for heavy-duty vehicles using machine learning models such as random forest, xgboost, neural networks, and a super-learner model. The super-learner model achieved an error as low as 0.0068 $/mile for mean absolute error and 0.0086 $/mile for root mean square error with a coefficient of determination/R-Squared of 97.28%. Specifically, the paper investigates the data collected from the maintenance and repair costs associated with delivery trucks using diesel and natural gas fuels. Since the availability of data is the major constraint, we leveraged the data collected by West Virginia University and the partnership with fleet companies. This allows for additional information related to maintenance costs and fleet-specific maintenance practices of alternative fuel vehicles. This study promotes clean fuel technologies and enables fleet management companies to adopt alternative fuel vehicles in case of similar or lower cost of maintenance compared to diesel vehicles resulting in reduced emissions and total cost of ownership.

Katreddi, Sasanka↗

Model Calibration with Markov Chain Monte Carlo Tutorial

The purpose of this tutorial is to demonstrate how to use Markov chain Monte Carlo (MCMC) to calibrate a model. By calibration, we mean the selection of model parameters (and, when relevant, structures). A common goal in model development and diagnostics is calibration, or the identification of model structures and parameters which are consistent with data. While models can be calibrated through hand-tuning parameters or minimizing simple error metrics such as root-mean-square-error (RMSE), these approaches can underrepresent the probabilistic nature of the data-generating process, as well as the potential for multiple model configurations to be consistent with the data. Probabilistic uncertainty quantification, which is the topic of this notebook, can address these concerns. This tutorial is presented as an appendix to the e-book: Addressing Uncertainty in MultiSector Dynamics Research.

Markov chain Monte Carlo↗