SEARCH · Engineering Papers

Results for “predictive state estimation”

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.

SourceResearch connectionFrom yearThrough year

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

At least 991 records · Page 55

All sources NASA NTRS DOE OSTI

DOE OSTI2024

Multimodal sensor fusion for real-time standoff estimation in directed energy deposition

In Laser Powder-based Direct Energy Deposition (LP-DED) systems, achieving consistency, precision and quality of produced parts requires tight control over printing parameters. One of the critical parameters is the standoff distance. Maintaining an optimal standoff height is crucial for achieving correct laser power density and powder catchment efficiency, as both laser and powder streams are focused at this distance. Here, this study introduces a novel approach using multimodal sensor fusion to predict standoff height in real-time. The proposed system integrates two low-profile, cost-effective sensors: an RGB coaxial camera and a high frequency and high dynamic range microphone. By utilizing a simple fully connected neural network, trained on a limited dataset, data fusion of these sensors allowed for the real-time prediction of the standoff height. The results demonstrate high resolution and accuracy of the predictions across multiple geometries and a wide range of standoff heights. This approach offers a simple, and cost-effective solution for real-time standoff height monitoring and lays the groundwork for future integration into commercial LP-DED systems.

42 ENGINEERING↗

DOE OSTI2025

A detailed study of pre-heating effects in electron beam melting powder bed fusion process

Metal-based additive manufacturing processes, such as powder bed fusion with electron beam (PBF-EB) process, also referred to as electron beam melting (EBM), can produce high-density parts with minimal residual stresses due to the uniform and coherent preheating of the powder bed. However, understanding and controlling the multiple stages of preheating is required to enable the production of high-quality, consistent parts of various materials. This work presents a large-scale, multi-layer, three-dimensional numerical analysis focused on studying the preheating stages for predicting thermal history during the PBF-EB process. The model follows a continuous multi-stage cyclic process, that incorporates all the main stages of the PBF-EB process for 316 L stainless steel. This includes the gradual deposition of a new powder layer, the first and second preheating levels of the powder bed, and the energy deposition during melting (excluding the actual melt-pool behavior simulation). The model employs an adaptive time-scaling approach that automatically adjusts the energy deposition for each solution time-increment. This allows for localized changes in time-resolution over an otherwise computationally expensive multi-layer procedure. The material property variations are also taken into account, with an emphasis on the subtle irreversible changes in powder effective thermal conductivity after the two requisite preheating stages of the powder bed. This effect is studied using simplified conductivity models from the literature for partially sintered powder, validated by a dedicated experiment and numerical simulation. The large-scale model is then used to estimate the actual temperatures during first and second preheating levels for 316 L steel, which is not yet fully supported commercially for PBF-EB. Model predictions are corroborated by experiments, using and analyzing IR images, taken at the completion of each layer by the machine’s built-in infrared camera. The current model also incorporates a qualitative assessment for the effects of conductivity change during pre-heating, as well as evaluates the applicability of the time-scaling approach.

36 MATERIALS SCIENCE↗

DOE OSTI2026

Oil in Alaska North Slope Gas Hydrate Reservoir: Micro-CT and Flow Simulation Insights into Permeability

Gas hydrate-bearing sands on the Alaska North Slope (ANS) host minor volumes of crude oil whose impact on formation permeability has never been quantified relative to a gas hydrate reservoir system. Here, we combine in situ pressure-core microcomputed-tomography (μ-CT), thermogravimetric analysis (TGA), gas-chromatography–mass-spectrometry (GC–MS), scanning-electron microscopy coupled with energy-dispersive spectroscopy (SEM-EDS), and Stokes-flow simulation to (i) quantify oil saturation, (ii) infer its source, and (iii) evaluate its influence on permeability after gas-hydrate dissociation. μ-CT and SEM-EDS imaging identify the host sediment as silty, with some grain-coating clay present. Eight representative 3003-voxel subvolumes extracted from a preserved pressure core (Core 15P-3c, from the B1 sand (unit B), HYDRATE 02 Geo Data Well (GDW), 886.96–887.29 m measured depth, MD) exhibit porosities of 0.34–0.37 and an average oil saturation of 0.09 ± 0.03, in comparison to an independent TGA-based estimate of 0.16. GC-MS fingerprints obtained from an adjacent depressurized core (Core 17P-1, 891.37–891.44 m MD), together with oil pore habit revealed by μ-CT, suggest that the oil is partly native but also includes contributions from mineral oil-based drilling fluid contamination. Further analysis of oil–porewater interactions and flushing tests confirms that native oil saturation in the analyzed samples ranges from 0.04 to 0.08 with an average of approximately 0.06, and the oil is highly immobile. Flow simulations demonstrate that the native oil saturation of ≈0.06 reduces permeability by approximately 50% due to both pore blockage and increased flow-path tortuosity. These findings reveal the potential presence of native oil within the ANS gas hydrate reservoir and suggest that future simulation models may need to account for its impact on permeability to improve long-term performance predictions of gas and water production.

02 PETROLEUM↗

DOE OSTI2024

Quantifying Investment Risk: Analysis of the Purchase Decision of a Nuclear Power Plant (Presentation)

Cost overruns are an ill-fated part of the deployment history of nuclear power plants (NPPs) in the United States, and yet studies increasingly show the important role nuclear technologies must play in decarbonizing the U.S. economy. Paradoxically, then, a key piece of a coherent decarbonization strategy depends on attracting investor action to a purchase where historical cost overruns have been sizable. To address this challenge, this study aims to develop a financial model that quantifies the risk of cost overruns in the decision-making process for purchasing advanced reactor concepts. Using the concept of value at risk (VaR), the model is built to evaluate financial risk nuclear construction with the aim to identify risk mitigation strategies. The objective is to identify strategies to mitigate cost-risk challenges and assess the potential reduction in investor risk exposure. This paper presents the initial development and preliminary verification of the financial risk analysis model. The development of this model involved a comprehensive approach to estimating financial risk over the operating life of NPP that stems from construction uncertainties. By utilizing net present value (NPV) with discounted cash flows, the model captures the complex interconnections of project costs, construction timelines, revenue, and uncertainties. Verifying the model involved testing historical data from previous reactor construction projects against the construction project of Vogtle 3 and 4. This paper’s results present the comparison of the preconstruction cost overrun prediction with the current cost estimates from a nearly complete Vogtle 3 and 4.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2020

Impacts of High Variable Renewable Energy Futures on Electric-Sector Decision Making: Demand-Side Effects

Previous work by the Berkeley Lab describes how high shares of variable renewable energy (VRE) such as wind and solar power could change wholesale electricity price dynamics. These include the timing of when electricity is cheap or expensive, locational differences in the cost of electricity, and the degree of regularity or predictability in those costs. Many decentralized decision-makers on the demand-side may not yet have considered the implications of these possible future changes. In this report, we evaluate the sensitivity of a set of demand-side decisions to different levels of VRE penetration ranging from a low of 5-20% to a high of 40-50%. The analysis builds on hourly wholesale energy and capacity prices in different VRE scenarios for four wholesale markets in the United States for the year 2030 (CAISO, ERCOT, NYISO, and SPP). The principal question for this exploration is whether private and public electric-sector decisions that are made based on assumptions reflecting low VRE levels still achieve their intended objective in a high VRE scenario with 40-50% wind and solar? This scoping report evaluates the impacts of changing patterns of peak system needs on the benefits of demand reductions by examining the altered value of different energy efficiency (EE) measures. Similarly, we investigate new opportunities for large energy consumers that may arise from periods with very low wholesale electricity prices. We calculate the value of new process investments (e.g., hydrogen production and other generalized electro-commodities), showcase the varying value of new product storage investments (such as reservoir extensions at a desalination plant), and estimate the benefits of increased process flexibility that uses electricity as a process-input in addition to traditional fossil fuels (e.g., district energy systems). Finally, many decentralized decision-makers and end-use customers are not directly exposed to wholesale electricity prices but instead receive price signals from their retail electricity rates. As wind and solar shares increase, we compare the economic efficiency of flat retail rates relative to more dynamic time-of-use tariffs with and without critical peak-pricing events.

14 SOLAR ENERGY↗

DOE OSTI2026

A Kinetic Model-Driven Techno-Economic Analysis of Plastic Pyrolysis: Linking Process Dynamics to Economic Viability

This study employs a kinetic model integrated into Aspen Plus to predict pyrolysis product distribution under various conditions. A techno-economic assessment calculated the minimum selling price (MSP) of pyrolysis oil under different operating conditions for the baseline capacity of 100 kta, and across eight processing capacities ranging from 30 to 150 kta. The lowest MSP under the baseline capacity is estimated at $\$$420/ton, which is 33% lower than the 2023 average US crude oil price ($\$$74.6/bbl, equivalent to $\$$634/ton based on the density of pyrolysis oil). Under Monte Carlo simulation, accounting for variability in key economic and technical parameters, the mean MSP is estimated at $\$$1137/ton. The economic viability depends on feedstock price remaining below $\$$320/ton, defining the break-even feedstock price threshold. Sensitivity analysis further identifies capital investment and transportation cost as key economic drivers. Capacities beyond 90 kta show limited economies of scale benefits. Reducing product storage time cuts capital costs by 7% but raises operational risk. Uncertainty analysis suggests the economic feasibility of pyrolysis oil is unlikely to compete with crude oil without policy incentives.

petrochemicals↗

DOE OSTI2024

Quantifying Investment Risk: Analysis of the Purchase Decision of a Nuclear Power Plant

Cost overruns are an ill-fated part of the deployment history of nuclear power plants (NPPs) in the United States, and yet studies increasingly show the important role nuclear technologies must play in decarbonizing the U.S. economy. Paradoxically, then, a key piece of a coherent decarbonization strategy depends on attracting investor action to a purchase where historical cost overruns have been sizeable. To address this challenge, this study aims to develop a financial model that quantifies risk of cost overruns in the decision-making process for purchasing advanced reactor concepts. Using the concept of Value at Risk (VaR), the model is built to evaluate financial risk nuclear construction with the aim to identify risk mitigation strategies. The objective is to identify strategies to mitigate cost-risk challenges and to assess the potential reduction in investor risk exposure. The paper presents the initial development and preliminary verification of the financial risk analysis model. The development of this model involved a comprehensive approach to estimating financial risk over the operating life of NPP that stems from construction uncertainties. By utilizing net present value (NPV) with discounted cash flows, the model captures the complex interconnections of project costs, construction timelines, revenue, and uncertainties. Verification of the model involved testing historical data from previous reactor construction projects against the construction project of Vogtle 3 and 4. The results of this paper present the comparison of the preconstruction cost overrun prediction with the current cost estimates from a nearly complete Vogtle 3 and 4.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

DOE OSTI2025

Cold Gas and Star Formation in the Phoenix Cluster with JWST

We present integral field unit observations of the Phoenix Cluster with the JWST Mid-infrared Instrument’s Medium Resolution Spectrometer. We focus this study on the molecular gas, dust, and star formation in the brightest cluster galaxy (BCG). We use precise spectral modeling to produce maps of the silicate dust, molecular gas, and polycyclic aromatic hydrocarbons (PAHs) in the inner ∼50 kpc of the cluster. We measure the optical depth from silicates by comparing the observed H 2 line ratios to those predicted by excitation models. We provide updated measurements of the total molecular gas mass of $1.9^{+0.5}_{-0.04}$ x 10 10 M ⊙ , which agrees with CO-based estimates, providing an estimate of the CO-to-H 2 conversion factor of α CO = 0.8 ± 0.2 M ⊙ pc -2 (K km s -1 ) -1 ; an updated stellar mass of M * = 2.6 ± 0.5 × 10 10 M ⊙ ; and star formation rates (SFRs) averaged over 10 and 100 Myr of $\langle$SFR$\rangle$ 10 = 1340 ± 100 M ⊙ yr −1 and $\langle$SFR$\rangle$ 100 = 740 ± 80 M ⊙ yr −1 , respectively. The H 2 emission seems to be powered predominantly by shocks and star formation within the central ∼20 kpc, induced by stellar feedback and radio jets from the active galactic nucleus. Additionally, we find nearly an order-of-magnitude drop in the SFRs estimated by PAH fluxes in cool core BCGs compared to field galaxies, suggesting that hot particles from the intracluster medium are destroying PAH grains even in the central-most tens of kiloparsecs.

Active galaxies↗

DOE OSTI2025

Quantifying CO 2 Plume Stabilization at Carbon Storage Projects, North Dakota, USA

This study presents an approach for quantifying when injected carbon dioxide (CO 2 ) stabilizes pursuant to carbon capture and storage (CCS) project permitting and site closure requirements. The distribution of mobile-phase CO 2 (CO 2 plume) will evolve within the storage reservoir during and after injection through both physical and chemical trapping mechanisms. CCS policies generally agree that the CO 2 plume’s migratory behavior in post-injection should demonstrate nonendangerment to the environment but do not provide specific guidance on how to meet the definition of plume stabilization, generating some uncertainty for operators. Plume stability herein means the CO 2 plume 1) changes size minimally and predictably in the storage reservoir such that it will not cross key boundaries identified in the permit and 2) does not pose a threat to human health, underground sources of drinking water (USDWs), and the environment because of lateral migration to areas where leakage pathways may exist. Published literature on plume metrics was reviewed to determine which metric(s) may be most appropriate for determining CO 2 plume stability. A technical approach that defines plume stabilization by estimating the rate of change in the geographic footprint of the CO 2 plume with respect to time was developed and illustrated using a case study from North Dakota, USA, as a proposed solution for CCS operators to apply at the project permitting stage. Any prospective CCS operator may benefit from using the same approach to inform the selection of pore space lease and monitoring areas and develop post-injection site care plans.

03 NATURAL GAS↗

DOE OSTI2021

Prediction of Detonation-Induced Disturbances Propagating Upstream into Inlets of Rotating Detonation Combustors

Disturbances caused by the detonation wave in a rotating detonation combustor (RDC) propagate upstream through the inlet, and can potentially affect and couple to upstream components, such as turbomachinery or isolators. These disturbances can potentially also affect the operation of the RDC itself. By drawing from the analogy of a detonation wave bounded by an inert gas, the pressure disturbances observed upstream of the inlet are explained as the consequence of the passage of an upstream propagating oblique shock. In this study, the pressure rise in the plenum from the oblique shock is measured in an axial air inlet RDC. The speed of the upstream propagating wave is estimated to be moderately above the acoustic speed of the oxidizer in the plenum. The wave propagates into the plenum despite local regions of choking in the inlet. It is estimated that the time it takes a fluid particle to transit from the plenum to the detonation channel through the inlet is much larger than the rotational time of the detonation wave. This implies that a fluid particle experiences multiple shocks prior to entering the detonation channel. The oblique shock propagating upstream through the inlet area change is modeled by leveraging an analogy with a quasi-1D shock wave moving in a variable area duct with mean (incoming) flow. Due to flow expansion along the area change, fluid particles are found to experience stronger shocks in the inlet than in the plenum, thereby creating different thermodynamic states within the fill region as the oxidizer emerges from the inlet.

Feleo, Alexander↗

DOE OSTI2025

Evaluation of Maximum Allowable Working Pressure and Svensson Burst Pressure Recommended in API 579-1 2021 Edition

Abstract API 579-1/ASME FFS-1 2021 Edition provides the minimum wall thickness, the maximum allowable working pressure (MAWP), and the membrane stress equations for thin and thick-walled cylindrical shells subject to internal pressure in Section 2C.3.3.1 of Appendix 2C – Thickness, MAWP, and Stress Equations for an FFS Assessment. The minimum wall thickness and MAWP are determined using the hoop stress and the Tresca yield criterion. Section 2C.7 – Estimation of Burst Pressure newly added the Svensson method for calculating burst pressure of cylindrical shells under internal pressure, where the plastic yielding is characterized by the von Mises yield criterion. For thin-walled cylinders, the von Mises flow solution of burst pressure in Equation (2C.179) was recommended. For thick-walled cylinders, an implicit burst pressure solution in an integral equation (2C.176) was recommended. But this integral equation is inconvenient to use in practice. It is well known that the classic plasticity theory includes the Tresca and von Mises yield criteria, with the Tresca criterion predicting a lower bound solution and the von Mises criteria predicting an upper bound solution. In addition, the present author developed an average shear stress yield criterion that can determine more accurate limit and burst pressures for thin and thick-walled cylinders. This work uses these three yield criteria to evaluate the minimum required wall thickness, MAWP and Svensson burst pressure recommended in the API 579 code.

burst pressure↗

DOE OSTI2025

Evaluation of Maximum Allowable Working Pressure and Svensson Burst Pressure Recommended in API 579-1 2021 Edition

ABSTRACT API 579-1/ASME FFS-1 2021 Edition provides the minimum wall thickness, the maximum allowable working pressure (MAWP), and the membrane stress equations for thin and thick-walled cylindrical shells subject to internal pressure in Section 2C.3.3.1 of Appendix 2C – Thickness, MAWP, and Stress Equations for an FFS Assessment. The minimum wall thickness and MAWP are determined using the hoop stress and the Tresca yield criterion. Section 2C.7 – Estimation of Burst Pressure newly added the Svensson method for calculating burst pressure of cylindrical shells under internal pressure, where the plastic yielding is characterized by the von Mises yield criterion. For thin-walled cylinders, the von Mises flow solution of burst pressure in Equation (2C.179) was recommended. For thick-walled cylinders, an implicit burst pressure solution in an integral equation (2C.176) was recommended. But this integral equation is inconvenient to use in practice. It is well known that the classic plasticity theory includes the Tresca and von Mises yield criteria, with the Tresca criterion predicting a lower bound solution and the von Mises criteria predicting an upper bound solution. In addition, the present author developed an average shear stress yield criterion that can determine more accurate limit and burst pressures for thin and thick-walled cylinders. This work uses these three yield criteria to evaluate the minimum required wall thickness, MAWP and Svensson burst pressure recommended in the API 579 code.

burst pressure↗

DOE OSTI2025

Graphite waste classification and disposal cost estimation for high temperature gas and salt reactors

As high-temperature reactor designs progress to demonstration, managing the radioactive wastes from these systems presents unique challenges. This work explores the irradiated graphite source term produced by three reactor designs: The Modular High Temperature Gas reactor (MHTGR), a pebble-bed High Temperature Gas Reactor (pb-HTGR), and a Fluoride-cooled High-temperature Reactor (FHR). We predicted a C-14 concentration of 4.3 Ci/m 3 for the MHTGR, 1.2 Ci/m 3 for the pebble bed HTGR, and 2.5 Ci/m 3 for the gFHR after 20 years of operation. The final C-14 concentration highly depended on the graphite nitrogen impurity, a major precursor for C-14. The C-14 concentration in all reactor types exceeded the 0.8 Ci/m3 threshold, resulting in a Class C waste classification. The costs associated with accepting the graphite after 20 years in a low-level waste disposal facility were projected to be 255 dollars per kWe for the MHTGR, 248 dollars per kWe for the pb-HTGR, and 56.8 dollars per kWe for the FHR.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

DOE OSTI2026

Assessing the Impact of a Forest Canopy on Near-Surface Wind Statistics

Representing the forest canopy in atmospheric numerical models should improve simulated winds within and above the canopy up to a few hundred meters above the ground. Here, in this study, we implement a forest canopy parameterization into the Weather Research and Forecasting (WRF) Model in a large-eddy simulation (LES) mode by applying drag forces across multiple layers within the canopy height. We use unique observations from the Lidar Experiments for Assessing Flow over Forests (LEAFF) field campaign at the Wind River Experimental Forest (WREF) in the U.S. Pacific Northwest to evaluate model performance. In a 2-day case study, the canopy parameterization improved wind predictions both within and above the canopy, particularly during the daytime and at finer grid resolution. Without it, winds were frequently overpredicted above the canopy. Similarly, derived quantities such as the wind shear index also yielded estimates closer to observations with the canopy parameterization implemented. These findings suggest that representing the canopy using drag forces alone can improve simulated mean winds up to 200 m above the surface. Furthermore, second-order statistical moments of wind were more sensitive to canopy density than first-order moments, especially during the daytime. This increased sensitivity and the improved daytime performance in wind speed—evidenced by the lowest bias from observations (3% compared to 20% over diurnal cycle)—imply that winds above the canopy layer are strongly influenced by how well turbulence above the canopy is modeled. The results of this study can serve as a foundation for parameterizing forest canopy effects in coarser weather forecast models.

Energy - Wind↗

DOE OSTI2025

Tracing the impacts of Mount Pinatubo eruption on regional climate using spatially-varying changepoint detection

Significant events, such as volcanic eruptions, can have global and long-lasting impacts on climate. These global impacts, however, are not uniform across space and time. Understanding how the Mt. Pinatubo eruption affects global and regional climate is of great interest for predicting the impact on climate due to similar events as well as understanding the possible effect of the stratospheric aerosol injections proposed to combat climate change. While many studies illustrated the impact of the Pinatubo eruption on a global scale, studies at a fine regional scale are scarce. Here, we propose a Bayesian spatially-varying changepoint detection and estimation method to trace the impact of Mt. Pinatubo eruption on regional climate. Our approach takes into account the diffusing nature and spatial correlation of the climate changes attributed to the volcanic eruption. We illustrate our method and demonstrate its advantages over an existing changepoint detection method through simulations. Finally, we apply our method to monthly stratospheric aerosol optical depth and surface temperature data from 1985 to 1995 to detect and estimate changepoints following the 1991 Mt. Pinatubo eruption. Our results quantitatively characterize the spatial pattern of the eruption’s impact on regional climate, complementing the previous studies on the global impact of the Pinatubo eruption.

Aerosol optical depth↗

DOE OSTI2025

An Observational Evaluation of RKW Theory over the U.S. Southern Great Plains

The theory of Rotunno et al. (“RKW” theory) addresses the behavior of squall-line cold pools in vertically sheared flows. It predicts that, within a given thermodynamic environment, a balance between baroclinic vorticity generation by the cold pool and low-level environmental vertical wind shear induces an upright updraft along the gust front that maximizes the initiation of new convective cells. Although this theory has been evaluated numerically, its applicability to observed systems remains unclear and is limited by a lack of critical measurements, including high-frequency thermodynamic and wind profiles across the gust front. Herein, observations from the Atmospheric Radiation Measurement Southern Great Plains (ARM-SGP) observatory near Lamont, Oklahoma, are used to evaluate RKW theory for 10 well-observed squall lines over a 11-yr period. For this evaluation, RKW parameters including cold-pool intensity (c), low-level ambient, line-normal vertical shear (ΔV n ), subcloud and cloud-layer updraft tilts, and multiple measures of system intensity are estimated. Furthermore, the c estimates rely on thermodynamic retrievals from the Atmosphere Emitted Radiance Interferometer (AERI), which are uncertain but verify reasonably well against independent observations. As predicted by the theory, for c/ΔV n ≥ 1, c/ΔV n correlates positively with updraft tilt and negatively with system intensity, but these results are not always statistically significant and are also sensitive to the method by which ΔV n is evaluated. Specifically, ΔV n evaluations that extend above the cold-pool top yield greater consistency with RKW predictions. Also, some measures of intensity correlate more strongly with standard moist instability metrics than with RKW parameters.

Cold pools↗

DOE OSTI2025

Numerical simulation of asteroid geometry variance on airburst threat

For an atmospheric airburst the primary source of concern when assessing uncertainty is the size and velocity. Determining these properties provides the basis for threat assessment, as the total energy of the asteroid may then be estimated, and the threat investigated thoroughly. Even with clarity as to how much energy an asteroid may deposit, a great deal of uncertainty still exists for the actual energy deposition process. One such source of uncertainty is the geometry of the incoming asteroid. The geometry of an asteroid will alter the stress distribution during entry, which adds uncertainty to when fracture will occur. Here, in this study, we use Smoothed Particle Hydrodynamics to model the atmospheric airburst of Tunguska-scale asteroids with varying geometric profiles, including a sphere, ellipsoid, binary and superellipsoid. Each asteroid is modeled as a homogenous structure with strength. We assess uncertainty through a series of planar 2D simulation cases for each geometry, comparing the source of stochasticity across geometries. A single 3D airburst simulation for each geometry is also analyzed. Additionally, the 3D cases are compared to the highly uncertain Tunguska event, predicting variance in burst height across geometries, but all bounded by theoretical burst heights proposed for Tunguska.

Airburst↗

DOE OSTI2021

Machine-learning predictions of the shale wells’ performance

The ultra-low permeability nature of shale reservoirs leads to an extended linear flow and necessitates horizontal wells with multi-stage engineered fractures to efficiently extract hydrocarbons resources. These artificially-generated and naturally-occurring fractures form complex networks that create complex flow regimes which control oil production. These fractures are neither identical nor equally-spaced, which leads to a production profile with a masked onset of the boundary-dominated flow. The combination of the extended linear flow with the indeterminate onset of the boundary-dominated flow challenges the current deterministic analytic approaches to forecast the estimated ultimate recovery (EUR). In this work, we propose a novel machine-learning approach which overcomes these challenges and provides reliable EUR estimates based on field-wide analyses. We implement a novel unsupervised machine learning (ML) methodology, which allows for automatic identification of the optimal number of features (signals) present in the data based on non-negative matrix/tensor factorization coupled with k-means clustering incorporating regularization and physics constraints. In the presented analyses, the input data to the ML algorithm is the available (public) production history from the field collected at existing unconventional reservoirs. We validate our approach through hindcasting of the production data, where we achieved an excellent agreement. In addition, our approach is able to identify the poorly-performing wells, which could benefit from early refracing. Our approach provides fast and accurate estimations of the well performance without presumptions about the state of the well or the flow regime.

03 NATURAL GAS↗