To make sound investment decisions, decision makers need accurate estimates of the uncertainties present in forecasts of reservoir performance. In this work I propose a method, the integrated mismatch method, that incorporates the misfit in the history match into the estimation of uncertainty in the prediction. I applied the integrated mismatch method, which overcomes some deficiencies of existing methods, to uncertainty estimation in two reservoir studies and compared results to estimations from existing methods. The integrated mismatch method tends to generate smaller ranges of uncertainty than many existing methods. When starting from nonoptimal reservoir models, in some cases the integrated mismatch method is able to bracket the true reserves value while other methods fail to bracket it. The results show that even starting from a nonoptimal reservoir model, but as long as the experimental designs encompass the true case parameters, the integrated mismatch method brackets the true reserves value. If the experimental designs do not encompass all the true case parameters, but the true reserves value is covered by the experiments, the integrated mismatch method may still bracket the true case. This applies if there is a strong correlation between mismatch and closeness to the true reserves value. The integrated mismatch method does not need a large number of simulation runs for the uncertainty analysis, while some other methods need hundreds of runs.

To make sound investment decisions, decision makers need accurate estimates of the uncertainties present in forecasts of reservoir performance. In this work I propose a method, the integrated mismatch method, that incorporates the misfit in the history match into the estimation of uncertainty in the prediction. I applied the integrated mismatch method, which overcomes some deficiencies of existing methods, to uncertainty estimation in two reservoir studies and compared results to estimations from existing methods. The integrated mismatch method tends to generate smaller ranges of uncertainty than many existing methods. When starting from nonoptimal reservoir models, in some cases the integrated mismatch method is able to bracket the true reserves value while other methods fail to bracket it. The results show that even starting from a nonoptimal reservoir model, but as long as the experimental designs encompass the true case parameters, the integrated mismatch method brackets the true reserves value. If the experimental designs do not encompass all the true case parameters, but the true reserves value is covered by the experiments, the integrated mismatch method may still bracket the true case. This applies if there is a strong correlation between mismatch and closeness to the true reserves value. The integrated mismatch method does not need a large number of simulation runs for the uncertainty analysis, while some other methods need hundreds of runs.

For any product development process, limited time and resources are always a focus for the engineer. However, will the overall program goals be achieved with the provided time and resources? Uncertainty analysis is a tool that is capable of providing the answer to that question. Product development process uncertainty analysis employs previous knowledge in modeling, experimentation, and manufacturing in an innovative approach for analyzing the entire process. This research was initiated with a pilot project, a four-bar-slider mechanism, and an uncertainty analysis was completed for each individual product development step. The uncertainty of the final product was then determined by combining uncertainties from the individual steps. The uncertainty percentage contributions of each term to the uncertainty of the final product were also calculated. The combination of uncertainties in the individual steps and calculation of the percentage contributions of the terms have not been done in the past. New techniques were developed to evaluate the entire product development process in an uncertainty sense. The techniques developed in this work will be extended to other processes in future work.

With limited resources and time available for a typical design project, it is difficult to decide how to allocate these resources and time to produce an optimum design. Also, the question arises, ?Given the design process, available resources, and available time, will the design meet the program goals?? Uncertainty analyses of design processes addresses these issues and could substantially improve design quality, cost, and cycle time. Research to examine uncertainty in the design process employs previous experience in experimental, model, and manufacturing uncertainty, in an innovative approach for analyzing the entire design process. This research was initiated with a pilot project, a 4-bar-slider mechanism. Two new theories for the research have arisen from this pilot project. First, design optimization techniques could be used to compare steps of the design process and to help determine the overall uncertainty of the final manufactured product. Second, manufacturing uncertainty can be included as an additional random uncertainty in the analysis of the final manufactured product. While more research needs to be completed to test, apply, and expand on these theories, the pilot project has been a positive step forward. It has already produced two proposals with one funded and one awaiting a decision. This research, although in its beginning stages, could substantially improve the design process.

This thesis documents significant contributions to the quantification of input and modelling uncertainties in the simulation of nuclear power plants. This work is intended to support the simulations that are performed to demonstrate the safety of nuclear power plants in general, and in CANDU reactors specifically. The work presented in this thesis extends the methodologies for uncertainty propagation established internationally to CANDU plants and pioneers the integration of these tools with important plant features in CANDUs, such as online fueling. This thesis documents a series of simulation studies performed to quantify the impact of uncertainties (primarily nuclear data uncertainties), on simulations of CANDU stations and light water reactors (LWRs). The novel part of this work includes quantifying the role of operational feedbacks such as online refuelling and reactor control systems, and important modelling uncertainties, on CANDU simulations. To achieve this objective, this thesis examines 4 important areas as documented in journal papers. To demonstrate understanding of the tools developed for the UAM-LWR benchmark and to support the ongoing international effort, select studies from the UAM-LWR benchmark study exercises were performed and published in the first journal paper. Time-dependent PWR neutronics exercises, considering both nuclear data and manufacturing uncertainties, were completed. This work found that the relative importance of nuclear data uncertainties and manufacturing uncertainties depended on whether the parameter of interest was “local”, such as pin power factors, or “global”, such as homogenized assembly properties. The second publication in this thesis documents the adaption of the tools from the first paper to consider CANDU specific features, such as spatial control systems and online refuelling. This paper demonstrated the significant effect that consistent feedback from fuelling operations has on reducing the total uncertainty in core level simulations of CANDU plants. The tools developed for this work were used to support downstream studies by generating extensive sets of realistic initial conditions for many different possible nuclear datasets. The next publications utilized the tools developed above and then extends the methods to include operational aspects of CANDUs in the assessments for the first time. In the third paper these methods were then used to demonstrate the tools’ capabilities to simulate an operational transient (a power maneuver from 100% full power to 59% full power) in a CANDU station and compared the resultant prediction and uncertainties to measure plant responses. A further study, on the role of nuclear data and initial burnup distribution uncertainty on a CANDU plant’s response to perturbations to liquid zone controller levels, was also performed to examine the effect of the commonly used “superposition principle” utilized in industry to make safety analysis of CANDU’s various fueling states more tractable. In both cases the role of nuclear data uncertainties was generally found to be similar in magnitude to the role of uncertainty in the core initial conditions. The results of this work support the continued safe operation of CANDU nuclear generating stations in Canada by quantifying the role of select uncertainties on safety simulation outputs, informing future BEPU analysis for CANDU plants and demonstrating the exceptional flexibility of the CANDU reactor design. This is reflected in one of the major conclusions of these works, which demonstrates that the natural feedbacks in CANDU operation help to minimize the effect of uncertainties in the outcome of many safety analysis. / Thesis / Candidate in Philosophy

International Telemetering Conference Proceedings / October 28-31, 1985 / Riviera Hotel, Las Vegas, Nevada / In many engineering applications, a systems analysis is performed to study the effects of random error propagation throughout a system. Often these errors are not independent, and have joint behavior characterized by arbitrary covariance structure. The multivariate nature of such problems is compounded in complex systems, where overall system performance is described by a q-dimensional random vector. To address this problem, a computer program was developed which generates Taylor series approximations for multivariate system performance in the presence of random component variablilty. A summary of an application of this approach is given in which an analysis was performed to assess simultaneous design margins and to ensure optimal component selection.

ITC/USA 2005 Conference Proceedings / The Forty-First Annual International Telemetering Conference and Technical Exhibition / October 24-27, 2005 / Riviera Hotel & Convention Center, Las Vegas, Nevada / There are increasing demands, particularly from government agencies, to perform uncertainty analysis in order to assign accuracy bounds to telemetered data from environmental measuring transducers (pressure, acceleration, force, strain, temperature, etc.). Several requirements must be fulfilled before measurement uncertainty analysis is justified. These requirements include good measurement system design practices such as adequate low- and high-frequency response and data-sampling rates, appropriate anti-aliasing filter selection^(1), proper grounding and shielding, and many more. In addition, there are applications (e.g., flight test) in which the environment of the transducer varies with time and/or location. In these applications, it is a requisite that data-validation be performed to establish that an individual transducer responds only to the environmental stimulus that it is intended to measure. Without this validation component designed into the telemetry system, assigned accuracy bounds can be totally meaningless. This paper presents examples and describes techniques for data validation of signals from environmental measuring transducers.

M.Ing. / Experiments are being conducted with regard to heat exchange systems. However, there are errors and uncertainties attached to each system. Journals, which publish articles concerning heat transfer experiments, require an estimate of this uncertainty. These uncertainties must be calculated in order to determine how valid a set of results is. The uncertainty describes to what level one may rely on a set of experimental results and conclusions. The uncertainty was calculated by the formulation of an uncertainty equation with the use of various statistical methods. Adjustments or modifications had to be made to the present uncertainty equations in order to calculate the uncertainty in heat transfer systems. Uncertainty based on a general uncertainty equation by Schultz and Cole (1979) enabled the derivation of the equations to calculate the necessary uncertainty factor for heat transfer systems. Implementation of the equations in various experimental set-ups was achieved. The uncertainty equations yielded results that seemed consistent with the subjective view of the experimenter. Therefore, the equations were considered valid.

In distribution networks, the traditional way to eliminate network stresses caused by increasing generation and demand is to reinforce the primary network assets. A cheaper alternative is active network management (ANM) which refers to real-time network control to resolve power flow, voltage, fault current and security issues. However, there are two limitations in ANM. First, previous ANM strategies investigated generation side and demand side management separately. The generation side management evaluates the value from ANM in terms of economic generation curtailment. It does not consider the potential benefits from integrating demand side response such as economically shifting flexible load over time. Second, enhancing generation side management with load shifting requires the prediction of network stress whose accuracy will decrease as the lead time increases. The uncertain prediction implies the potential failure of reaching expected operational benefits. However, there is very limited investigation into the trade-offs between operational benefit and its potential risk. In order to tackle the challenges, there are two aspects of research work in this thesis. 1) Enhanced ANM. It proposes the use of electric vehicles (EVs) as responsive demand to complement generation curtailment strategies in relieving network stress. This is achieved by shifting flexible EV charging demand over time to absorb excessive wind generation when they cannot be exported to the supply network. 2) Uncertainty management. It adopts Sharpe Ratio and Risk Adjust Return On Capital concepts from financial risk management to help the enhanced ANM make operational decisions when both operational benefit and its associated risk are considered. Copula theory is applied to further integrate correlations of forecasting errors between nodal power injections (caused by wind and load forecasting) into uncertainty management. The enhanced ANM can further improve network efficiency of the existing distribution networks to accommodate increasing renewable generation. The cost-benefit assessment informs distribution network operators of the trade-off between investment in ANM strategy and in the primary network assets, thus helping them to make cost-effective investment decisions. The uncertainty management allows the impact of risks that arise from network stress prediction on the expected operational benefits to be properly assessed, thus extending the traditional deterministic cost-benefit assessment to cost-benefit-risk assessment. Moreover, it is scalable to other systems in any size with low computational burden, which is the major contribution of this thesis.

Understanding uncertainty in emissions inventories is critical for evaluating both air quality modeling results as well as impacts of emissions reduction strategies. This study focused on quantification of uncertainty due to non-road emissions specifically for the state of Georgia using the EPA NONROAD emissions model. Nonroad engines contribute significantly to anthropogenic emissions inventories, with national estimates for various criteria pollutants ranging from 14% to 22%. The NONROAD model is designed to estimate emissions for any area in the United States based on population, activity, and emissions data. Information used in the model comes from a variety of sources collected over many years. A sensitivity analysis of the model determined the input variables that have significant effects on emissions. Results showed that model estimated emissions are significantly sensitive to increases in equipment population, activity, load factor, and emission factor. Increases in ambient temperature, fuel RVP, fuel sulfur (except on SO2), and average useful life have smaller effects. Emissions and activity data used in the NONROAD model were analyzed using statistical techniques to quantify uncertainty in the input parameters. Expert elicitation was also used to estimate uncertainties in emission factors, equipment population, activity, load factors, and geographic allocations of the emissions to the county level. A Monte Carlo approach using the derived parameter uncertainties and different input probability distributions was used to estimate the overall uncertainty of emissions from the NONROAD model for the state of Georgia. The uncertainties resulting from this analysis were significant, with 95% confidence intervals about the mean ranging from ?? to +61% for THC, -46 to +68% for NOx, -43% to 75% for CO, and ?? to +75% for PM. The sensitivity of ozone and CO for different regions in Georgia to NONROAD emissions in Georgia was also estimated. The analysis suggests that uncertainties in ozone and CO simulations due to NONROAD emissions uncertainties, averaged over the regions of interest, are not large, with resulting maximum coefficients of variation of 1% and 10% respectively.