Advanced Workflows

Advanced workflows are available around Sensor that promise to significantly improve productivity and profitability through assisted history matching, optimization, and uncertainty quantification.  These can be generally described as circular workflows with integrated Sensor pre- and post-processing capabilities, and evaluation, optimization and run control algorithms.  A basic capability and a scalable benefit is the management of serial runs on multiple CPUs to simultaneously evaluate multiple sensitivities.

"The first rule of any technology used in a business is that automation applied to an efficient operation will magnify the efficiency.  The second is that automation applied to an inefficient operation will magnify the inefficiency." - Bill Gates

Energy Components - Integrated Asset Modeling (ECIAM)

ECIAM, formerly called Pipe-it, is a unique software application that allows you to graphically and computationally integrate models and optimize petroleum assets.  ECIAM can launch any software application you are currently running, on any operating system.  ECIAM chains together your applications, in series and in parallel, to create workflows.  Any number of available optimizers can be used to automatically and iteratively adjust any graphically-defined optimization variables to maximize or minimize any objective function that is definable from the workflow outputs.  User-defined (auxiliary) variables and equations provide for great flexibility and generalization in automatic iterative optimization.  With Sensor, we have achieved automatic probabilistic global predictive optimization and forecasting under uncertainty, close to the holy grail of simulation. ECIAM is applicable to any set of applications for deterministic cases (not considering uncertainties in the inputs).  See our Simulation Goals page.

Automated Pipe-it workflows:

SensorMatch - Deterministic Upscaling, and Deterministic or Probabilistic History Matching

SensorCast - Deterministic Predictive Optimization and Forecasting

SensorPcast - Probabilistic Predictive Optimization and Forecasting

SensorPOpt - Probabilistic Alternative Optimization

Probabilistic workflows use our tools SensorPx / Makespx / MakeOptDat along with other programs that are provided with Sensor.

MatchingPro

MatchingPro® is the latest innovative history matching (HM) technology from NITEC that is very easy to use, fully utilizes available computing resources, and provides fast, reliable HM solutions. MatchingPro only requires a reservoir simulation model combined with historical production, injection, and pressure data to accurately and efficiently determine the best HM solution based on the selected HM parameters. No thinning of the historical data is required.

MatchingPro can function in auto or manual modes to achieve a high quality HM solution. In the MatchingPro auto mode, simulation cases are designed, run, and evaluated without user intervention. The process is expedited by utilizing multiple computer CPUs, when available.

MatchingPro’s cutting edge technology makes use of Artificial Neural Networks (ANN), Genetic Algorithms (GA), and statistical methods. These algorithms quickly reveal the complex relationships among the HM parameters and the simulation results (the mismatch in simulated versus historical volumes and pressures). MatchingPro quantifies the quality of each HM case and determines the best combination of HM parameters to achieve the best HM solution. MatchingPro can also find multiple HM solutions (model characterizations) and provide a quantification of their probabilities of occurrence.

MatchingPro can be used with Sensor and other commercial simulators.

The MatchingPro Process: Auto Mode

The observed performance data sets are easily loaded into MatchingPro. This includes: individual well monthly production or injection volumes; static reservoir pressures from build-up tests or shut in wells; and dynamic bottom hole pressures from flowing wells.

The HM parameters and their associated physical ranges are defined via an interactive window. The parameters can be continuous or discrete by defining them as real (linear or logarithmic), or as integer.  

In order to search for the best solution (defined as a set of HM parameters) or multiple solutions, an objective function is defined. The objective function weighs each “mismatch variable” (oil, water, and gas production, injection, and pressure) for use in the minimization calculations. The mismatch can be minimized for selected wells, gathering centers, or the total field. MatchingPro includes a default objective function that is applicable to most situations. The objective function can be modified at any time during the history matching exercise (auto or manual mode) and does not require repeating of the simulation cases or regeneration of the correlation model.

MatchingPro commences a pre-determined iterative process. Simulation cases are created, run, analyzed, and displayed as each case is completed until the entire routine is finished. Once all of the runs are executed, the final plot reveals the best HM solution.

One of the unique features of MatchingPro is the ability to generate multiple history match solutions for the same HM problem. MatchingPro also calculates the probability of occurrence of the set of HM parameters determined for each acceptable HM solution. The minimized objective function value is typically very similar for each solution. However, these different solutions may result in very different performances in the prediction cases. This highlights the well-known non-uniqueness of HM solutions.

Once the MatchingPro process has completed and a satisfactory history match solution has been found, as exhibited by a very low objective function value relative to other simulation runs, the user can utilize MatchingPro’s solution clustering algorithm to determine multiple HM solutions. These solutions may be equally acceptable from the perspective of the objective function (overall HM error), but will have different values for the individual HM parameters.

The user can utilize MatchingPro’s Monte Carlo technology to sample the correlation model, say one million times. From this sample, the user can request the number of multiple HM solutions desired. MatchingPro’s clustering technology groups these solutions accordingly and calculates the probability of occurrence for each.

Each of these solutions will have different HM parameter values, but the objective function value will be very similar indicating an equally acceptable solution.

While the HM parameter values may appear similar, experience indicates that very minor differences in the final HM parameters can result in significant variations in prediction results.

A display of the predicted performance for a well in a large field where three HM solutions were generated shows the predicted performance for the same operating scenario can be quite different for the three equally acceptable, HM solutions.

PlanningPro

PlanningPro^® is an extension of the technologies encompassed in NITEC’s MatchingPro software. It is designed to aid the user in developing optimal drilling locations and schedules. The software utilizes all available computing resources and provides flexibility to the user in defining and scheduling the drilling program.

PlanningPro utilizes a proprietary process to evaluate prospective new well locations, rank the new wells based on cumulative oil, gas or BOE production potential and determines the optimal combination )locations and timing) of new wells. The technology developed in PlanningPro allows a minimal number of simulation cases to be utilized to determine the optimal case. The process is completely automated once the user specifies certain parameters.

PlanningPro can currently be utilized with Sensor and other commercial simulators.

The PlanningPro Process

The prospective new wells to be evaluated and scheduled are first imported or input to PlanningPro. PlanningPro requires only the well name. The simulation data deck (template file) must include these same wells and the typical well information – I, J, K, PI, etc. The simulation deck should also include the typical well, field, gathering center constraints, gathering center identification, well types, etc. There is no limit to the number of prospective new wells that can be input at this time. The user will later be able to select the maximum number of wells to evaluate.

The drilling schedule parameters must be input. These include days to drill each well, completion days per well, number of drilling rigs available and the start date for drilling. Simulation well types must be identified – Liquid producer, Oil producer, Water injector, etc. This should be consistent with the simulation data deck. The user will then define the maximum rate constraint for each of the wells. PlanningPro will write the well rate constraints to the simulation data deck.

Having defined the basic well and simulation parameters, the user is asked to identify the actual new wells to be evaluated. Typically, if the original list of wells input to PlanningPro included 50 wells, all of these wells would be selected. However, this need not be the case. The user can then define the objective function to be based on cumulative oil, gas or BOE production. Economic parameters can also be input in order to calculate a pseudo Net Present Value for each case.

PlanningPro’s automated process first ranks all of the new wells to be evaluated. This is based on a proprietary objective function. This ranking is then used to simulate the model performance starting with the highest ranked well and adding each additional new well through to the lowest ranked well according to the drilling schedule parameters. The user can select all wells previously identified to be evaluated or a lesser number from the ranked list.

The process of ranking the wells is carried out with simulation runs of a relatively short prediction period; shorter than the overall simulation prediction period desired. Three to five years may be simulated in the ranking runs if the full prediction period is 20 years. Once the ranking has been established the full prediction runs are made.

As the process progresses a bar chart is generated to display the objective function value for each simulation case – ranking through scheduling cases. PlanningPro reports the cases being processed and those waiting to process. If economic parameters were input on the Objective Function screen, the “pseudo” economic analysis, discounted net present value is displayed as a Black line on each bar in the figure.

The resulting drilling schedule can be displayed along with the production for each well. PlanningPro allows the user to display production rate and cumulative plots for any of the simulated cases and compare one case to another. Results can also be exported to .csv files for use in other applications.

ForecastingPro

ForecastingPro^® is the latest technology from NITEC for evaluation of uncertainty associated with prediction case scenarios in reservoir simulation. ForecastingPro uses defined uncertainties in reservoir and operating parameters and simulation model runs to evaluate the probability of the performance results. The results are displayed as rate and cumulative production and injection profiles as a function of time for P10 through P90.

ForecastingPro utilizes proprietary technology to assess the impact on forecasted field performance of an unlimited number of user defined reservoir and operating parameters which are varied over user specified ranges. The evaluation can be initiated at time zero (a Greenfield) or from history match run results from MatchingPro with their associated probabilities or from history match runs that have been made independent of MatchingPro. The process is completely automated once the user specifies the required parameters.

ForecastingPro can currently interface with the Sensor and other commercial simulators. Only Eclipse currently allows changes to reservoir parameters on a restart run.

The ForecastingPro Process

ForecastingPro needs very little information about the particulars of the prediction case or the reservoir being evaluated. Reservoir parameters and well and field data are only provided in the simulation data deck and need not be imported into ForecastingPro. The simulation data deck is treated as a template file which contains the user defined variable names for the uncertain parameters used in the analysis. These parameters must be identified along with the range over which they can vary. If this is a prediction study from an existing history match model, these parameters are typically those which do not have an impact on the history match period, but may have an impact on the predictions.

The user can identify a large number of uncertain reservoir and operating parameters, but choose to vary only a few in the analysis; others can be set to constant values. This provides the user with flexibility during the analysis process.

ForecastingPro’s automated process uses distributed computing, hence it can take advantage of cluster servers and multiple CPUs to speed processing of the simulation runs required in the analysis. Once the user has identified the uncertain parameters to use in the analysis, the software determines the number of simulation runs that will be required to achieve reliable results. For more than five variables, testing has indicated that the number of simulation cases needed is 6 to 8 times the number of uncertain reservoir parameters being evaluated. This is significantly fewer simulation runs than required by other software that rely on Latin Hypercube search methods for experimental design.

Proprietary technology is used to develop an accurate response surface of the simulated field performance of oil, water, and gas production profiles, as well as gas and water injection profiles. Initial experimental design (scoping) runs are followed by a series of simulation (investigation) runs that sequentially improve the ability of the response surface to predict performance from any given set of uncertain parameters. By default, all simulation prediction constraints are honored in all simulation runs, hence the predicted performance is only impacted by the perturbed parameters.

Once the auto process has completed the user can view the difference between the predicted performance of the response surface and the actual simulation run results for each run in a simple bar chart.

At this point the response surface has been calibrated to the actual simulation runs. To assess the performance profile for any combination of parameters and the associated probability, Monte Carlo analysis is used. The user can select any combination of the parameters which have been varied or can specify that some should have a specific value. The number of Monte Carlo samples is input and the calculations are made. A sample of 50,000 has been found to be generally satisfactory, but a larger sample can be specified. This typically takes 30 seconds using a generic Windows personal computer.

Once the calculations have been made a cumulative performance profile (oil, water, gas production and gas, water injection) is available for display. P10 through P90 results are shown at 10 percent increments. Rate profiles are also available.

The default frequency for the time interval used in the Monte Carlo analysis is annual. However, the user can select less frequent time periods to speed the analysis. Statistics in the form of distribution and Tornado charts can be displayed if desired. The user can then evaluate the statistics associated with any probability result at any of the specified times during the prediction.