Petroleum Engineering

Highly specialised requests are first checked for feasibility from this information. Please do not send confidential data through the public form; describe it first and agree secure transfer afterwards.

1. 01Enquiry and requirement assessment
2. 02Technical scoping and proposal
3. 03Execution
4. 04Quality review and delivery, with an agreed round of revisions

Read the full project workflow.

In depth: how the work is done

1. Decline-curve and production-data analysis

   Decline-curve analysis fits exponential, hyperbolic or more flexible models to production rates to forecast future output and estimate recoverable volumes. The fitted parameters depend on the period used, flow regime and operating changes, so forecasts can differ widely. Rate-transient analysis uses pressure and rate data to infer reservoir properties. We emphasise out-of-sample testing, in which the model is fitted to early data and compared with later production.

2. Reservoir-modelling concepts

   Reservoir simulation solves multiphase flow in porous media on a grid populated with porosity, permeability and fluid properties from geological and petrophysical models. History matching adjusts parameters until simulated production matches observations. Many parameter sets can achieve a match, so the match alone does not validate the forecast. We discuss grid resolution, upscaling and the effect of geological uncertainty.

3. Data-driven prediction

   Machine learning is applied to predict production, estimate properties from logs or identify patterns in drilling data. With few wells, overfitting is a concern, and spatial correlation between wells can leak information across random splits. Validation by well or by field block, feature importance analysis with caution and comparison with simple baselines are expected. Physics-informed approaches can improve credibility.

4. Uncertainty quantification

   Uncertainty in reservoir performance arises from geology, fluid properties, well performance and economics. Ensembles of models, experimental design to explore parameters and proxy models speed up the analysis. Outputs are distributions of outcomes, such as P10, P50 and P90 estimates, with a clear statement of how they were derived.

5. Drilling and well-performance data

   Drilling data such as rate of penetration, torque and mud properties support analysis of performance and events. Quality control for sensor errors and depth alignment is essential. We support statistical and data-driven analysis, treating conclusions as research findings, not operational advice.

6. Energy transition applications

   Reservoir methods are used for geothermal systems, hydrogen storage and CO2 storage, where long-term containment and pressure response matter. Such studies need careful treatment of geomechanics, chemistry and monitoring data. Results are presented with uncertainty and appropriate caution about leakage and induced-seismicity risks, which require specialised assessment.

7. Data quality in production and well-test records

   Production records may contain allocation errors, shut-in periods, changes of choke or lift method and unreported workovers. Before fitting, we mark such events and analyse segments separately or model them explicitly. Pressure gauges drift, and well-test interpretations depend on assumed models. Documenting data corrections and the fraction of data excluded protects the credibility of any forecast built on the records.

8. Economics, risk and decision framing

   Technical results are often used in economic evaluations, where prices, costs and timing add more uncertainty. Research papers can show how the distribution of production outcomes feeds into value distributions without making investment recommendations. Sensitivity to price and cost assumptions is shown with clear labelling of the scenarios, so readers do not mistake an illustration for a forecast of project value.

9. Reproducible workflows for subsurface analysis

   Subsurface studies involve many software tools and file formats. Recording versions, input decks and random seeds, and scripting the analysis from raw data to figures, allows results to be reproduced and updated when new wells come online. Sharing de-identified data and code where licences permit lets others test the methods on their own fields, which is the best protection against methods that work only on one dataset.

Common pitfalls this service helps avoid

• Fitting decline curves to unstable periods. Operating changes and transient flow produce misleading parameters.
• Equating a history match with a validated forecast. Many models match the same history.
• Random train-test splits across wells. Spatial correlation inflates apparent accuracy.
• Reporting a single forecast. The outcome is uncertain and should be shown as a distribution.

What good reporting looks like

A petroleum engineering report describes the data and their quality, the decline or reservoir models and fitting method, the validation approach, the uncertainty analysis, the data-driven model evaluation and the limitations. It does not provide reserves estimates, field-development recommendations or drilling programmes, which require qualified professionals and applicable reporting standards.

An illustrative scenario

Hypothetical example written for explanation. It is not a client case, a result or a testimonial.

Imagine a researcher forecasting shale-well production with several decline models. Fitting to the first year of data gives widely different cumulative forecasts for different models. Testing on wells with longer histories shows that a more flexible model forecasts better on average but with large scatter. The report presents forecasts as ranges, notes that completions differed among wells and avoids stating a single recoverable volume.

Questions to settle before you begin

• Is the production period I fit stable and representative?
• How will I test the forecast out of sample?
• Which uncertainties dominate outcomes?
• Does my validation separate wells and areas?
• What does my model not represent?

Key terms

Decline curve

A fitted curve of production rate against time.

History matching

Adjusting a model to reproduce observed production.

P10/P50/P90

Percentile values from a distribution of outcomes.

Permeability

A measure of rock's ability to transmit fluid.

Rate-transient analysis

Analysis of rate and pressure to infer reservoir properties.

Working together on Specialised Applied Sciences projects

Specialised applied-science engagements start with the field-specific question, data and methods, and with an honest check of whether the necessary domain expertise is available for your topic. We ask for protocols, datasets, instrument descriptions and relevant literature. Where a request is highly specialised, feasibility is confirmed before a proposal is issued. Analyses are documented and reproducible, results are reported with uncertainty, and any interpretation beyond the data is labelled. The service does not provide clinical, regulatory or safety advice.

Few projects need a single service in isolation. These are the services most often combined with this one.

• AI & Machine Learning for Engineering

  Machine-learning methods for engineering problems with physical data, designed around data quality, leakage-free validation and a clear account of when the model should not be trusted.

• Time-Series Analysis & Forecasting

  Analysis of sequential engineering data such as loads, flows, temperatures or production signals, with decomposition, modelling and forecasting validated on held-out time periods.

• Geology & Geophysics

  Analysis and interpretation support for geological and geophysical datasets, including seismic, gravity, magnetic and electrical methods, with attention to processing steps and non-uniqueness.

• Multiphase Flow Simulation

  Simulation of flows with more than one phase, such as gas-liquid, liquid-liquid and particle-laden systems, where model choice (Eulerian, Lagrangian, interface tracking) strongly shapes the result.