Forecasting

In general, in order to forecast, you need to calibrate your model by history matching. This helps you get model parameters that allow you to estimate / forecast future performance over time.

History Matching

History matching is the process of creating and modifying a model until it mimics the known past behaviour of a reservoir. The purpose of history matching is to develop a model from which to base predictions of future behaviour.

The three steps in a typical history match are:

1. Select a model. The type of model used can range from a simple empirical fit of the data to a rigorous reservoir multi-parameter model.

2. Set Appropriate Ranges for the model parameters. It is extremely important to set appropriate ranges for all model parameters based on interpretations from other measurement and testing sources. Model matches that require some variables to be manipulated far outside their appropriate range may “look nice” but invariably lead to poor forecasts of future scenarios.

3. Determine final match. The model parameters are manipulated so that the model matches known measured values within an acceptable range. It is extremely important that the acceptable range for the solution be determined prior to the matching process and that range be adhered to throughout the process. Matching measured data too finely always leads to inappropriate manipulation of model parameters or selection of an inappropriate model. Although this process can be automated, it is often a trial and error process. Once the model values bear an acceptable resemblance to the measured values, the history match is complete. What is acceptable will vary depending on the quality of the measured data and the model used.

It is very important to understand the limitations of the model selected for matching. Some models can be very simple and so are not capable of modeling changes in operating conditions, such as decline analysis. In the case of decline analysis, it is assumed that the user has reviewed the data and determined which portions of the dataset are appropriate for extrapolation.

Forecasting

Forecasting is the process of estimating an entity’s (well, reservoir, field, etc) future performance as a function of time. Depending upon the model employed, there may be options for making changes to how certain variables may change in the future that could have an impact on the rate forecast. For example, a decline analysis assumes that operating conditions such as line pressure are constant. If line pressure can be expected to change in the future, a forecast from a decline analysis would not be able to predict the impact on the rate forecast. On the other hand, if a gathering system model tied to a reservoir deliverability model were employed, then a forecast could include the impact of expected changes to the line pressure in its rate forecast.

Creating the forecast is often as simple as extrapolating the history match model to a desired time. For decline analysis, an empirical fit developed by Arps is generally used to create a forecast. An endpoint for a forecast is usually based on economic or operational constraints and in some cases by multiple criteria.