There are several models for predicting US natural gas production. The most referenced model is the one developed and used by the Energy Information Administration (EIA). Recently a model was developed at Texas A & M University that used neural networks in order to learn from the past when predicting the future. One of the major short comings of the existing models is the fact that they appear to be too certain about their predictions. In other words, they fail to take into account the uncertainties associated with predicting many of the parameters that are inputs into their models and treat them as a certain single values.

In this paper authors present a new model with several advantages over the existing models. First, this model treats the US Natural Gas Production as a time series with all its inherent complexities and incorporates the required methodologies during the development process. Second, this model does not use parameters such as gas price as an input. If we know the future price of gas, then predicting the production levels will not be so difficult. Third, this model does not assume that crisp and exact information is available on the future values of all the input parameters such as GDP, Population and average depth of oil and gas wells. Instead by incorporating a Monte Carlo simulation approach, it uses a probability distribution function for each of its input parameters, and hence the outcome of the model (US Natural Gas Production) is not a crisp number for each year rather it is a range that includes a minimum, a maximum and a most likely value.

The model is developed using the state-of-the-art in intelligent systems by using data up to year 1998. The model is then validated by comparing its results with actual US Natural Gas production for years 1998-2002 as blind (verification) years. During this five-year period this model outperforms all the existing models in the literature. The US Natural Gas production is then forecasted until year 2020.

You can access this article if you purchase or spend a download.