Artificial Intelligence as a Price Forecaster in Oil & Gas

Artificial Intelligence as a Price Forecaster in Oil & Gas

Artificial Intelligence as a Price Forecaster in Oil & Gas

In a paper written four years ago, nine scholars led by Haruna Chiroma, of the Federal College of Education (Technical) in Gombe, Nigeria, undertook to synthesize the research then available on “wavelet analysis and computational intelligence techniques … in the domain of crude oil price forecasting.”  In short, they were looking at AI as a way of predicting O & G prices, in the hope that better forecasting in turn would “create stability in the oil market.”

AI continues to exercise a hypnotic sway over observers and participants in the O&G markets.  Just in recent weeks, the editor of Oil & Gas IQ has suggested that China’s superiority in AI has given the PRC the potential to “disrupt the global energy market enough to undermine major exporters’ position.”  The hypothesis is that China is improving the efficiency of its operations both in the oil fields and in the refineries: it is shrinking oil well downtime, lowering labor intensivity, etc. Since China is “a major world energy consumer,” its AI improvements could make it more self-sufficient, reducing its imports and this other nation’s exports in a way that would “shake up the traditional … hierarchy.”

Also, a contributor to Forbes has detailed how Royal Dutch Shell in particular is using an AI technique called “reinforcement learning” (think of Skinner’s behavioral techniques applied to machines instead of pigeons) to improve its performance at every link in the “entire oil and gas supply chain,” from the withdrawal of raw hydrocarbons out of the earth to retail sales.

 

AI as a Price Forecaster in O & G 

So let us take a look at AI in O&G, starting with a retrospective look at that review article by Chiroma et al. It focused on the issue of price forecasting, because volatility in this effect is a major source of economic and political uncertainty in many countries, exporting or importing. Early on, the authors take note of the development of “fuzzy logic” as a crucial piece of this puzzle. Fuzzy logic, they explain, consists of “methods that are composed of cognitive human principles” whereby “imprecise data and vague statements are accepted as inputs and a decision is yielded.”

Fuzzy logic includes “fuzzy regression … a modification of the conventional regression model used to predict dependent variables dependent on independent variables.” The conventional regression model breaks down in circumstances of inadequate or vague data sets. Fuzzy regression can plough onward, as can the hardware in which it is instantiated.

After going over those and other fundamentals with regard to contemporary computing, the article turns to the other half of its equation, the O&G industry. There has been a lot of effort put into the creation of forecasting models.  Some of them have used Box-Jenkins models, that is, models in which an autoregressive moving average is applied to find the best fit to actual historical prices. In a 2005 paper, Shouyang et al integrated the Box-Jenkins model into a hybrid intelligence system (HIS) in order “to model linear constituents of the crude oil price time series.”

It is in such contexts that “wavelets” entered the discussion. Wavelets are mathematical functions that cut data into different frequency components, allowing for the study each component with a resolution matched to its scale. A 2012 paper by Jammazi and Aloui used wavelets to incorporate crude oil price data into a neural network in order to forecast crude oil prices.

 

Can Everybody Do It?

To cut a long story short: a good deal of progress has been made. Chiroma et al anticipated more. But this raises another issue. What is the upshot as such AI programs become more generally prevalent in the markets for O&G, both the physical market and the derivatives markets? For example: if everybody were using the same sophisticated AI programs, would that lessen price volatility or increase it, by allowing for a disastrous feedback loop?

One safe inference is that the traders of oil and gas are finding themselves forced into an expensive arms race. Whatever the consequence might be when everyone does it, the consequence is surely disastrous for the Smith Crude Trading Co. if the SCTC is the only entity that isn’t doing it.

A more recent survey indicates that the use of AI in the oil and gas market, not specifically for trading or price forecasting but throughout the industry, will be a market worth $2.85 billion by 2022. That combines “the adoption of the big data technology” with “digitization of the Oil & Gas industry, investments in AI related startups, and rising pressure to reduce production costs and increase efficiency.”

 

GET THE LATEST STORIES AND TIPS FROM CATERPILLAR – SIGN UP FOR OUR E-NEWSLETTER

 


Beyond the Iron

Beyond the Iron

Journey with us beyond our iron as we explore trending topics in oil & gas.

Learn more

Cat Oil & Gas Ratings Guide

Cat Oil & Gas Ratings Guide

Get the detailed specifications of Cat oil & gas solutions needed for your projects right at your fingertips.

Download Now