The oil and gas industry is highly risky capital-intensive field of business. Many companies are working hard to perform projects on a global scale in order to get, produce and deliver their final products. No matter the economic conditions, it is vital for organizations to efficiently manage their capitals projects, which facilitates to control expenditure, handle priorities and mineral resources, and make assets productive as quickly as possible. It is also critical to efficiently and safely maintain and improve these assets. Probably the most volatile item of the project cost in oil and gas industry is the market price of resources or assets. Both sides (stakeholders) seek for efficient profitable price for selling and buying final product. This paper provides the description of application developed for fair oil price estimation using Fuzzy Sets and Logic approach of Soft Computing and FRIL inference language with its environment installed on UNIX virtual machine.
Managing the complexity and profit of major projects in
today’s oil and gas landscape has never been more critical.
Against the backdrop of a decline in both global economic
conditions and corporate revenues, stakeholders are
demanding improved return on investment (ROI), reduced
risk and exposure and greater transparency. Since capital
construction projects in the upstream oil and gas industry
comprise a significant percentage of company spend, there
must be a particular focus on predictability, transparency
and reliability, including estimation of profit, controlling
and reducing the costs associated with these projects. The
best opportunity to make a positive impact on the life cycle
of capital project in this industry is during early planning,
even before the capital outlay occurs. In order to control
planning it is useful to develop an integrated cost
management function that aligns all cost-related processes
and functions and incorporates data developed or
maintained in other processes. Emphasis should be placed
on budget control, approved corporate budget changes and
project management internal budget transfers.[
In the direction of application of fuzzy logic to similar
economic problems, the following research was made: the
problem of developing automated system of technical –
economic estimation in oil and gas fields was considered
by Yu.G. Bogatkina [
No study was made in Kazakhstan connected with problems of uncertainty in oil prices and project costs in terms of the principles of fuzzy logic, which could give a more complete picture of price changes and their influence on the general economic situation in the country, which allows forecasting of rising inflation and determining the most optimal range of prices taking into account various economic factors.
The given application offers solution to the problem of the project profit estimation considering the price of oil as a fuzzy set. Such types of applications are developed for project managers in oil and gas industry to make evaluation of project profit for the future decision. Also all investors and financial institutions of this industry could be interested in using the given tool.
The total profit of the project could be generally expressed as follows:
P= Oil_price * Supply_scope
According to the research [
Price=Brent_Price (or Urals_Price) +/
Market_Differential Key benchmarks ($/bbl)
Data code PCAAT00 PCAAU00 PCAAV00 AAWSA00 AAWSB00 AAWSC00 AAJMS00 PCAAS00
AAOFD00 Dubai (SEP) Dubai (OCT) Dubai (NOV) MEC (SEP) MEC (OCT) MEC (NOV) Brent/Dubai Brent (Dated) Dated North Sea Light Brent (AUG) Brent (SEP) Brent (OCT) Sulfur Deescalator WTI (AUG) WTI (SEP) WTI (OCT) ACM AUG)* ACM (SEP)* ACM (OCT)*
PCAAP00 PCAAQ00 PCAAR00 AAUXL00 PCACG00 PCACH00 AAGIT00 AAQHN00 AAQHO00 AAQHP00
Usually traders make an agreement for the value of market differential. Market price construction for Kazakhstani oil depends on CIF August’s (CIF stands for cost, insurance, freight) market differential.
The main idea is to consider these two parameters (the price of Brent dtd. oil and market differential) as fuzzy sets, because the former changes every day, and the second is a result of contract between traders.
The research is based on the theory of fuzzy sets and
logic. Fuzzy sets were initially offered by Lotfi A. Zadeh
[
Using fuzzy sets in estimation of oil price gives opportunity to remove straight principles of crisp sets. For example, in the case of crisp sets we should take only three exact numbers for oil price and market differential and calculate assessment for the worst, standard, and the best cases. In opposite, with the help of fuzzy sets it is possible to take into account the variation of price in time, in other words statistical or historical changes. Another significance of fuzzy sets is in possibility to manage uncertainty. It means that we can more precisely define which numbers that represent the price of oil can be called normal or higher/lower than that and with which membership degree.
Concerning all of conditions described above three universal sets can be defined:
UX1 = [80, 140] – price of Brent dtd.
UX2 = [
UY = [50, 160] – oil price
The approximate values of borders of sets are taken from
statistical values of mentioned parameters for the previous
year [
Also fuzzy sets on inputs X1 and X2 and output Y were created with membership values as follows:
Fuzzy sets on X1(for the price of Brent dtd. see fig.1): ‘less than usual’ = {1/80, 0.8/90, 0.7/100, 0.5/110, 0.2/113, 0.1/115, 0/118, 0/120, 0/130, 0/140} ‘more than usual’ = {0/80, 0/90, 0/100, 0/110,0.2/113,0.4/115,0.5/118,0.7/120,0.9/130, 1/140}
The given sets were constructed following principle: as
the price for Brent dtd. has changed between 80 and 140
(according to statistics [
($/barrel)
Fuzzy sets on X2 (for the value of market differential see fig.3) ‘significantly high’={1/-5, 0.8/-4, 0.5/-3, 0.2/-2, 0/-1, 0/0, 0.1/0.1, 0.2/1, 0.4/2, 0.6/3, 0.9/4, 1/5} ‘significantly low’={0/-5, 0.2/-4, 0.4/-3,0.7/-2, 0.8/-1, 1/0, 0.9/0.1, 0.8/1, 0.6/2, 0.3/3, 0.1/4, 0/5} The most frequently used market differentials vary from -5 to 5 as was mentioned in the list of universal sets for three parameters. The higher the value of market differential (without sign) the more significant influence it has to the final price of resource, so that the members -5 and 5 have the highest membership degree in the set ‘significantly high’. Other members of this universal set, which are close to 0, have higher degree in the set ‘significantly low’ respectively.
Finally, the resulting universal set for oil price was divided into three approximate subsets: cheap, normal or optimal, and expensive. The membership degrees of the elements were assigned following the same principle described above.
Fuzzy sets on Y (for the resulting value of oil price see fig.4): ‘cheap’={1/50, 0.9/60, 0.8/70, 0.7/80, 0.6/90, 0.4/100, 0.2/110,0/115, 0/120,0/140,0/160}
The given application was developed using FRIL. With the help of FRIL method of inference number of cases can be calculated.
FRIL was initially an acronym for "Fuzzy Relational Inference Language", which was the predecessor of the current FRIL language, which was developed in the late 1970's following Jim Baldwin's work on fuzzy relations.
FRIL is an uncertainty logic programming language which not only comprises Prolog as one part of the language, but also allows working with probabilistic uncertainties and fuzzy sets as well.
The theory of mass assignments developed by J.F. Baldwin in the Artificial Intelligence group of the Department became a foundation to FRIL language. A fuzzy set is defined as a possibility distribution which is equivalent to a family of probability distributions. FRIL supports both continuous and discrete fuzzy sets.
FRIL gives opportunity to express uncertainty in data,
facts and rules with the help of fuzzy sets and support
pairs. There are three special types of uncertainty rule
besides the Prolog rule. They are: the basic rule, the
extended rule, and the evidential logic rule. The second,
extended, rule is widely used for causal net type
applications, and the evidential logic rule is appropriate to
case-based and analogical reasoning. Every rule can have
related conditional support pairs, and the method of
inference from these rules is based on Jeffrey's rule which
is connected with the theorem of total probability. Fuzzy
sets can be used to represent semantic terms in FRIL
clauses. In addition, a process of partial matching of fuzzy
terms like this, which is called semantic unification, gives
support for FRIL goals [
For instance, if the price of Brent dtd equals to 120 (which belong to the set ‘more than usual’) and CIF market differential is -3 the application outputs defuzzyfied value equal to 116.199. If we apply initially described formula to this data, we will get 120 – 3 = 117. So, the result, which was received by fuzzy sets application, is close to the output of calculations following formula. However, there is a difference in 1 dollar per barrel that can considerably affect the final project cost. More results for the values of the Brent dtd. price and market differential from different sets are listed in appendix below.
The application gives the fair price to the sellers’ side, so that the price by contract, that is lower than this, is not profitable, and the price that higher than this, is not competitive on the market. On the other side, buyer takes the reasonable price for the product, which correspond reality with taking into account situation in the market.
Finally, the project profit can be obtained by multiplication of output value to whatever value of the supply scope following the volume in the contract.
As a result let us notice that Fuzzy Sets and Logic can be successfully applied to estimate project profit by assuming several parameters as fuzzy sets, constructing those sets and applying fuzzy logic to them in order to reduce uncertainty. There are still many opportunities to improve the precision of calculations.
The market price of Kazakhstani oil depends on several
fundamental factors, which can be divided into fixed and
variable. Variable factors, such as the level of
consumption of petroleum and petroleum products in a
specific period of time, the amount of energy resource
available on the market, conditions of delivery, the number
of traders, significantly affect on the fluctuation of oil
prices. Moreover, the quality of exported oil (density,
content of sulfur and wax, etc.), maintenance of quality,
stable production and supply, and the cost of oil production
in a particular region have an impact on market price [
Yet another example, oil price in Kazakhstan also depends on dollar variation. So, additional fuzzy set representing the value of dollar in Kazakhstani tenge can be added and application will calculate it without any problems. In addition, the application can be easily customized to calculate the price of gas and prices of other mineral resources.
Source Code %% oil-price.frl %% NOTE: we use FRIL method of inference %% (different from Mamudani, Sugeno,etc.) % estimate the price of oil depending on Brent dtd. and market differential % INPUTS: price of Brent, market differential % OUTPUT: oil price %% universe of diccourse set (dom-brent 80 140) set (dom-market_dif -5 5) set (dom-oil_price 50 160) %% fuzzy sets on Brent dtd. (less_than_usual [80:1 90:0.8 100:0.7 110:0.5 113:0.2 115:0.1 118:0 120:0 130:0 140:0] dom-brent) (more_than_usual [80:0 90:0 100:0 110:0 113:0.2 115:0.4 118:0.5 120:0.7 130:0.9 140:1] dom-brent) %% fuzzy sets on market differential (significantly_high [-5:1 -4:0.8 -3:0.5 -2:0.2 -1:0 0:0 0.1:0.1 1:0.2 2:0.4 3:0.6 4:0.9 5:1] dom-market_dif) (significantly_low [-5:0 -4:0.2 -3:0.4 -2:0.7 -1:0.8 0:1 0.1:0.9 1:0.8 2:0.6 3:0.3 4:0.1 5:0 ] dom-market_dif) %% fuzzy sets on oil price (cheap [50:1 60:0.9 70:0.8 80:0.7 90:0.6 100:0.4 110:0.2 115:0 120:0 140:0 160:0] dom-oil_price) (normal [50:0 60:0 70:0.1 80:0.3 90:0.5 100:0.7 110:0.9 115:1 120:0.5 140:0 160:0] dom-oil_price) (expensive [50:0 60:0 70:0 80:0 90:0 100:0 110:0 115:0 120:0.7 140:0.9 160:1] dom-oil_price) %% Fuzzy Associative Matrix (FAM) % % b\d | L | H | % -------------% L | C | N | % M | N | E | % %% fuzzy rules based on FAM ((price cheap)
(brent significantly_low)) ((price normal)
(brent significantly_high)) ((price normal)
(brent significantly_low)) ((price expensive)
(brent significantly_high)) more_than_usual)(market_dif more_than_usual)(market_dif ((simulation-v B D) (addcl ((brent B))) (addcl ((market_dif D))) (p 'Price of Brent:' B ', ' 'Market differetial:' D)(pp) (qsv ((price X))) (delcl ((brent B))) (delcl ((market_dif D)))) %% eof %%
Price of Brent: 80 , Market differetial: 1 ((price 76.9223)) : (1 1) Fuzzy set [49.9978:0 50:0.8 60:0.8 70:0.82 71.6666:0.826667 90:0.68 110:0.36 115:0.2 ] defuzzifies to 76.9223 over sub-domain (50 140) no (more) solutions yes Fril >?((simulation-v 120 -3)) Price of Brent: 120 , Market differetial: -3 ((price 116.199)) : (1 1) Fuzzy set [115:0.65 119.545:0.872727 120:0.86 140:0.72 160:0.72 160.011:0] defuzzifies to 116.199 over sub-domain (60 160) less_than_usual)(market_dif less_than_usual)(market_dif no (more) solutions no (more) solutions yes
Price of Brent: 90 , Market differetial: -1 yes
Price of Brent: 140 , Market differetial: 5 ((price 142.216)) : (1 1) Fuzzy set [115:0 120:0.7 140:0.9 160:1 160.011:0] defuzzifies to 142.216 over sub-domain (115 160) no (more) solutions yes