Prediction Of Gasoline Yield Of Crude Oil Refinery Using Long-Short-Term-Memory Time Series

Electrical Electronics Engineering | Engineering

This study investigates the prediction of gasoline yield in a crude oil refinery using Long-Short-Term-Memory (LSTM) time series modeling. Gasoline yield prediction is crucial for refinery operations and optimization. The research employs LSTM, a type of recurrent neural network (RNN) known for its ability to capture long-term dependencies in sequential data. Historical refinery data, including feedstock properties, process conditions, and operational parameters, are used to train and validate the LSTM model. Key factors such as feedstock composition, processing techniques, and market demand are integrated into the modeling process. The results demonstrate the effectiveness of LSTM in accurately forecasting gasoline yield, enabling refineries to make informed decisions for production planning and resource allocation. This study contributes to the advancement of predictive modeling in the petroleum industry and underscores the importance of leveraging advanced data-driven techniques for optimizing refinery operations and meeting market demands.

ABSTRACT

Crude oil and petroleum products are among the critical inputs of industrial production and have an essential role in logistics and transportation. Hence, sudden increases and decreases in oil prices cause particular problems in global economies and thus, they have a direct or indirect effect on economies. Furthermore, due to crises in developing economies, trade disputes between major economies, and the dynamic nature of the oil price effect on demand and supply for oil and petroleum products, and time to time volatility in the oil price are very severe. The uncertainty in oil prices can leave both consumers and producers with heavy potential losses. Due to this rapid variability, predicting oil prices has global importance. In this study, to increase the accuracy and stability, the Long-Short Term Memory (LSTM) was applied to foresee future tendencies in Brent oil prices considering their previous prices. The coefficient of determination (R2) for the LSTM was found as 0.92 in the training stage, and 0.89 in the testing stage, respectively. According to the results obtained, the LSTM model has good results to predict the trend of oil prices.

Key words: Crude oil, crude oil yield, yield correlations, gasoline yield of crude, oil refinery, long-short-term-memory time series.

TABLE OF CONTENTS

TITLE PAGE

APPROVAL PAGE

DEDICATION

ACKNOWLEDGEMENT

ABSTRACT

TABLE OF CONTENT

CHAPTER ONE

1.0 INTRODUCTION

1.1 BACKGROUND OF THE PROJECT

1.2 PROBLEM STATEMENT

1.3 AIM AND OBJECTIVES OF THE PROJECT

1.4 SIGNIFICANCE OF THE PROJECT

1.5 SCOPE OF THE PROJECT

CHAPTER TWO

2.0 LITERATURE REVIEW

2.1 OIL REFINERY OVERVIEW

2.2 OPERATION OIL REFINARY

2.3 MAJOR PRODUCT OF CRUDE OIL

2.4 REVIEW OF RELATED STUDIES

2.5 REVIEW OF CRUDE OIL PRICE PREDICTING TECHNIQUES

CHAPTER THREE

3.1 MATERIALS AAND METHOD

CHAPTER FOUR

RESULT ANALYSIS

4.1 RESULTS AND DISCUSSION

CHAPTER FIVE

5.0 CONCLUSION AND RECOMMENDATION

5.1 CONCLUSION

5.2 RECOMMENDATION

REFERENCES

CHAPTER ONE

1.0 INTRODUCTION

1.1 BACKGROUND OF THE STUDY

Crude oil comes in diverse types depending on how it was formed. After the extraction of crude oil from producing fields, it is transported to the processing units usually via oil tankers or pipelines, with pipelines being the safest and economical means of transportation (Marfo et al., 2018). On average, hundreds of different crude oil are transformed in larger or smaller quantities in the world’s refineries (MathPro, 2011). There are more than 660 refineries operating in about 116 countries that generate over 85 million barrels of refined products each day (Haghi and Torrens, 2018). The physico-chemical properties of crude oil determine the level of yields of petroleum products in a refinery. The recoveries, yields, and properties of disparate fractions are the vital variables assessed in a laboratory distillation method for crude oil (Michael et al., 2009). This information is crucial to the refiner in knowing the processes that would be required to generate their desired product yields and help in monetary evaluation in profit margins on these products yet to be obtained. Petroleum yields are physically refined primarily by two methods; ASTM D86 and D1160 with detailed information about the methods found in (Michael et al., 2009; Perumal, 2014). At present, the ASTM processes have been substituted by gas chromatography simulation of distillation which offers more control in the operating conditions of a refinery and product specification (Michael et al., 2009).

Petroleum refineries are large, capital-intensive industrial facilities with multifaceted processing systems. Their complexities depend on a fixed configuration that yields defined output reliant on the crude inputs and refinery’s capacity (Morsali, 2017). All refineries have distinctive physical configuration, operating characteristics and economics (Gary et al., 2007). A refinery configuration and performance characteristics are determined mainly by its location, funds availability for capital investments, accessible crude oil, product demands, product quality stipulations, environmental rules and standards, and market conditions and specifications of the refined products (Parkash, 2003; MathPro, 2011). The principal economic goal of a refinery is to maximise the value added in transforming crude oil into end products (Ali and Rashid, 2013). Typical products obtained from refinery processes are Liquefied Petroleum Gases (LPG), gasoline, aviation turbine kerosene, diesel, petrochemical feedstocks, lubricating oils and waxes, heating and fuel oil and asphalt (Jones and Pujado, 2006; Morsali, 2017). Crude oil varies significantly with respect to their API gravity, component mixture and level of metals, sulphur and various salts. Crude oil assay descriptions of different crude oil with a distillation range are to indicate the number of components in each boiling point range. This information is used to determine the relative flow out of the atmospheric crude oil fractionating column (Fahim et al., 2009).

The complex refinery processes makes it nearly impossible to develop a thorough model for them. This is because their modelling usually requires the application of mass and heat transfer, fluid mechanics, thermodynamics, and kinetics (Al-Enezi and Elkamel, 2000). The outcome is a system of nonlinear, together with algebraic and differential equation systems. Many equations are typically required for their description and many variables have to be estimated. Nevertheless, refinery processes are essentially flexible processes that can take a variety of feedstock and run at several conditions in order to reach product demand targets (Al- Qahtani and Elkamel, 2010). Increasing feed weight, stringent demands on product qualities and environmental regulations, make it crucial to develop models that can optimise the several processes, and most importantly predict the yields of the refined products during refinery operations (Fisher, 1990; Al-Enezi and Elkamel, 2000). Numerous refinery research scientists have developed kinetic models (Farag and Tsai, 1987) and correlations for predicting yields of crude oil using the fluid catalytic cracking (FCC) processes and taking into account American Petroleum Institute (ºAPI) gravity of the feed and the conversion alone (Gary and Handwerk, 1978); Volumetric Average Boiling Point (VABP), specific gravity, aniline point and feed sulphur content (Castiglioni, 1983), molecular description for catalytic cracking of vacuum gas oil (Pitault et al., 1994). However, these models gave quite unsatisfactory predictions.

Based on the models shortcomings, Maples (2000) developed new correlations for predicting product yields of FCC process independent of variables including Watson characterisation factor K of the feed, the percent conversion, the feed ºAPI as well as the sulphur content of feed of which gave consistently better predictions. These mathematical correlations have become integral to the refinery processes as refiners have the opportunity to estimate the yield before the actual refining process commences (Ancheyta-Juarez and Murillo-Hernandez, 2000).

Before any refinery goes into production, simulation is done on the type of crude oil about to be fractionated as it gives them the fore knowledge about the yield range of every crude oil type based on the assay. This helps in both the economic aspect of the refinery in obtaining margin profitability of the yields to expect as well as helping refiners on what operating conditions to prepare for to obtain maximum production. The Tema Oil Refinery (TOR) simulates crude oil using the yield pattern model which has been helpful over the years in predicting their yields until December 20, 2016 where they refined ten million barrels of crude oil from the Tweneboa, Enyenra and Ntomme (TEN) fields without yielding Aviation Turbine Kerosene (ATK) (Anon, 2017). Having discovered the flaw in the previous yield model used in predicting the crude oil yields, it has become very necessary to combine other existing yield pattern models together with the yield pattern model in the refinery and then design a friendly-user software application that can simulate the crude oil yield before the actual crude oil refining commences. This paper, therefore, presents long short term memory application to predict the crude oil yields based on their assay to increase the level of certainty in yields predictions and also determine the most sensitive crude oil property to the production of maximum gasoline.

1.2 PROBLEM STATEMENT

All countries, even producers, are petroleum and petroleum products consumers, so the price of petroleum depends on economic activities around the world. The cost of other goods and services depends on changes in oil prices directly or indirectly. Moreover, globalization has revealed the effect of the prices of goods and services on each other more clearly. Many countries’ economy depends on oil production, oil, and petroleum products trade, so estimating oil prices is an important task. Besides, some sectors are directly dependent on oil prices, such as manufacturing, logistics, and transportation. Therefore, oil prices affect not only these sectors but also the political and economic processes that determine countries’ economic growth and development. For this reason, there is need for predict the yield of the oil production in refinery.

Yield prediction is an integral part of every refinery process due to the intricacy of the process and composition of crude oil. Inability to sufficiently predict the yields before refinery process creates challenges such as inadequacy in planning the operating conditions to meet product target, product optimisation failure and inability to meet product market specification. Predicting yields using existed practised by the Tema Oil Refinery are time consuming and tedious. Using long-short-term-memory time series makes the prediction fast and easy.

1.3 AIM AND OBJECTIVES OF THE STUDY

The main aim of this work is to carry out a prediction of gasoline yield of crude oil refinery using long-short-term-memory time series. The objectives of the study are:

To predict the present and future availability of crude oil
To determine the production rate of the crude oil

To carry out tests on the system model

1.4 SIGNIFICANCE OF THE STUDY

Oil prices are rising due to an increase in demand and a decrease in supply. This study will serve a tool for predicting oil yield thereby ensuring there will be no oil scarcity.

This study will ensure that there is a crucial input into macroeconomic projections, in particular owing to the impact that oil prices have on inflation and output and, hence, on monetary policy. Using futures to forecast oil prices provides a transparent and simple tool which is easy to communicate.

1.5 SCOPE OF THE STUDY

Prediction of oil yield of important in that it ensure that there is no scarcity of oil due to the importance. Oil is a very important resource as one of major sources of energy. It can generate heat, drive machinery, and fuel vehicles and airplanes. Its components are used to manufacture almost all chemical products, such as plastics, detergents, paints, and even medicines.

Different oil prediction model have existed in the past but the scope of this work covers predicting the oil yield of crude oil refinery using long-short-term-memory time series.