MACHINE LEARNING'S INFLUENCE ON SUPPLY CHAIN AND LOGISTICS OPTIMIZATION IN THE OIL AND GAS SECTOR: A COMPREHENSIVE ANALYSIS

Computer Science & IT Research Journal, Journal Year: 2024, Volume and Issue: 5(3), P. 725 - 740

Published: March 28, 2024

Machine Learning (ML) is revolutionizing supply chain and logistics optimization in the oil gas sector. This comprehensive analysis explores how ML algorithms are reshaping traditional practices, leading to more efficient operations cost savings. enables predictive analytics, demand forecasting, route optimization, inventory management, improving overall performance. Supply sector inherently complex, involving numerous interconnected processes stakeholders. adept at handling this complexity by analyzing vast amounts of data identify patterns optimize operations. By leveraging historical data, can predict future demand, enabling companies adjust their levels production schedules accordingly. also play a crucial role helping minimize transportation costs reduce carbon emissions. factors such as traffic patterns, weather conditions, road determine most routes for transporting goods equipment. Furthermore, maintenance, which essential prevent equipment failures downtime. sensor from equipment, when maintenance required, allowing schedule proactively avoid costly disruptions. In conclusion, transforming maintenance. power ML, improve operational efficiency, costs, enhance performance. Keywords: Machine’s Learning, Chain, Logistics, Optimization, Oil Gas.

Language: Английский

---

Generative AI and process systems engineering: The next frontier

Computers & Chemical Engineering, Journal Year: 2024, Volume and Issue: 187, P. 108723 - 108723

Published: May 9, 2024

Language: Английский

---

Energy Usage Forecasting Model Based on Long Short-Term Memory (LSTM) and eXplainable Artificial Intelligence (XAI)

Information, Journal Year: 2023, Volume and Issue: 14(5), P. 265 - 265

Published: April 29, 2023

The accurate forecasting of energy consumption is essential for companies, primarily planning procurement. An overestimated or underestimated value may lead to inefficient usage. Inefficient usage could also financial consequences the company, since it will generate a high cost production. Therefore, in this study, we proposed an model and parameter analysis using long short-term memory (LSTM) explainable artificial intelligence (XAI), respectively. A public dataset from steel company was used study evaluate our models compare them with previous results. results showed that achieved lowest root mean squared error (RMSE) scores by up 0.08, 0.07, 0.07 single-layer LSTM, double-layer bi-directional In addition, interpretability XAI revealed two parameters, namely leading current reactive power number seconds midnight, had strong influence on output. Finally, expected be useful industry practitioners, providing LSTM offering insight policymakers leaders so they can make more informed decisions about resource allocation investment, develop effective strategies reducing consumption, support transition toward sustainable development.

Language: Английский

---

Energy Consumption Prediction Based on LightGBM Empowered With eXplainable Artificial Intelligence

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 91263 - 91271

Published: Jan. 1, 2024

The precise prediction of energy consumption is crucial for businesses, companies, and households especially when it comes to planning purchases. An underestimated or overestimated forecast value may result in the use inefficiently. companies will face financial consequences inefficient usage because production requires high costs. In this research, an model proposed employing Light Gradient-Boosting Machine (ECP_LightGBM) explainable artificial intelligence (XAI), respectively forecasting. A household dataset used study evaluation our also compare results with previously published approaches. According results, achieved lowest root mean square error. Furthermore, interpretability investigation using XAI indicated that feature name sub_metering_3 had a very strong impact on model's output which shows by air conditioner water heater. Lastly, can be helpful practitioners, offering LightGBM giving guidance leaders policymakers, so they allocate investments resources more intelligently.

Language: Английский

---

Implementation of deep neural networks and statistical methods to predict the resilient modulus of soils

International Journal of Pavement Engineering, Journal Year: 2023, Volume and Issue: 24(1)

Published: Sept. 20, 2023

ABSTRACTThe Resilient Modulus (Mr) is perhaps the most relevant and widely used parameter to characterise soil behaviour under repetitive loading for pavement applications. Accordingly, it a crucial controlling mechanistic-empirical design. Nonetheless, determining Mr by laboratory tests not always possible due high consumption of time financial resources. Thus, developing new indirect approaches estimating MR necessary. Precisely, this article investigates application Deep Neural Networks (DNNs) statistical methods predict soils. For that purpose, Long-Term Pavement Performance (LTPP) database was implemented. It includes 64 701 datasets resulting from coarse-grained fine-grained samples considering wide range grain size distribution subjected different stress levels. The input parameters were bulk stress, octahedral shear percentage particles passing through sieves (3", 2", 3/2", 1", 3/4", 1/2", 3/8", No. 4, 10, 40, 80, 200) output Mr. results suggest while conventional mathematical models are unable influence level on Mr, proposed DNNs able reproduce very accurate predictions. Notably, computational have been uploaded GitHub repository become valuable tool forecasting when experimental measurements feasible.KEYWORDS: neural networksresilient modulusstatistical methodsUS soils Disclosure statementNo potential conflict interest reported author(s).Data availability statementThe authors publicly share investigation; data can be accessed in following link: https://github.com/rpoloe/MR_DNN.Additional informationFundingThe appreciate support given Czech Science Foundation grant 21-35764J. first third acknowledge institutional Center Geosphere Dynamics (UNCE/SCI/006).

Language: Английский

---

State-Space Compression for Efficient Policy Learning in Crude Oil Scheduling

Mathematics, Journal Year: 2024, Volume and Issue: 12(3), P. 393 - 393

Published: Jan. 25, 2024

The imperative for swift and intelligent decision making in production scheduling has intensified recent years. Deep reinforcement learning, akin to human cognitive processes, heralded advancements complex found applicability the domain. Yet, its deployment industrial settings is marred by large state spaces, protracted training times, challenging convergence, necessitating a more efficacious approach. Addressing these concerns, this paper introduces an innovative, accelerated deep learning framework—VSCS (Variational Autoencoder State Compression Soft Actor–Critic). framework adeptly employs variational autoencoder (VAE) condense expansive high-dimensional space into tractable low-dimensional feature space, subsequently leveraging features refine policy augment network’s performance efficacy. Furthermore, novel methodology ascertain optimal dimensionality of presented, integrating reconstruction similarity with visual analysis facilitate informed selection. This approach, rigorously validated within realm crude oil scheduling, demonstrates significant improvements over traditional methods. Notably, convergence rate proposed VSCS method shows remarkable increase 77.5%, coupled 89.3% enhancement reward punishment values. substantiates robustness appropriateness chosen dimensions.

Language: Английский

---

Artificial intelligence and machine learning in future energy systems (state-of-the-art, future development)

Elsevier eBooks, Journal Year: 2024, Volume and Issue: unknown, P. 3 - 30

Published: Jan. 1, 2024

Language: Английский

---

Back propagation neural network model for controlling artificial rocks petrophysical properties during manufacturing process

Petroleum Science and Technology, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 27

Published: Oct. 12, 2024

Artificial rocks are increasingly used in experiments, and it is important to make artificial rocks' petrophysical properties similar real obtain more valuable results. The effect of widely five factors, grain size (GS), distribution (GSD), mass fraction cementing agent (MC), pressing pressure (PP), time (PT), on rock analyzed. Three GS, MC, PP, which suitable for establishing a quantitative model opted. relationships 80 plugs' manufacturing factors studied, indicates MC PP negatively correlated with porosity permeability, GS has significant permeability but little porosity. Furthermore, novel back propagation (BP) neural network proposed, can be determine factor values during process. A series core plugs manufactured relying calculated by the BP model, their tested compared design value verify model. Verification results show that comprehensive average error rock, including calculating error, 2.38 18.68%, respectively.

Language: Английский

---

Hierarchical Reinforcement Learning for Crude Oil Supply Chain Scheduling

Algorithms, Journal Year: 2023, Volume and Issue: 16(7), P. 354 - 354

Published: July 24, 2023

Crude oil resource scheduling is one of the critical issues upstream in crude industry chain. It aims to reduce transportation and inventory costs avoid alerts limit violations by formulating reasonable strategies. Two main difficulties coexist this problem: large problem scale uncertain supply demand. Traditional operations research (OR) methods, which rely on forecasting demand, face significant challenges when applied complicated short-term operational process To address these challenges, paper presents a novel hierarchical optimization framework proposes well-designed reinforcement learning (HRL) algorithm. Specifically, (RL), as an upper-level agent, used select operators combined various sub-goals solving orders, while lower-level agent finds viable solution provides penalty feedback based chosen operator. Additionally, we deploy simulator real-world data execute comprehensive experiments. Regarding alert number, maximum penalty, overall cost, our HRL method outperforms existing OR two RL algorithms majority time steps.

Language: Английский

---

Assessing the carbon footprint of soccer events through a lightweight CNN model utilizing transfer learning in the pursuit of carbon neutrality

Frontiers in Ecology and Evolution, Journal Year: 2023, Volume and Issue: 11

Published: July 24, 2023

Introduction Soccer events require a lot of energy, resulting in significant carbon emissions. To achieve neutrality, it is crucial to reduce the cost and energy consumption soccer events. However, current methods for minimization often have high equipment requirements, time-consuming training, many parameters, making them unsuitable real-world industrial scenarios. address this issue, we propose lightweight CNN model based on transfer learning study strategies carbon-neutral context. Methods Our proposed uses downsampling module human brain efficient information processing learning-based speed up training progress. We conducted experiments evaluate performance our compared with existing models terms number parameters computation recognition accuracy. Results The experimental results show that network has advantages over while achieving higher accuracy than conventional models. effectively predicts event data proposes more reasonable optimize costs accelerate realization neutral goals. Discussion promising method studying use allows faster indicate outperforms can predict optimization strategies. contribute goals sports industry.

Language: Английский

---