The oil and gas industry operates in a complex and ever-changing environment with uncertainties that threaten return on investment(ROI).
Upstream Volatility, midstream challenges, and shifting consumer demands mean that oil and gas companies need accurate forecasts for planning, operational efficiency, and economic investment.
Reservoir production, demand, and price are all areas in which oil and gas companies look to forecast and simulate. However, conventional methods of forecasting often fall short due to multiple complexities in the industry.
The repercussions of these shortcomings lead to missed opportunities and safety concerns, among other issues. For instance, inaccurate demand forecasting could lead to overstocking or understocking, which could lead to losses.
Artificial intelligence(AI) is not constrained to the world of science fiction anymore. AI research has given rise to real-world applications such as self-driving cars and speech recognition, with powerful AI techniques like computer vision and natural language processing(NLP).
AI for oil and gas continues to help the industry, with advanced algorithms powering more efficient processes and workflows. Combining AI with technologies like big data is ushering in new ways of forecasting by harnessing even more value from data. Our article explores how oil and gas companies can leverage AI for forecasting.
Forecasting different dynamics in the oil and gas industry is fraught with complexity due to the immense number of variables, voluminous data, and uncertainties.
But what are some of the conventional methods of forecasting and their challenges [1]?
Numerical simulation: The simulations depend on geological models and historical data matching.
Challenge: Demands vast computation resources, time and is difficult to scale due to history matching challenges [2].
Decline curve analysis: A popular technique that forecasts short-term oil production. It leverages historical production data to determine future trends.
Challenge: It assumes that production decline takes a steady pattern, which can be unreliable and misleading due to unsteady and complex production scenarios.
Econometric models: A quantitative technique that forecasts consumer demand. It captures the relationship between demand and the variables affecting it, such as economic factors.
Challenge: They rely on assumptions, abstractions, and simplifications, failing to capture the reality of real-world constraints.
What quantities will the wells produce? What is the optimal recovery method? Reservoir modeling and forecasting aims to answer such questions. The output from these processes allows companies to determine the viability of their investment.
Accurate forecasting is critical for predicting the characteristics of a reservoir. Engineers need effective strategies to predict reserve and product rates accurately. Some of the properties to consider include production history, well properties, reservoir drive mechanism, and external factors.
Production forecasting with AI analyzes large amounts of data, optimizes parameters, and identifies patterns. Algorithms like artificial neural networks (ANN), long short-term memory, and genetic algorithms (GA) can build ensemble models for forecasting oil and gas production [3].
Machine learning algorithms can determine the correlation between reservoir properties and well production. Models built with ANN can take in parameters like the oil’s porosity and thickness to predict production [4].
Additionally, other forecast models correlate past data sequences to determine future values. This method works with time series training data to accurately predict the oil production rate of wells.
Oil demand forecasting is crucial for decision-making and energy planning. Accurate forecasts help companies determine pricing and investment strategies. Traditional forecasting methods often fail to capture the intricate dynamics and factors, such as price volatility, that are common in the oil and gas industry.
AI-powered demand forecasting offers greater accuracy compared to traditional methods. By using AI to anticipate oil and gas demand, the industry can optimize production, inform pricing decisions, and find ways to adapt to shifting market demands.
ML algorithms like ANN, fuzzy regression, and GA can be used to build and optimize predictive models for forecasting oil and gas demand [5]. These models learn from a given dataset, recognizing the interdependencies between different variables to determine forecasts. Additionally, AI-based solutions can use reinforcement learning to get better continuously and learn from new scenarios.
ExxonMobil has built an AI-powered demand forecasting model that uses data from multiple sources, including market trends, production facilities, and transportation networks. The company has seen a reduction in forecast errors by 25% [6].
Crude oil and gas is a primary energy source today, impacting livelihoods globally. Predicting the price of oil and gas is a complex problem due to the myriad of factors that affect it. Current forecasting models include time series and econometric models [7].
AI for oil and gas price prediction can be tailored to consider factors like historical and current price information, weather patterns, market volatility, demand, and supply.
Ensemble techniques that combine multiple algorithms like ANN, GA, and SVM can build models to predict crude oil prices [8]. However, to make such an AI-driven solution complete, it must access current news and trends and not solely rely on historical data [9]. Despite the complexities, AI systems do a good job of utilizing data and other indicators like market sentiments and geo-political factors for predictions.
Forecasters still believe that AI has some way to go before it can accurately predict oil prices.
When asked whether AI could predict oil prices, Al Salazar, the senior VP at Enverus Intelligence Research (EIR), recognized AI’s advantage over convection methods. According to the VP, AI’s advantages include potent data processing and computing capabilities. However, Mr. Salazar was weary of AI technology's ability to track OPEC actions combined with other geo-political influences [10].
The oil and gas industry is a high-stakes sector with a constant need to monitor and evaluate risk. Companies must track a wide range of potential threats, including operational hazards, regulatory compliance, and environmental risks. Non-optimized risk assessment and management processes could lead to oversight, resulting in reactive issue resolution.
Traditionally, the industry uses preliminary risk analysis(PRA) to manage potential threats [11]. The method includes extensive processes that may not prioritize risks correctly and may suffer from human bias.
Digital transformation allows the industry to forecast, visualize, and even simulate potential threats. Some of the areas in which AI can forecast risk include:
AI-powered risk forecasting can help oil and gas industries avert risks associated with exploration, production, and operations.
There’s no doubt that AI offers the oil and gas industry several advantages in forecasting. Here are the main ones:
Forecasting helps oil and gas enterprises adjust and optimize their strategies. Traditional forecasting methods can be limited and subjective to human bias.
AI and ML are revolutionizing how predictions are made in the industry. AI’s ability to analyze large datasets and process specific tasks of varying complexities has ushered in a new generation of forecasting solutions. AI-powered forecasting has the potential to help the oil and gas industry optimize production, navigate volatile markets, and manage risks effectively.
Written by Baptiste Aelbrecht & Jacques Mojsilovic
