Data analytics approach for online produced fluid flow rate estimation in SAGD process

Sedghi, Shabnam; Tan, Ruomu; Huang, Biao

doi:10.1016/j.compchemeng.2020.106766

Cited by 8 publications

(2 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Soft sensors have been developed and implemented for flow rate (m 3 h −1 ) measurements in steam‐assisted gravity drainage (SAGD) wells using data‐driven models based on partial least squares regression or multivariate linear regression by Sedghi et al [ 56 ]…”

Section: And Da Applications In Upstream Petroleum Industrymentioning

confidence: 99%

Identifying Applications of Machine Learning and Data Analytics Based Approaches for Optimization of Upstream Petroleum Operations

2020

View full text Add to dashboard Cite

In the mid-19th century, the first oil well was drilled, which marked the beginning of the world oil economy. This era embarked on the phase of evolution and development of society in every aspect, with huge dependence on petroleum products, which subsequently led to tremendous growth in terms of technological advancements in the exploration and production activities over time. With the existing petroleum fields approaching toward depletion, the global need for energy resources is increasing simultaneously. This continuously rising demand has led us to move to complex environments for oil and gas exploration and production activities. Moreover, the real-time monitoring of reservoirs and well operations produces enormous data, which require efficient processing and analysis tools for decisionmaking regarding field development and asset management. [1,2] Subsequently, the need for optimum facilities design, [3] transportation [4] and refining [5] activities at reduced cost has also grown. Several attempts [6-9] are also being made to develop unconventional fields commercially based on technoeconomic viability. The implementation of computer-based techniques has proven to be promising toward many challenging problems for various oilfield operations. Several researchers have shown that the machine learning (ML) and data analytics (DA) approaches hold promising solutions toward the operational challenges being encountered at present in the industry and can efficiently help in resolving the issues related to interpretation and analysis of large datasets. [10-15] Nikravesh and Aminzadeh [16] have shown the use of neural networks (NNs) and fuzzy logic for mining petroleum data. Li and Li [17] combined a NN and cluster analysis to put forth a predictive model for identifying complex lithology. Fath et al. [18] proposed a novel approach for bubble point pressure prediction using an NN model with reservoir temperature, solution gas oil ratio (GOR), oil gravity, and gas specific gravity as input attributes. Ahmadi and Bahadori [19] showed the use of a supervised learning algorithm to determine the well placement and conning occurrence in horizontal wells. Maucec et al. [20] performed data mining and ML on well stimulation data for enhancing the prediction capabilities. Gupta et al. [21] applied DA for safeguarding real-time electrical submersible pumping operations. Cadei et al. [22] forecasted operational upsets using advanced analytics in an upstream production system. Dongxiao et al. [23] presented in their research that long short-term memory (LSTM), cascaded LSTM, and a fully connected NN can be utilized for generating synthetic well logs by supplementing the missing logging input data. Ghorbani et al. [24] proposed an artificial Intelligence (AI)based approach to predict the flow rate of oil from an orifice

show abstract

Section: And Da Applications In Upstream Petroleum Industrymentioning

confidence: 99%

Identifying Applications of Machine Learning and Data Analytics Based Approaches for Optimization of Upstream Petroleum Operations

2020

View full text Add to dashboard Cite

show abstract

“…For example, data can be used for prediction, monitoring, control, optimization, and so on. − Data-driven virtual sensors or soft-sensors are a significant application of process data analysis techniques in manufacturing with the goal of predicting the hard-to-measure product quality or key performance indicators (KPIs) from other easy-to-measure process variables. , Compared with soft-sensors derived from the first-principles, data-driven soft-sensors do not require sufficient prior process knowledge and can be derived directly from the data. In recent years, data-driven soft-sensors have attracted growing attention and achieved successful applications in various areas. ,− Although data-driven soft-sensors can be constructed in a variety of ways, this research focuses on the just-in-time (JIT) learning method because it has the capacity of dealing with both nonlinearity and changes in process characteristics. , As illustrated in the questionnaire survey in Japan, model maintenance is considered to be the most important issue associated with soft-sensors because the prediction accuracy of soft-sensors may degrade due to changes in process characteristics. The JIT learning method provides a solution to prevent a decrease in prediction accuracy and has attracted widespread attention in various industrial processes. − The JIT learning method constructs a local model using the relevant samples stored in a database each time when an output estimate is required for a query sample.…”

Section: Introductionmentioning

confidence: 99%

Fast Locally Weighted PLS Modeling for Large-Scale Industrial Processes

Zhang

Wei

Song

2020

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

Locally weighted partial least-squares (LW-PLS) is an efficient just-in-time (JIT) modeling method, which can handle process collinearity, nonlinearity, and time-varying characteristics. However, it is not suitable for modeling large-scale industrial processes due to its huge computational cost. To solve this issue, the present work proposes a novel fast LW-PLS (FLW-PLS) method that can handle large-scale process data well. FLW-PLS is designed based on the exact Euclidean locality-sensitive hashing algorithm. Unlike standard LW-PLS which implements a brute-force linear search of similar samples over a reference data set, FLW-PLS searches for similar samples in sublinear time. Thus, significant computational savings can be obtained by FLW-PLS. The effectiveness of the proposed FLW-PLS method was validated through the case studies of predicting the silicon content and phosphorus content in the ironmaking process. The application results have shown that when faced with large-scale data, FLW-PLS can significantly reduce the prediction time without significantly reducing the prediction accuracy in comparison to the conventional LW-PLS method.

show abstract

A comprehensive overview of the applications of kernel functions and data-driven models in regression and classification tasks in the context of software sensors

Ngu,

Yeo,

Thien

et al. 2024

Applied Soft Computing

View full text Add to dashboard Cite

Data analytics approach for online produced fluid flow rate estimation in SAGD process

Cited by 8 publications

References 19 publications

Identifying Applications of Machine Learning and Data Analytics Based Approaches for Optimization of Upstream Petroleum Operations

Identifying Applications of Machine Learning and Data Analytics Based Approaches for Optimization of Upstream Petroleum Operations

Fast Locally Weighted PLS Modeling for Large-Scale Industrial Processes

A comprehensive overview of the applications of kernel functions and data-driven models in regression and classification tasks in the context of software sensors

Contact Info

Product

Resources

About