Generative ai agents for production engineering in oil and gas

Abstract

A method for performing production engineering tasks includes receiving a question or instruction related to an oil and/or gas industry. The question or instruction is received by a large language model (LLM) agent. The method also includes selecting one or more tools from a plurality of tools using the LLM agent based upon the question or instruction. The tools include a data tool, a retrieval augmented generation (RAG) tool, a simulation tool, and an analytical tool. The method also includes generating output data related to the question or instruction using the one or more identified tools. The method also includes generating a response to the question or instruction using the LLM agent based upon output data.

What is claimed is: 
     
         1 . A method for performing production engineering tasks, the method comprising:
 receiving a question or instruction related to an oil and/or gas industry, wherein the question or instruction is received by a large language model (LLM) agent;   selecting one or more tools from a plurality of tools using the LLM agent based upon the question or instruction, wherein the tools comprise a data tool, a retrieval augmented generation (RAG) tool, a simulation tool, and an analytical tool;   generating output data related to the question or instruction using the one or more identified tools; and   generating a response to the question or instruction using the LLM agent based upon output data.   
     
     
         2 . The method of  claim 1 , wherein the data tool is selected in response to the question or instruction including structured data. 
     
     
         3 . The method of  claim 2 , wherein the question or instruction is related to:
 a production summary for a predetermined time period; and/or   a corrective maintenance (CM) work order for a compressor.   
     
     
         4 . The method of  claim 1 , wherein the RAG tool is selected in response to the question or instruction including unstructured data. 
     
     
         5 . The method of  claim 4 , wherein the question or instruction is related to an effect of a production separation temperature on basic sediment and water (BSW) specifications for crude treatment as well as liquid in a gas processing train. 
     
     
         6 . The method of  claim 1 , wherein the simulation tool is selected in response to the question or instruction relating to simulations. 
     
     
         7 . The method of  claim 6 , wherein the question or instruction is related to:
 an impact on total production in response to decreasing a speed of an electrical submersible pump (ESP) by a particular amount in a well; and/or   emissions from the compressor.   
     
     
         8 . The method of  claim 1 , wherein the analytical tool is selected in response to the question or instruction including a data analysis involving machine learning (ML) models and physics-based calculations. 
     
     
         9 . The method of  claim 8 , wherein the question or instruction is related to carbon dioxide (CO 2 ) emissions for the well in response to switching from electricity to natural gas. 
     
     
         10 . The method of  claim 1 , further comprising performing a physical action based upon the response. 
     
     
         11 . A computing system, comprising:
 one or more processors; and   a memory system comprising one or more non-transitory computer-readable media storing instructions that, when executed by at least one of the one or more processors, cause the computing system to perform operations, the operations comprising:
 receiving a question or instruction related to an oil and/or gas industry, wherein the question or instruction is received by a large language model (LLM) agent; 
 selecting one or more tools from a plurality of tools using the LLM agent based upon the question or instruction, wherein the tools comprise:
 a data tool that is configured to be selected in response to the question or instruction including structured data; 
 a retrieval augmented generation (RAG) tool that is configured to be selected in response to the question or instruction including unstructured data; 
 a simulation tool that is configured to be selected in response to the question or instruction relating to simulations; and 
 an analytical tool that is configured to be selected in response to the question or instruction including a data analysis involving machine learning (ML) models and physics-based calculations; 
 
 generating output data related to the question or instruction using the one or more identified tools; and 
 generating a response to the question or instruction using the LLM agent based upon output data. 
   
     
     
         12 . The computing system of  claim 11 , wherein the one or more tools are also identified based upon a context on a screen of a user. 
     
     
         13 . The computing system of  claim 12 , wherein the context comprises a region and/or asset that the user is working on and an ongoing conversation history of the user. 
     
     
         14 . The computing system of  claim 11 , wherein the question is related to identifying pumps with a cavitation risk and suggesting immediate corrective actions, and wherein the output data comprises:
 a list of the pumps with the cavitation risk, generated by the data tool, wherein the list also comprises a type of the pumps, start and end dates for running the pumps, and a severity level of the cavitation risk; and   the immediate corrective actions, generated by the RAG tool, wherein the immediate corrective actions comprise reducing a speed of the pumps and/or increasing a suction head of the pumps.   
     
     
         15 . The computing system of  claim 11 , wherein the output data requests additional information from the user, and wherein the response shows one or more sources that were used to generate the output data. 
     
     
         16 . A non-transitory computer-readable medium storing instructions that, when executed by one or more processors of a computing system, cause the computing system to perform operations, the operations comprising:
 receiving a question or instruction related to an oil and/or gas industry, wherein the question is related to identifying pumps with a cavitation risk and suggesting immediate corrective actions, and wherein the question or instruction is received by a large language model (LLM) agent;   selecting one or more tools from a plurality of tools using the LLM agent based upon the question or instruction, wherein the one or more tools are also identified based upon a context on a screen of a user, wherein the context comprises a region and/or asset that the user is working on and an ongoing conversation history of the user, and wherein the tools comprise:
 a data tool that is configured to be selected in response to the question or instruction including structured data; 
 a retrieval augmented generation (RAG) tool that is configured to be selected in response to the question or instruction also including unstructured data; 
 a simulation tool that is configured to be selected in response to the question or instruction relating to simulations; and 
 an analytical tool that is configured to be selected in response to the question or instruction including a data analysis involving machine learning (ML) models and physics-based calculations; 
   generating output data related to the question or instruction using the one or more identified tools, wherein the output data requests additional information from the user, and wherein the output data comprises:
 a list of the pumps with the cavitation risk, generated by the data tool, wherein the list also comprises a type of the pumps, start and end dates for running the pumps, and a severity level of the cavitation risk; and 
 the immediate corrective actions, generated by the RAG tool, wherein the immediate corrective actions comprise reducing a speed of the pumps and/or increasing a suction head of the pumps; 
   generating a response to the question or instruction using the LLM agent based upon output data, wherein the response shows one or more sources that were used to generate the output data; and   displaying the response, wherein displaying the response comprises generating a graph.   
     
     
         17 . The non-transitory computer-readable medium of  claim 16 , wherein the simulation tool and the analytical tool are not selected due to the question or instruction not being related to simulation, ML models, and/or physics-based calculations. 
     
     
         18 . The non-transitory computer-readable medium of  claim 16 , wherein the operations further comprise performing an action in response to and/or based upon the response. 
     
     
         19 . The non-transitory computer-readable medium of  claim 18 , wherein the action comprises generating and/or transmitting a signal that recommends, instructs, or causes a physical action to occur. 
     
     
         20 . The non-transitory computer-readable medium of  claim 19 , wherein the physical action implements the immediate corrective actions.