Virtual esp model to detect system degradation for preventive maintenance

A non-transitory, computer-readable medium having instructions stored thereon that are executable by a processor, the instructions comprising instructions to:

• select a plurality of wellbore variables Y(i), each wellbore variable representative of an aspect of a wellbore operation, where i is an integer between 1 and N;
• receive, at the processor, training data associated with each wellbore variable Y(i), the data collected over a plurality of time intervals j, where j is an integer between 2 and P;
• train, via the processor, a machine learning model M(j) of wellbore operation for each of the plurality of time intervals j based on the data for wellbore variables Y(i) collected over each respective time interval j;
• sample wellbore variables Y(i) over a time interval j, where j=P+1;
• identify a change in the wellbore operation, wherein identifying includes comparing model output responses from selected trained machine learning models of wellbore operation based on the wellbore variables Y(i) sampled over the time interval j=P+1;
• and
• modify the wellbore operation based on the modifying the wellbore operation based on the change identified in the wellbore operation.

A wellbore system, comprising:

• a processor;
• and
• a computer-readable medium connected to the processor, the computer-readable medium having instructions stored thereon, wherein the instructions are executable by the processor to cause the processor to,
• select a plurality of wellbore variables Y(i), each wellbore variable representative of an aspect of a wellbore operation in the wellbore system, where i is an integer between 1 and N;
• receive, at the processor, training data associated with each wellbore variable Y(i), the data collected over a plurality of time intervals j, where j is an integer between 2 and P;
• train, via the processor, a machine learning model M(j) of wellbore operation for each of the plurality of time intervals j based on the data for wellbore variables Y(i) collected over each respective time interval j;
• sample wellbore variables Y(i) over a time interval j, where j=P+1;
• identify a change in the wellbore operation, wherein identifying includes comparing model output responses from selected trained machine learning models of wellbore operation based on the wellbore variables Y(i) sampled over the time interval j=P+1;
• and
• modify the wellbore operation based on the modifying the wellbore operation based on the change identified in the wellbore operation.