Integrated ESP Wells Platform for Optimization and Management Based on Mechanism Analysis and Data-Driven Model

Abstract The optimization and management for ESP wells is an integrated engineering technology. Only if the oil field have complete data of Geological, Reservoir, Wellbore, ESP, and even electrical, the ESP’s production status can be accurate judgment. So, there is a great demand of an integrated method to diagnose faults, analyze fault causes, and propose optimization measures. Then, using an integrated platform to visualize production data and analysis results. In this study, a mechanism analysis logic has been built first based on the onsite experience and integrated database including Geological, Reservoir, Wellbore, and ESP data. Then, different data type uses different machine learning method. For example, the vibration analysis using frequency/time domain, the Current card using image recognition (CNN), the pressure and production data using Random Forest. After that, using a neural network to combine the mechanism analysis logic and all the machine learning models to improve the accuracy. The platform is built to data and results visualization including three modules: Data monitor, Process automation, Rapid abnormal response. The model in this platform includes both experience, mechanism, and data-driven, with higher reliability, controllability, and interpretability. Moreover, the model has a full process function. By inputting the initial dataset, it can quickly respond to the diagnosis results of ESP faults and then corresponding process optimization measures. By using such a platform, on-site work can be greatly simplified, and the management of ESP wells can be more standardized and modular. With the automatic data collection and real-time transmission, the platform can diagnose the current production status of ESP wells in real time, warn of possible future failures, and propose trouble shooting measures. The visualized ESP management platform integrates data-driven methods, with built-in algorithms for automatic updating, achieving self-iterative updates and upgrades, making the management of ESP wells more efficient. After the deployment of the oil field, the model will continue to optimize and improve based on real-time data, which can effectively increase the MTBF (Mean time before faults) of the ESP.