Executive Summary: A Canadian oil and gas company collaborated with Zeroik Company to develop an AI-powered solution that automates the reading of complex meter data. This includes recognizing data from printed graphs, barcodes, and handwritten notes using machine learning and computer vision. The solution improved data accuracy, reduced manual errors, and increased operational efficiency.

Problem Statement: Traditional manual data reading from meters is time-consuming, prone to errors, and inefficient. With growing demands on resource extraction, accuracy in processing oil and gas meter data is crucial for efficiency and decision-making.

Proposed Solution: By implementing AI and computer vision technologies, the system can:

  • Automatically process round, graphical meter data by unfolding discs for easier reading.
  • Accurately recognize multiple color-coded curves from printed graphs despite noise.
  • Detect and process data from various formats and equipment.
  • Recognize and interpret handwritten data using trained neural networks, reducing human error.

Benefits:

  1. Increased Accuracy: The solution achieved over 80% accuracy in processing meter data, reducing manual input errors and inconsistencies.
  2. Efficiency Gains: Automated processing significantly cuts down time spent on manual data collection, boosting productivity.
  3. Scalability: The system can handle various formats and integrate with global partners’ equipment, making it adaptable across regions.
  4. Reduction of Human Error: With automated recognition of handwritten data and barcodes, the solution minimizes the risk of errors from manual data entry.

Financial Impact: By improving data accuracy and reducing processing time, this AI solution can lead to significant cost savings. The reduction in human intervention translates into labor cost reductions and fewer operational delays, directly impacting the bottom line.

Challenges and Mitigation:

  1. Data Noise: Overcoming background noise in graph readings was managed through robust neural network training, ensuring clearer curve identification.
  2. Handwriting Variability: The quality of handwritten inputs remains a challenge, though accuracy was improved to a range of 30-70% depending on input quality.
  3. Diverse Equipment Formats: The AI was trained to recognize and adapt to different data formats, addressing compatibility issues across equipment used globally.

Conclusion: Adopting AI and computer vision for meter data processing in the oil and gas industry enhances both accuracy and operational efficiency. With the ability to handle various data formats and reduce human errors, the solution promises long-term cost savings and competitive advantage.