Toward fault tolerant modelling for SCADA based electricity distribution networks, machine learning approach

Area of Science:

• Electrical Engineering
• Computer Science
• Data Science

Background:

• Maintaining electrical energy is critical, especially in developing nations with limited resources.
• Increasing appliance use challenges optimal fault management and minimizing downtime.
• Existing research often overlooks fault duration minimization in resource-constrained environments.

Purpose of the Study:

• To propose accurate load redistribution estimation models for electrical grids.
• To develop a fault-tolerant approach using machine learning and SCADA to minimize fault duration.
• To aid electricity distribution companies in maintaining customer power supply and reducing outage times.

Main Methods:

• Utilized machine learning models: multiple linear regression, nonlinear regression, and classifier neural networks.
• Integrated information technology with electrical grids via a SCADA system.
• Trained and validated models using real smart grid data from zones with approximately 20 transformers.

Main Results:

• Achieved high accuracy of approximately 97% with a standard deviation of 2.3% in load redistribution estimation.
• Demonstrated the models' effectiveness in enabling the grid to redistribute loads strategically.
• Validated model performance using MATLAB and Anaconda-Python, confirming significant reduction in fault duration.

Conclusions:

• The proposed machine learning and SCADA-based fault-tolerant approach is effective in minimizing electrical fault durations.
• The models provide precise load redistribution strategies, enabling optimal actions within minimal time.
• This solution offers a viable method for enhancing power grid reliability, particularly in developing countries.