Construction and optimization of non-parametric analysis model for meter coefficients via back propagation neural...

Area of Science:

• Artificial Intelligence
• Machine Learning
• Data Science

Background:

• Traditional meter coefficient analysis methods suffer from low accuracy and long processing times.
• Existing techniques lack the efficiency and precision required for modern energy metering applications.

Purpose of the Study:

• To propose and evaluate a novel non-parametric analysis method for meter coefficient analysis.
• To leverage the classification and pattern recognition capabilities of the Back Propagation (BP) neural network.
• To overcome the limitations of traditional methods in terms of accuracy and processing time.

Main Methods:

• Utilized the United Kingdom Domestic Appliance-Level Electricity dataset for model training and testing.
• Employed non-parametric analysis for data pre-processing, feature extraction, and normalization.
• Implemented a Back Propagation (BP) neural network model for meter coefficient analysis.

Main Results:

• The proposed BP neural network model demonstrated significantly improved accuracy indicators (MAE, MRE) compared to the Least Squares Method (LSM).
• Achieved a Mean Absolute Error (MAE) of 0.025 and Mean Relative Error (MRE) of 1.32% on the testing dataset.
• The LSM method resulted in an MAE of 0.043 and an MRE of 2.56% on the same dataset.

Conclusions:

• The non-parametric analysis model based on the BP neural network offers superior accuracy for meter coefficient analysis.
• This novel approach provides a practical reference for the electricity industry in energy metering and load forecasting.
• The BP neural network presents a viable alternative to traditional methods, enhancing efficiency and precision.