A novel imputation approach for power load time series data based on tsDatawig.

Area of Science:

• Electrical Engineering
• Data Science

Background:

• Accurate power load forecasting is crucial for optimizing power grid operations and unit scheduling.
• Real-time load data from sensor networks is fundamental for forecasting models.
• Data gaps in sensor networks, caused by failures or interference, pose a significant challenge.

Purpose of the Study:

• To propose and evaluate a novel data imputation method for addressing missing power load data.
• To enhance the accuracy and reliability of power load forecasting models.

Main Methods:

• Historical power load data was analyzed using a data imputation approach.
• A time-coding method based on tsDataWig was developed for data preprocessing and encoding.
• The dataset was intentionally masked using three distinct missing data mechanisms.
• A power load data imputation framework was built utilizing the tsDataWig method.

Main Results:

• The proposed tsDataWig-based imputation method demonstrated significant advantages over existing approaches.
• Experimental results showed consistently lower prediction errors compared to other methods.
• The method effectively confirmed its capability in predicting missing power load values.

Conclusions:

• The developed time-coding imputation method effectively solves the problem of data missing loopholes in sensor networks.
• This approach enhances the reliability of power load data for improved grid operation and forecasting.