Towards Modified Entropy Mutual Information Feature Selection to Forecast Medium-Term Load Using a Deep Learning...

Area of Science:

• Electrical Engineering
• Data Science
• Energy Systems

Background:

• Accurate load forecasting is crucial for power companies to balance energy supply and demand.
• Medium-term load forecasting is essential for grid maintenance, electricity pricing, and energy sharing.
• Forecasting month-ahead electrical loads facilitates energy interchange between power companies.

Purpose of the Study:

• To propose an accurate medium-term load forecasting model for predicting month-ahead hourly electrical loads.
• To enhance forecasting accuracy and convergence speed using advanced machine learning and optimization techniques.
• To investigate and model dynamic consumer consumption behaviors for improved load prediction.

Main Methods:

• Utilized hourly electrical load and temperature data for forecasting.
• Applied modified entropy mutual information for feature selection to reduce data redundancy and irrelevancy.
• Employed Conditional Restricted Boltzmann Machine (CRBM) optimized with the Jaya meta-heuristic algorithm.
• Investigated consumer behavior using discrete-time Markov chains and adaptive k-means clustering.

Main Results:

• The proposed model demonstrated superior accuracy in month-ahead hourly electrical load forecasting.
• Achieved faster convergence and reduced execution time compared to existing models.
• Successfully clustered dynamic consumer consumption behaviors.

Conclusions:

• The integrated CRBM model with Jaya optimization offers a robust solution for medium-term load forecasting.
• The feature selection and consumer behavior analysis contribute to enhanced prediction accuracy.
• The model provides valuable insights for efficient power grid management and energy trading.