Improving energy consumption prediction for residential buildings using Modified Wild Horse Optimization with Deep...

Area of Science:

• Environmental Science
• Computer Science
• Engineering

Background:

• Building energy consumption significantly contributes to environmental issues.
• Accurate energy prediction is vital for energy conservation and informed decision-making.
• Machine Learning (ML) offers effective strategies for energy consumption prediction.

Purpose of the Study:

• To introduce a Modified Wild Horse Optimization with Deep Learning approach for Energy Consumption Prediction (MWHODL-ECP) in residential buildings.
• To enhance the precision and timeliness of energy consumption forecasts for residential structures.
• To leverage advanced ML techniques for optimizing energy usage and reducing environmental impact.

Main Methods:

• The MWHODL-ECP model employs rigorous data preprocessing, including merging, cleaning, conversion, and normalization.
• A deep belief network (DBN) is utilized as the core predictive model.
• The Modified Wild Horse Optimization (MWHO) algorithm is applied for hyperparameter tuning to optimize model performance.

Main Results:

• The MWHODL-ECP model demonstrated superior performance compared to existing deep learning models.
• Achieved highly effective prediction results with MSE of 1.10, RMSE of 1.05, and MAE of 0.41.
• Attained an R-squared value of 96.28, indicating strong model accuracy, with a training time of 1.23.

Conclusions:

• The MWHODL-ECP model provides an accurate and efficient method for predicting residential building energy consumption.
• This approach significantly improves prediction accuracy, supporting energy conservation and environmental sustainability.
• The study highlights the potential of integrating advanced optimization algorithms with deep learning for complex prediction tasks.