A genetic-algorithm-based remnant grey prediction model for energy demand forecasting

Area of Science:

• Energy economics
• Time series analysis
• Computational intelligence

Background:

• Accurate energy demand forecasting is vital for economic planning.
• Traditional large-scale data methods are often impractical.
• The GM(1,1) model is a common, simple approach for limited data.

Purpose of the Study:

• To enhance the forecasting accuracy of the standard GM(1,1) model.
• To introduce a novel genetic-algorithm-based remnant GM(1,1) (GARGM(1,1)) model.
• To improve upon existing remnant GM(1,1) variants.

Main Methods:

• Development of the GARGM(1,1) model incorporating sign estimation.
• Simultaneous optimization of original and residual model parameters using a genetic algorithm (GA).
• Experimental validation using real-world energy demand data from China.

Main Results:

• The proposed GARGM(1,1) model demonstrated superior forecasting accuracy.
• The method effectively optimizes parameters for both primary and residual models.
• Performance was validated against other remnant GM(1,1) models.

Conclusions:

• The GARGM(1,1) model offers a significant improvement in energy demand forecasting accuracy.
• Genetic algorithms provide an effective mechanism for optimizing complex time series models.
• This approach is a valuable tool for energy planning and policy development.