Data-driven adaptive GM(1,1) time series prediction model for thermal comfort

Area of Science:

• Building Environment and Energy
• Human Comfort Studies
• Predictive Modeling

Background:

• The predicted mean vote (PMV) index is crucial for assessing indoor thermal comfort.
• PMV is influenced by six complex, non-linear environmental variables.
• Simplifying PMV calculations is necessary due to variable interdependencies.

Purpose of the Study:

• To develop an advanced predictive model for the PMV index.
• To enhance the accuracy of future thermal comfort predictions.
• To facilitate proactive control of indoor climate systems.

Main Methods:

• Utilized an improved grey system prediction model, GM(1,1).
• Introduced an adaptive GM(1,1) model to handle time series fluctuations.
• Employed a sliding window approach for continuous dataset adaptation.

Main Results:

• The adaptive GM(1,1) model demonstrated enhanced prediction accuracy for PMV.
• The improved model achieved better performance grades compared to standard methods.
• The model effectively addresses the irregularity and uncertainty of PMV time series.

Conclusions:

• The adaptive GM(1,1) model provides a robust method for future PMV prediction.
• This research supports advanced, human-centric climate control in smart homes.
• Accurate PMV prediction enables pre-emptive air conditioning adjustments for optimal comfort.