To solve significant differences in the performance of solar-air source heat pump systems under different weather conditions for an office building of scientific research in Shanghai, it proposes to classify the meteor-ological data by cluster analysis in this paper. Mainly, it uses solar insolation and outdoor temperature as two primary indicators with ten secondary indicators for further analysis. By this means, this study classifies 90 days of winter meteorological data in Shanghai into eight categories. Data standardization, factor analysis, and k-means clustering are the critical methods, and Bayes discriminant verifies the correct rate of 98.9% in the paper. Furthermore, it selects the typical day of every selected class to analyze the heat pump operating time effect on the system COP. Finally, the maximum system COP was used to determine the heat pump operating time as the performance optimization target. Meaning-fully, the intra-class daily data was verified to prove that the clustering result was highly accurate and reliable for further research and provides a solid related system control strategy.
Keywords solar energy, air-source heat pump, k-means clustering, Bayes discriminant, factor analysis