A novel method for spatiotemporal temperature prediction in offices by using Raman-distributed fiber-optic sensor with the data fusion method

The uneven distribution of temperature fields and dynamic personnel disturbance in open-plan offices are key research issues for achieving high-precision prediction of temperature fields. Using Raman-distributed fiber-optic sensors, we developed a novel method for predicting working area temperature in open-plan offices. The novelty of this study is that the proposed temperature prediction method can incorporate thermal environment changes caused by dynamic internal disturbances, which not only improves the temperature prediction accuracy, but also enables low-cost, high-density temperature field prediction. The method aimed to achieve high spatial resolution and strong generalizations for the temperature field, considering uneven temperature distributions and random personnel disturbances. An environmental adaptability calibration method was developed, reducing the measurement error to 0.09 °C. Using Computational Fluid Dynamics (CFD) transient simulation, we established a human impact quantification dataset for open-plan offices. Combining distributed fiber-optic sensors, the YOLO personnel recognition algorithm, and the human impact quantification dataset, we achieved temperature field prediction of personnel activity areas with an error of approximately 0.1 °C. The method's performance was validated through field experiments conducted in an open-plan office in Tianjin. The proposed low-cost, high spatial resolution temperature field prediction method can significantly improve temperature data acquisition accuracy for intelligent operation and maintenance of office buildings. Furthermore, it can provide data support for pre-adjustment control of indoor thermal environments by air conditioning systems through early temperature prediction, enabling further exploration of energy-saving and carbon reduction potentials in office building operations.