M2SDF: a novel method for multi-seasonal and multiscale building energy prediction

Accurate and rapid prediction of building energy consumption is crucial for energy conservation. In this study, we introduce the M2SDF model to enhance the precision of building energy forecasts. First, we propose the MSTD module, which decomposes the data into multiple seasonal components and a trend component to capture complex patterns. Next, we introduce the MSF module, which integrates both macro and micro features. Finally, we obtain the prediction results by aggregating the output values from the predictors for each component. Experimental results demonstrate that our model surpasses other algorithms in both accuracy and stability.