Prediction method of carbon emissions of intelligent buildings based on secondary decomposition BAS-LSTM

In the face of the challenges of global climate change, smart building carbon emission prediction has become a key energy management tool. Therefore, this paper proposes an innovative BAS-LSTM carbon emission prediction method for intelligent buildings based on quadratic decomposition. This method combines variational mode decomposition and comprehensive empirical mode decomposition to decompose historical data of carbon emissions of intelligent buildings in detail, aiming to fully reveal their inherent time series patterns and dynamic characteristics. With this quadratic decomposition, we succeeded in obtaining the intrinsic modal function component of the carbon emission data, thereby building a rich input dataset. These components are integrated and fed into a long short-term memory network (LSTM) model. LSTM is a special recurrent neural network, which is especially suitable for processing time-dependent data. Through in depth learning and training, the model can effectively predict future carbon emissions. In order to further improve the accuracy and reliability of prediction, BAS algorithm is introduced to optimize the weights of LSTM model. The BAS algorithm is a global search strategy designed to find the optimal model parameters to obtain more accurate prediction results. In the experimental part, the proposed method is tested and evaluated comprehensively. The results show that this method not only has good quadratic decomposition effect, can capture all the components in the data, but also shows stable performance and powerful global search ability. In addition, by comparing with traditional carbon emission prediction methods, we find that the fitting coefficients of this method are all above 0.94, which indicates that it has excellent prediction accuracy.