Carbon emission prediction method for urban regional energy system based on LSSVM

In order to address the issues of low mining rate, low prediction accuracy, and long time in traditional prediction methods, a carbon emission prediction method for urban regional energy system based on least squares support vector machine (LSSVM) is proposed. Filter the influencing factors of carbon emissions in urban regional energy systems, identifies abnormal influencing factor data using the density-based spatial clustering of applications with noise (DBSCAN) algorithm, reconstructs the data using a stacked denoising autoencoder. Using the reconstructed data as input for the model and the predicted carbon emissions as output, constructs a carbon emission prediction model for urban regional energy systems based on LSSVM and obtain relevant prediction results. The experimental results show that the mining rate of the influencing factors of the proposed method ranges from 96.7% to 98.2%, with a maximum prediction accuracy of 98.4% and an average prediction time of 0.81 seconds.