Construction and Empirical Analysis of Multi-source Data Fusion Intelligent Algorithm Model for Dual Control of Carbon Emissions

Abstract

In recent years, Hainan has actively promoted the construction of zero-carbon demonstration parks, aiming to achieve dual control of regional carbon emissions total and intensity through clean energy utilization, low-carbon industrial layout, and smart energy management. However, the parks still face many bottlenecks in energy monitoring, carbon emission assessment, and dynamic regulation, such as high data heterogeneity, complex spatiotemporal coupling, and insufficient real-time feedback mechanisms, resulting in limited accuracy in carbon emission prediction and making it difficult to provide reliable support for park management and policy optimization. In response to this, this paper built a multi-source data fusion intelligent algorithm model to control dual carbon emission. This model automatically derives time-series features, industry-related features using multi-layer nonlinear mapping, and introduces two constraint terms of the total carbon emissions and intensity in the objective to optimize dynamically the emission prediction and control performance. The experimental findings indicate that the model is optimally performed with a deep autoencoding fusion approach, and R2 reaches 0.96, which means that this approach is effective in capturing the nonlinear relationships between time, space, and industry, which improve the capabilities of the model to represent features.