Energy consumption and carbon emissions forecasting for industrial processes: Status, challenges and perspectives

The industrial process consumes substantial energy and emits large amounts of carbon dioxide. With the help of accurate energy consumption and carbon emissions forecasting, industrial enterprises would find it easier to achieve cleaner production, optimize the energy structure, and reduce production costs and carbon emissions by gaining deeper control over the production situation. Due to the oversaturation of machine learning modeling methods, forecasting models face difficulties in improving accuracy and extracting data features. The deep learning method is introduced to address these issues. However, the inaccurate measurement of key parameters and anomalies in data transmission strengthens the uncertainty of industrial big data. This makes the machine learning-based forecasting models show strong uncertainty and poor generalizability. Thus, it is significantly difficult to improve the accuracy of the current industrial energy consumption and carbon emissions forecasting model in different industrial scenarios. This work summarizes the research on energy consumption and carbon emissions forecasting for industrial processes in recent years. Combined with the actual problems of current industrial processes, this work summarizes three kinds of forecasting models: (i) the multi-step forecasting model based on the combination of deep learning and model uncertainty, (ii) the combined mechanism and data-driven method-based forecasting model, and (iii) the intelligent algorithms-based forecasting model. These models could become a new path for establishing generic industrial process energy consumption and carbon emissions forecasting models in the future.