Study on Load Forecasting of Combined Cooling, Heat and Power System with Deep Learning Algorithm Based on Attention Mechanism

As an important distributed energy system in the context of dual-carbon, CCHP (combined cooling, heating, and power) systems are widely used due to their advantages in energy utilization and environmental protection, and CCHP systems not only significantly improve energy utilization under the carbon tax penalty mechanism, but also effectively reduce the overall operating costs of distributed energy systems. Due to the many and complex factors affecting the system, especially the existence of demand charges and time-of-day tariffs in the industry, it is difficult for CCHP systems to operate efficiently under the economic model through the existing control methods. The traditional model-based scheduling algorithm of CCHP system is affected by the dynamic uncertainty of load, which leads to the problems of sub-optimal scheduling strategy and long decision-making time; accurate load prediction is an important prerequisite for optimized scheduling of distributed energy systems. Aiming at the above problems, this paper proposes to construct a CCHP demand-side load forecasting model by combining the attention mechanism and deep learning technology, and examines the accuracy and generalization of the proposed forecasting model through the design of evaluation indexes.