Incorporating deep learning of load predictions to enhance the optimal active energy management of combined cooling, heating and power system

Abstract The energy management of combined cooling, heating and power (CCHP) system is essential for simultaneously improving its energy and economic performances. However, the conventional operation strategies are mainly logical control, which passively adapts to users’ demands. This paper proposes an optimal economic energy dispatch model of the CCHP system incorporating deep learning of load predictions to fulfill active control strategy in dynamic programming. A cross linear optimization method with a half update strategy of component efficiencies is developed to solve and calculate the variable component efficiencies in the CCHP system. Compared to the genetic algorithm, the proposed method achieves better results and the convergence time is reduced by 93%. The model predictive control of the CCHP system in load predictions of an artificial neural network is combined to the dynamic programming to realize the active energy dispatch strategy. The effects of short-term prediction and long-term prediction on forecast performances and operation costs are discussed. The case study demonstrates that the ideal prediction horizon of 8 h is recommended to fully realize the active functions of energy storage devices in the CCHP system. The proposed active strategy with model predictive control reduces the operational cost by 3.66% compared to the passive control strategy.