Power prediction and packed bed heat storage control for marine diesel engine waste heat recovery

The generation capability of the waste heat recovery system cannot be fully utilized by relying solely on the exhaust gas, as the mass flow rate and temperature of exhaust fluctuate in real time. The volatility of ship engine conditions and the instability of power demand can pose a challenge to real-time power allocation. Developing integrated prediction and control methods for power allocation becomes necessary. An integrated system is devised to fully utilize waste energy from marine diesel engines by combining a regenerative organic Rankine cycle with a controllable packed bed. The trends of thermodynamic properties and dynamic charging and discharging processes of the system are analyzed. To provide a plausible management strategy for allocating power in real time, a bidirectional long-short-term memory network power prediction method based on improved complete ensemble empirical mode decomposition with adaptive noise and improved marine predator algorithm is proposed. It is found that 315.6 kW of power can be obtained from the system, and its payback period is 6.84 years. In addition, by using the proposed model, the system can reduce diesel engine power generation. The findings suggest the proposed solution is a viable and meaningful strategy.