Abstract: The PV-diesel-battery hybrid power system can extend the range of unmanned surface vessel for maritime patrol, keep missions uninterrupted, and improve the level of maritime rights protection in remote sea area. However, fluctuations in load demand power will occur due to the time-varying characteristics of wave height, and the photovoltaic power will change significantly due to the intermittency of solar irradiation density under complex sea conditions, which brings challenges to energy management. A real-time energy management strategy based on wave height and solar irradiation density prediction is proposed. A hybrid CNN-LSTM model combined with convolutional neural networks(CNN) and long short-term memory(LSTM) is utilized to predict wave height and solar irradiation density for acquiring load demand power and photovoltaic power. Model predict control (MPC) strategy is subsequently employed to optimize energy flow. The comparison is carried out between the proposed method and MPC with traditional prediction method. The results of hardware-in-the-loop experiment show that the proposed method significantly improves energy efficiency under high sea states. The dimensionless sea condition factor, composed of the maximum photovoltaic power and the average load demand power, is proposed to quantify the characteristic of complex sea conditions. The correlation analysis shows that the dimensionless sea condition factor is significantly correlated with the energy efficiency improvement. The proposed method has the potential to enhance the complex environment adaptability of the energy management strategy employed by the PV-diesel-battery vessels, thereby providing valuable engineering guidance.

Key words: PV-diesel-battery vessels, energy management strategy, wave height, solar irradiation density, CNN-LSTM, dimensionless sea condition factor