Multi-Time-Scale Optimal Scheduling Strategy for Marine Renewable Energy Based on Deep Reinforcement Learning...

Area of Science:

• Marine Renewable Energy
• Energy Systems Engineering
• Artificial Intelligence in Power Systems

Background:

• The Bohai and Yellow Seas offer significant marine energy potential, but wind and solar power generation face challenges due to seasonality and intra-day variability.
• Integrating diverse renewable sources is crucial for grid stability, yet scheduling complementary energy generation presents a substantial challenge.

Purpose of the Study:

• To propose a multi-time-scale rolling optimization scheduling strategy for complementary dispatch of wave, offshore photovoltaic, and wind power generation.
• To address the seasonality and intra-day uncertainty of renewable energy sources in the Bohai and Yellow Sea regions.
• To reduce system costs and ensure grid stability through intelligent energy management.

Main Methods:

• Utilized actual meteorological data from the Shandong Peninsula region.
• Employed a CNN-LSTM neural network for day-ahead (24h) and intra-day (1h) power generation and load demand forecasting.
• Applied the DDPG algorithm for rolling optimization scheduling based on forecast data.

Main Results:

• The proposed strategy effectively meets load demands by complementary scheduling of wave, wind, and photovoltaic power.
• Demonstrated reduction in the negative impacts of energy source variability on the grid.
• Achieved system cost reductions of 16.1% (winter) and 22% (summer) compared to day-ahead scheduling alone.

Conclusions:

• The integrated wave, wind, and photovoltaic power generation strategy, coupled with rolling optimization, enhances grid stability and economic efficiency.
• The developed scheduling approach effectively leverages the climatic characteristics of the Bohai and Yellow Sea regions for reliable renewable energy integration.
• This study provides a robust framework for managing intermittent renewable energy sources in complex maritime environments.