Non-Causal Control For Wave Energy Conversion Based on the Double Deep Q Network

To harness maximal wave energy, control and optimization for wave energy converters(WECs) have been investigated for decades. It has been long recognized that WEC control is essentially a non-causal control problem, in which future wave determines current control decisions. This paper introduces double deep Q network into the foundation of the non-causal time variant PD control system, enabling real-time parameter adjustments for dynamic control responses. Additionally, this paper delves into a comparative assessment of the influence of different prediction horizons on the efficiency of energy harvesting. The primary objective of this study is to elevate the control performance of wave energy converters, facil-itating more efficient capture and conversion of wave energy into usable electrical power. The integration of deep reinforcement learning empowers researchers to adapt swiftly to fluctuating waves and ocean conditions, fine-tuning control parameters to enhance overall system efficiency and stability. Taking the point absorber as an example, the effectiveness of the proposed method has been verified. This method can be straightforwardly applied to other types of WEC, such as Dielectric Elastomer Generators and Dielectric Fluid Generators.