Real-time Latching Control Of Wave Energy Converters Based On LSTM Artificial Neural Network

Currently,with the initiate of a new round of global energy revolution,developing renewable energy has become a concerted and ambitious action goal for mankind to address climate change and global energy transformation.Wave energy is a kind of marine renewable energies with high energy density,cleanness and great development potential.Wave energy can be converted into usable electric energy by using Wave Energy Converter(WEC).Oscillating buoy device,also known as Point Absorber(PA),is one of the most popular WECs using ocean wave energy at present.Its heave motion will drive the Power Take-Off(PTO)system to generate electricity.Firstly,based on three-dimensional potential flow theory,this thesis analyzes the hydrodynamic characteristics of the buoy WEC device in the frequency domain,establishes the frequency domain motion equation of the device,converts the dynamic variables in the frequency domain-added mass,potential damping,wave excitation force transfer function into the state vector described in the control theory using the system identification method,and replaces the convolution term in the time domain motion equation with the state space,Runge-Kutta interpolation method is used to obtain the motion response of WEC in random waves.Based on this work,the accuracy of the approximate substitution method of state space model is verified by comparing the motion response results of time domain and frequency domain simulation.Secondly,traditional WEC can only maximize the wave energy conversion efficiency at its own single resonant frequency.But the wave frequency is often distributed in a specific range in the actual sea state,and the motion of the WEC is also quite random.Therefore,it is necessary to fully collect the energy of the WEC at any time and at any frequency.In this study,we aim to improve the wave energy conversion efficiency of the WEC device by taking effective real-time latching control strategy.The latching control is a brake-release two-phase control mode(bang-bang control)realized by a very large but limited alternating force,which makes the motion tend to the optimal control in the form of discrete approximation.By introducing Hamiltonian function,the latching control command within a certain time interval is determined according to the maximum principle of Pontryagin.The latching control can adjust the movement speed phase of the buoy to make it in phase of the wave excitation force,achieving a resonance effect similar to that at various frequencies.The width of the captured energy will be much larger than that without control.Next,the optimal control theory is based on the assumption that all wave excitation forces are known,but in actual sea conditions,it is impossible to know the all wave excitation forces.Therefore,the Model Predictive Control(MPC)method in intelligent control methods is combined with the time series prediction method based on deep learning.A Long Short Term Memory(LSTM)artificial neural network has been innovatively proposed as the main predictive model for predicting short-term wave forces.Compared with the traditional deep learning neural network,the LSTM neural network has one more cell state and multiple gates in the hidden layer,and it can effectively retain association information from a long time ago.Its prediction performance is more suitable for engineering applications that need a long prediction period,such as wave prediction.Based on prediction,the latching control module will be updated using the receding time domain and achieve optimal energy conversion within the prediction time domain.Therefore,the MPC method belongs to sub-optimal control.Due to the low requirements for model accuracy,including the prediction of external input,and taking the system constraints into account,it has more advantages than other methods.Then,based on the assumption that the power take-off system is simplified as a dampingspring system,numerical simulations were conducted on the motion response and wave energy absorption performance of the oscillating buoy wave energy converter under different control strategies are studied under the action of regular and random waves.The MPC strategy parameters were selected,and the energy capture width was used as an indicator to evaluate and optimize the damping and latching damping parameters of the PTO system,Use it as a real-time control parameter.Finally,to explore the practical application of real-time latching control algorithms and fill the gap in experimental research in this field in China,only use MATLAB to manually write the program.Research on device structure,data fusion and numerical simulation process of the experiment improving and design an oscillating buoy type WEC,and based on this device to conduct offline latching control experiments,analysis in depth and get summary of the phenomenon that occurred during the testing,obtained corresponding optimization strategies,and designed a simulation experimental system software based on the actual experimental process,providing a debugging platform for subsequent formal real-time experiments,quantifying the impact of experimental errors,and providing practical engineering experience.