DC / DC predictive control method under bistable probabilistic resonance wave power generation
By adopting long-term memory network and predictive control model in wave energy power generation system, combined with Runge-Kutta method and rolling optimization, the efficiency and accuracy problems of wave energy power generation system under non-fixed low-frequency vibration conditions are solved, and more efficient Voltage waveform stability and control effect.
Patent Information
- Application Number
- CN202411137308.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-12-03
AI Technical Summary
Under non-fixed low-frequency vibration conditions, traditional control methods lead to low power transmission efficiency, low control accuracy, and neglect the harmonics of each order, resulting in inaccurate modeling and slow response speed.
The DC/DC prediction control method is adopted for bistable probability resonant wave energy generation, and the long-term memory network combined with the predictive control model is used to verify the bistable characteristics of the mechanical system through the Runge-Kutta method, and the instantaneous power tracking of the three-phase rectifier is established. The high-frequency chain transmission power expression of the dual active bridge converter performs rolling optimization and feedback correction, and improves control accuracy and response speed.
Improve the power transmission efficiency and control accuracy of wave energy power generation systems under non-fixed low-frequency vibration conditions, solve the problems of inaccurate modeling and slow response speed in traditional methods, and achieve more efficient voltage waveform stability and control effect. .
Smart Images

Figure CN119070591A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of wave energy technology, and in particular to a DC / DC predictive control method under bistable probabilistic resonance wave energy power generation. Background Art
[0002] my country has nearly three million square kilometers of sea area and is a coastal country. Therefore, it attaches more importance to the development and utilization of the ocean, including the development of offshore oil and natural gas, mineral collection and the utilization of wave energy. However, the development of wave energy resources is much more difficult than that of land resources, which requires the use of a three-phase rectifier and a dual active bridge converter system. By inputting the non-fixed low-frequency AC signal emitted by the wave energy power system (equivalent to a bistable system) into the input end of the three-phase rectifier, a relatively stable DC voltage waveform is output through the direct power control of the rectifier. The waveform output by the rectifier is input into the dual active bridge converter in series at the rear end, and the input voltage can be operationally amplified through the high-frequency transformer between the primary and secondary sides. By performing predictive control based on the long short-term memory network on the output voltage of the dual active bridge converter, the output voltage can be kept high efficient while ensuring the stability of the output voltage waveform.
[0003] The long short-term memory network is a variant of the recurrent neural network. It solves the problem of the association between input information and can link previous information to the current task. Based on this, the long short-term memory network also solves the problem that the traditional recurrent neural network loses the ability to link distant information when the time interval between previous information and the current task is large. The long short-term memory network uses the "gate" structure to add, subtract or delete information to the cell state. The structure of the long short-term memory network mainly includes input gate, forget gate, output gate and cell state. Summary of the invention
[0004] According to the technical problems raised above, a DC / DC predictive control method under bistable probabilistic resonance wave power generation is provided. The present invention treats the wave power system as a bistable system, and uses a predictive control based on a long short-term memory network to estimate the dynamic response of the system under control in the future using a predictive model.
[0005] The technical means adopted by the present invention are as follows:
[0006] A DC / DC predictive control method for bistable probabilistic resonant wave energy generation, the specific steps comprising:
[0007] S1. Calculate the nonlinear equations of motion and perform numerical analysis using the Runge-Kutta method to verify the bistable characteristics of the mechanical system.
[0008] S2, based on the bistable characteristics in S1, the instantaneous power tracking of the three-phase rectifier is realized through direct power control;
[0009] S3, based on the control of the input current of the three-phase rectifier in S2, a long short-term memory network predictive control model is established;
[0010] S4, based on the long short-term memory network predictive control model in S3, perform offline and online training, predict the time domain, and fit the long short-term memory network of the dual active bridge converter;
[0011] S5. Perform rolling optimization and feedback correction based on the long short-term memory network predictive control model.
[0012] Furthermore, the step S1 specifically includes:
[0013] S11. Establish the nonlinear motion equation of the mass block in the bistable system:
[0014] For the wave energy power generation system, the mechanical part is equivalent to a bistable system to capture the energy of the waves. The wave energy power generation system consists of a mass block, a track and an elastic spring. The motion equation of the mass block is expressed as:
[0015]
[0016] Where m is the mass, x is d is the displacement starting from the central axis of symmetry, F is the elastic restoring force, and θ is the angle between the axis of the elastic spring and the horizontal direction;
[0017] Considering the spring restoring force, the equation of motion can be rewritten as:
[0018]
[0019] Where K is the spring constant, h is the vertical distance from the support point to the center of the block mass, l 0 is the initial length of the elastic spring, x t is the displacement starting from the symmetric center axis of the support;
[0020] S12. Calculate elastic potential energy:
[0021]
[0022] Among them, U represents the elastic potential energy of the spring, and x represents the relative displacement between the mass block and the support point;
[0023] The displacement equation formed by differentials can be expressed as:
[0024]
[0025] The root formula is as follows:
[0026]
[0027] Among them, x 1 、x 2 、x 3 is the extreme value of the elastic potential energy corresponding to the three static positions of the mass block in the motion model; x 1 and x 3 The elastic potential energy at a point is at a minimum, so the motion model corresponds only to the stopping position of the block-like mass at rest;
[0028] S13, the control equation is:
[0029]
[0030]
[0031] Among them, Δ, φ, and ψ represent the coefficient relationship. The potential energy calculation formula is Taylor expanded to obtain the following formula:
[0032]
[0033] S14. Numerical analysis using the Runge-Kutta method
[0034] After establishing the motion equation of the bistable motion model, considering the damping effect of friction, the motion equation is expressed as:
[0035]
[0036] Where c is the damping coefficient, Substitute the elastic force calculation and the elastic spring geometry into the excitation acceleration of the basic excitation:
[0037]
[0038] By analyzing the elastic potential energy, the bistable characteristics of the mechanical system are verified.
[0039] Furthermore, the step S2 specifically includes:
[0040] S21. Write the equation at the topological N node of the three-phase rectifier in the form of a state equation:
[0041]
[0042] Among them, s a 、s b 、s c They are the switching functions of phases abc, e a 、e b 、ec is a three-phase AC power supply, L and C are filter inductance and filter capacitor respectively, R L is the load resistance; R represents the equivalent resistance of the filter inductor;
[0043] S22, DQ stationary coordinate transformation:
[0044]
[0045] Among them, i α 、i β is the component of the current in the DQ stationary coordinate system, e α 、e β is the component of voltage in the DQ stationary coordinate system;
[0046] S23. The calculation formula of instantaneous power in three-phase abc stationary coordinate system is:
[0047]
[0048] Among them, p represents instantaneous active power, q represents instantaneous reactive power;
[0049] Instantaneous mathematical model of DQ stationary coordinate system:
[0050]
[0051] Where, ω represents the angular frequency of the three-phase rectifier system, u o Indicates the three-phase rectifier output voltage.
[0052] Furthermore, the step S3 specifically includes:
[0053] S31. Perform Fourier modeling on the dual active bridge and perform Fourier decomposition on the primary and secondary voltages of the high-frequency transformer:
[0054]
[0055] Among them, v H1 、v H2 is the transformer primary and secondary voltage, V 1 、V 2 Respectively represent the input and output voltages of the dual active bridge converter, T hs represents half of a switching cycle, d f It indicates the duty cycle of the phase shift time between the primary and secondary sides within half a switching cycle;
[0056] S32, high frequency link current expression is:
[0057]
[0058] Among them, L DABRepresents filter capacitor;
[0059] The dual active bridge inductor current is symmetrical within one cycle:
[0060]
[0061]
[0062] Where n represents the harmonic order; ω o Indicates the angular frequency of the DAB system at the fundamental frequency;
[0063] S33. Based on S31 and S32, the expression of high-frequency link transmission power in a switching cycle is obtained:
[0064]
[0065] S34, using a nonlinear autoregressive sliding average model to establish the input-output relationship of the dual active bridge converter system, as shown below:
[0066] V o (k+1)=f[V o (k),···,V o (k-g+1),d f (kd),···,d f (kd-j+1)]
[0067] Where f is the nonlinear autoregressive moving average model function; V o is the output voltage of the dual active bridge converter; d f is the ratio of the primary to secondary side shift of the transformer; k is the current moment; g is the output dimension of the system; j is the input dimension of the system; d is the delay time between the control input and the system response;
[0068] Use long short-term memory network to fit and approximate f:
[0069] V o (k+1)=f LSTM [V o (k),···,V o (k-g+1),d f (kd),···,d f (kd-j+1)]
[0070] Among them, f LSTM is the long short-term memory network model function; V o (k+1) is the predicted value at time k+1 predicted by the system based on the information at the current time and past times.
[0071] Furthermore, the step S4 specifically includes:
[0072] S41. Offline and online training based on long short-term memory network model:
[0073] The output voltage of the three-phase rectifier in S3 is used as the input power supply of the dual active bridge. The predictive control of the dual active bridge uses the original secondary side phase shift angle duty cycle as input and the secondary side output voltage as output. After the open-loop operation of the system is stable, different phase shift duty cycles and corresponding output voltages are sampled. After the sampling is completed, the data is sent to the long short-term memory network for offline training.
[0074] When the system is running in closed loop, the duty ratio of the primary and secondary side phase shift angle is directly output by the controller. For different inputs and different loads, the duty ratio of the primary and secondary side phase shift angle of the transformer is obtained by sampling the closed loop controller output. f , the output voltage under different operating conditions is compared with d f Input the long short-term memory network and train it to achieve predictive control under different working conditions;
[0075] S42, calculation prediction time domain and control time domain:
[0076] In predictive control, the number of steps used to predict the future output or state of the system in each operation cycle is defined as the prediction time domain, and the predicted output voltage is defined as V H , which is expressed as follows:
[0077]
[0078] The system requires the size of the prediction time domain H to be:
[0079] H≥d+M+1
[0080] Among them, M is the control time domain;
[0081] S43, Long short-term memory network fitting of dual active bridge converter:
[0082] Use historical data to train the long short-term memory network model to capture the law of changes in system output over time, verify the predictive performance of the network model on the reserved test set, and evaluate accuracy and generalization ability;
[0083] Assuming the output voltage of the dual active bridge converter is V o Depending on the historical input and output information of the converter, the fitting formula of the long short-term memory network for the dual active bridge converter is:
[0084] V o (k+1)=f LSTM [z(k),…,z(k-T+1)]
[0085] Where T represents the time step of the LSTM input vector; z(k) represents the input and output information of the LSTM at time k, and the expression of z(k) is as follows:
[0086] z(k)=[V o (k),d f (kd)]
[0087] When the system predicts the next H steps, the expression is:
[0088]
[0089] Here, i represents the time step.
[0090] Furthermore, the step S5 specifically includes:
[0091] S51. Rolling Optimization Based on Long Short-Term Memory Network Predictive Control:
[0092] Construct a value function to evaluate the effect of predictive control and optimize the response speed of the converter. Define the value function J as:
[0093]
[0094] Among them, V ref (k+i|k) represents the output voltage reference value at time k+i, r j is the suppression weight coefficient of the control quantity;
[0095] V ref (k+i|k) constitutes the reference trajectory sequence, which is expressed as:
[0096] V r (k)=[V ref (k+1k),…,V ref (k+Hk)]
[0097] Let R r =[r 0 ,…r M-1 ], then the value function is written in matrix form as:
[0098] J=[V r (k)-V H (k)] T [V r (k)-V H (k)]+[Δd f (k)] T R r [Δd f (k)]
[0099] The reference trajectory of the system output is defined as:
[0100]
[0101] For a given output voltage reference value V ref ,The goal of predictive control is to minimize the error between the output voltage and the reference value in the prediction time domain;
[0102] S52. Feedback correction based on long short-term memory network predictive control:
[0103] At time k+1, the actual output voltage of the system is V o (k+1), and the output voltage V predicted at time k o (k+1), the error e(k+1) of the long short-term memory network prediction model at time k+1 is:
[0104] e(k+1)=V o (k+1)-V o (k+1)
[0105] The error is used to correct the long short-term memory network prediction model, and the predicted voltage obtained by the long short-term memory network prediction model is rewritten as:
[0106]
[0107] Among them, λ i is the weighted correction coefficient of the predicted voltage obtained by the prediction model, where i = 1, 2, …, H.
[0108] Compared with the prior art, the present invention has the following advantages:
[0109] The DC / DC predictive control method under bistable probabilistic resonance wave power generation provided by the present invention is based on the predictive control of the long short-term memory network, which combines the long short-term memory network with the predictive control, and uses the predictive model to estimate the dynamic response of the system under control in the future. Then, combined with the set restrictions and goals, the optimal control law is calculated, and the predictive model is adjusted through real-time calculated data. In general, predictive control includes three main links: predictive model, rolling optimization and feedback correction.
[0110] The DC / DC predictive control method for bistable probabilistic resonant wave energy power generation provided by the present invention is analyzed based on the bistable model of the mechanical part, which proves that the proposed model has a larger amplitude in the steady state than the traditional cantilever beam bistable model. At the same time, it also effectively solves the problem of low efficiency of the monostable system under vibration power generation conditions such as wave energy power generation.
[0111] The DC / DC predictive control method under bistable probabilistic resonant wave energy power generation provided by the present invention performs predictive control based on a long short-term memory network on the dual active bridge converter connected in series at the rear end of the rectifier, thereby solving the problem of inaccurate model caused by traditional fundamental wave modeling ignoring each harmonic; at the same time, it also solves the shortcoming of slow response speed of traditional PI control.
[0112] The DC / DC predictive control method under bistable probabilistic resonance wave power generation provided by the present invention has low power transmission efficiency and low control accuracy when the system is in a non-fixed low-frequency wave power generation condition. For the control system of traditional fundamental wave modeling, ignoring each harmonic under such conditions can easily cause problems such as inaccurate modeling, inaccurate control effect and slow response speed. Even if the input voltage signal at the front end of the system can ensure that the instantaneous active and reactive power are in a relatively stable state through direct power control of the three-phase rectifier, the output voltage is still not ideal. Based on this problem, predictive control based on long short-term memory network is introduced. The predictive model of long short-term memory network avoids the problem of inaccurate modeling brought by traditional mathematical modeling. The rolling optimization based on the golden section method and feedback correction make the control response speed faster and the control accuracy higher. It can be widely used in engineering practice.
[0113] Based on the above reasons, the present invention can be widely promoted in the field of wave energy technology. BRIEF DESCRIPTION OF THE DRAWINGS
[0114] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0115] Figure 1 This is a flow chart of the DC / DC predictive control method under bistable probabilistic resonance wave energy power generation in the present invention.
[0116] Figure 2 This is the overall framework diagram of the DC / DC predictive control system under the bistable probabilistic resonance wave energy power generation in the present invention.
[0117] Figure 3 It is a non-fixed low-frequency three-phase voltage waveform output by the wave energy power generation in the present invention.
[0118] Figure 4 Schematic diagram of a wave energy bistable motion system in an embodiment of the present invention.
[0119] Figure 5 This is a block diagram of the direct power control of the three-phase rectifier in the present invention.
[0120] Figure 6 This is the direct power control effect of the three-phase rectifier in the embodiment of the present invention.
[0121] Figure 7 The three-phase rectifier direct power control output voltage waveform in the embodiment of the present invention.
[0122] Figure 8 It is the multi-step recursive model of the long short-term memory network in the present invention.
[0123] Fig. 9 This is the optimization process of the golden section method in the present invention.
[0124] Fig.10 This is a block diagram of the long short-term memory network predictive control in the present invention.
[0125] Fig.11 1 is a comparison of the output voltage during the medium voltage startup process in the embodiment of the present invention.
[0126] Fig.12 1 is a comparison of the output voltage when the load changes suddenly in the embodiment of the present invention (online). DETAILED DESCRIPTION
[0127] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.
[0128] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0129] like Figure 1 As shown, the present invention provides a DC / DC predictive control method under bistable probabilistic resonance wave energy generation, and the specific steps include:
[0130] S1. Calculate the nonlinear equations of motion and perform numerical analysis using the Runge-Kutta method to verify the bistable characteristics of the mechanical system.
[0131] In specific implementation, as a preferred embodiment of the present invention, the step S1 specifically includes:
[0132] S11. Establish the nonlinear motion equation of the mass block in the bistable system:
[0133] For the wave energy power generation system, the mechanical part is equivalent to a bistable system to capture the energy of the waves and improve the power generation efficiency of the system as much as possible. The wave energy power generation system consists of a mass block, a track and an elastic spring. The motion equation of the mass block is expressed as:
[0134]
[0135] Where m is the mass, x is d is the displacement starting from the central axis of symmetry, F is the elastic restoring force, and θ is the angle between the axis of the elastic spring and the horizontal direction;
[0136] Considering the spring restoring force, the equation of motion can be rewritten as:
[0137]
[0138] Where K is the spring constant, h is the vertical distance from the support point to the center of the block mass, l 0 is the initial length of the elastic spring, x t is the displacement starting from the symmetric center axis of the support;
[0139] S12. Calculate elastic potential energy:
[0140]
[0141] Among them, U represents the elastic potential energy of the spring, and x represents the relative displacement between the mass block and the support point;
[0142] The displacement equation formed by differentials can be expressed as:
[0143]
[0144] The root formula is as follows:
[0145]
[0146] Among them, x 1 、x 2 、x 3 is the extreme value of the elastic potential energy corresponding to the three static positions of the mass block in the motion model; x 1 and x3 The elastic potential energy at a point is at a minimum, so the motion model corresponds only to the stopping position of the block-like mass at rest;
[0147] S13, the control equation is:
[0148]
[0149]
[0150] Among them, Δ, φ, and Ψ represent the coefficient relationship. Taylor expansion of the potential energy calculation formula gives the following formula:
[0151]
[0152] S14. Numerical analysis using the Runge-Kutta method
[0153] After establishing the motion equation of the bistable motion model, considering the damping effect of friction, the motion equation is expressed as:
[0154]
[0155] Where c is the damping coefficient, Substitute the elastic force calculation and the elastic spring geometry into the excitation acceleration of the basic excitation:
[0156]
[0157] The bistable characteristics of the mechanical system are verified by analyzing the elastic potential energy. A random excitation load is applied to the mass block, and a weak periodic excitation load is applied at the same time, so that the total energy exceeds the elastic potential energy at the position of the symmetry axis. The mass block will pass through the symmetry axis and jump to both sides, forming a bistable vibration model. Therefore, this bistable motion model has the characteristics of maintaining stable vibration under large amplitude.
[0158] S2, based on the bistable characteristics in S1, the instantaneous power tracking of the three-phase rectifier is realized through direct power control;
[0159] In specific implementation, as a preferred embodiment of the present invention, step S2 specifically includes:
[0160] S21. Write the equation at the topological N node of the three-phase rectifier in the form of a state equation:
[0161]
[0162] Among them, s a 、s b 、s c They are the switching functions of phases abc, e a、e b 、e c is a three-phase AC power supply, L and C are filter inductance and filter capacitor respectively, R L is the load resistance; R represents the equivalent resistance of the filter inductor;
[0163] S22, DQ stationary coordinate transformation:
[0164]
[0165] Among them, i α 、i β is the component of the current in the DQ stationary coordinate system, e α 、e β is the component of voltage in the DQ stationary coordinate system;
[0166] S23. The calculation formula of instantaneous power in three-phase abc stationary coordinate system is:
[0167]
[0168] Among them, p represents instantaneous active power, q represents instantaneous reactive power;
[0169] Instantaneous mathematical model of DQ stationary coordinate system:
[0170]
[0171] Where, ω represents the angular frequency of the three-phase rectifier system, u o Indicates the output voltage of the three-phase rectifier system.
[0172] Direct power control uses instantaneous power theory to achieve precise control of the rectifier input current. The core of instantaneous power theory is that it does not rely on the sinusoidal waveforms of voltage and current, but directly defines active power and reactive power based on instantaneous values. The system structure of the direct power control method usually adopts double closed-loop control, in which the DC side voltage is used as the outer loop and the instantaneous power control is used as the inner loop. When the AC side voltage is constant, the instantaneous input current is controlled by controlling the instantaneous active power and reactive power of the rectifier to achieve the required power factor and power flow direction.
[0173] S3, based on the control of the input current of the three-phase rectifier in S2, a long short-term memory network predictive control model is established;
[0174] In specific implementation, as a preferred embodiment of the present invention, step S3 specifically includes:
[0175] S31. Perform Fourier modeling on the dual active bridge and perform Fourier decomposition on the primary and secondary voltages of the high-frequency transformer:
[0176]
[0177] Among them, v H1 、v H2 is the transformer primary and secondary voltage, V 1 、V 2 Respectively represent the input and output voltages of the dual active bridge converter, T hs represents half of a switching cycle, d f It indicates the duty cycle of the phase shift time between the primary and secondary sides within half a switching cycle;
[0178] S32, high frequency link current expression is:
[0179]
[0180] Among them, L DAB Represents filter capacitor;
[0181] The dual active bridge inductor current is symmetrical within one cycle:
[0182]
[0183]
[0184] Where n represents the harmonic order; ω o Indicates the angular frequency of the DAB system at the fundamental frequency;
[0185] S33. Based on S31 and S32, the expression of high-frequency link transmission power in a switching cycle is obtained:
[0186]
[0187] S34, using a nonlinear autoregressive sliding average model to establish the input-output relationship of the dual active bridge converter system, as shown below:
[0188] V o (k+1)=f[V o (k),···,V o (k-g+1),d f (kd),···,d f (kd-j+1)]
[0189] Where f is the nonlinear autoregressive moving average model function; V o is the output voltage of the dual active bridge converter; d f is the primary-to-secondary shift phase of the transformer; k is the current moment; g is the output dimension of the system; j is the input dimension of the system; d is the delay time between the control input and the system response.
[0190] Use long short-term memory network to fit and approximate f:
[0191] V o (k+1)=f LSTM [V o (k),···,V o (k-g+1),d f (kd),···,d f (kd-j+1)]
[0192] Among them, f LSTM is the long short-term memory network model function; V o (k+1) is the predicted value at time k+1 predicted by the system based on the information at the current time and past times.
[0193] Different from the traditional transfer function-based modeling method, the LSTM-based modeling method is implemented through a data-driven approach, relying on a large amount of historical data to train the model and find the input-output relationship.
[0194] S4, based on the long short-term memory network predictive control model in S3, perform offline and online training, predict the time domain, and fit the long short-term memory network of the dual active bridge converter;
[0195] In specific implementation, as a preferred embodiment of the present invention, step S4 specifically includes:
[0196] The predictive control method based on the long short-term memory network combines the long short-term memory network prediction model, the golden section rolling optimization strategy and the feedback correction mechanism. It effectively improves the problems of inaccurate model, low control accuracy and slow response speed of the traditional control method.
[0197] S41. Offline and online training based on long short-term memory network model:
[0198] The output voltage of the three-phase rectifier in S3 is used as the input power supply of the dual active bridge. The predictive control of the dual active bridge uses the original secondary side phase shift angle duty cycle as input and the secondary side output voltage as output. After the open-loop operation of the system is stable, different phase shift duty cycles and corresponding output voltages are sampled. After the sampling is completed, the data is sent to the long short-term memory network for offline training.
[0199] When the system is running in closed loop, the duty ratio of the primary and secondary side phase shift angle is directly output by the controller. For different inputs and different loads, the duty ratio of the primary and secondary side phase shift angle of the transformer is obtained by sampling the closed loop controller output. f , the output voltage under different operating conditions is compared with d f Input the long short-term memory network and train it to achieve predictive control under different working conditions;
[0200] S42, calculation prediction time domain and control time domain:
[0201] In predictive control, the number of steps used to predict the future output or state of the system in each operation cycle is defined as the prediction time domain, and the predicted output voltage is defined as V H , which is expressed as follows:
[0202]
[0203] The system requires the size of the prediction time domain H to be:
[0204] H≥d+M+1
[0205] Among them, M is the control time domain;
[0206] S43, Long short-term memory network fitting of dual active bridge converter:
[0207] The specific method of online prediction is to collect input data (here the input data is the original secondary side phase shift duty cycle df) and output response (here the output response is the secondary side output voltage Vo). The input and output data are sent to the long short-term memory network to pre-process the collected data, including normalization, denoising, etc., to improve the prediction performance of the model.
[0208] Use historical data to train the long short-term memory network model to capture the law of changes in system output over time, verify the predictive performance of the network model on the reserved test set, and evaluate accuracy and generalization ability;
[0209] In the actual control process, the input quantity obtained in real time is used as the input of the network model for online prediction. The predicted output quantity can be used as the basis for control decision-making.
[0210] Assuming the output voltage of the dual active bridge converter is V o Depending on the historical input and output information of the converter, the fitting formula of the long short-term memory network for the dual active bridge converter is:
[0211] V o (k+1)=f LSTM [z(k),…,z(k-T+1)]
[0212] Among them, T represents the time step of the LSTM input vector, that is, the input and output information of the previous moments contained in the LSTM; z(k) represents the input and output information of the LSTM at moment k, and the expression of z(k) is as follows:
[0213] z(k)=[V o (k),d f (kd)]
[0214] When the system predicts the next H steps, the expression is:
[0215]
[0216] Here, i represents the time step.
[0217] S5. Perform rolling optimization and feedback correction based on the long short-term memory network predictive control model.
[0218] In specific implementation, as a preferred embodiment of the present invention, step S5 specifically includes:
[0219] S51. Rolling Optimization Based on Long Short-Term Memory Network Predictive Control:
[0220] Output voltage control mainly involves two aspects: first, the output voltage needs to be adjusted quickly and accurately to reach the reference value and be as error-free as possible; second, ensure that the adjustment process is smooth to avoid excessive overshoot and oscillation.
[0221] Construct a value function to evaluate the effect of predictive control and optimize the response speed of the converter. Define the value function J as:
[0222]
[0223] Among them, V ref (k+i|k) represents the output voltage reference value at time k+i, r j is the suppression weight coefficient of the control quantity; when r j When r is zero, the value function is only committed to reducing the error of the output voltage, and the size of the control quantity is not considered in the optimization process. j When it is large, it will limit the range of control amount.
[0224] V ref (k+i|k) forms a reference trajectory sequence, which makes the output voltage rising process smoother, and the expression is:
[0225] V r (k)=[V ref (k+1|k),…,V ref (k+H|k)],
[0226] Let R r =[r 0 ,…r M-1 ], then the value function is written in matrix form as:
[0227] J=[V r (k)-V H (k)] T [V r (k)-V H (k)]+[Δdf (k)] T R r [Δd f (k)]
[0228] The reference trajectory of the system output is defined as:
[0229]
[0230] For a given output voltage reference value V ref ,The goal of predictive control is to minimize the error between the output voltage and the reference value in the prediction ,time domain and ensure the smoothness of the regulation process.
[0231] Among them, r(k) is the set value of the system at time k, ɑ is the softening factor of the output voltage rising process, which ranges from 0 to 1. The larger the value of ɑ, the slower the output voltage follows the reference value, and the stability is improved; the smaller the value, the faster the output voltage follows the reference value, but it is easy to cause system instability.
[0232] The commonly used rolling optimization method in predictive control is the golden section method, which is a method of finding the optimal solution by the rule of gradual elimination. The principle involves setting two test points in the initial search range and gradually eliminating the area of non-optimal solutions by comparing the function values of these two points, while retaining the part containing the optimal solution. This process is repeated until the search interval is reduced to a certain length, and then the midpoint of the interval is used as an approximation of the optimal value. When selecting the test points, the golden section point is used to ensure that the proportion of intervals eliminated in each iteration is consistent, ensuring that the number of iterations to find the optimal solution remains unchanged.
[0233] S52. Feedback correction based on long short-term memory network predictive control:
[0234] At time k+1, the actual output voltage of the system is V o (k+1), and the output voltage V predicted at time k o (k+1), the error e(k+1) of the long short-term memory network prediction model at time k+1 is:
[0235] e(k+1)=V o (k+1)-V o (k+1)
[0236] The error is used to correct the long short-term memory network prediction model, and the predicted voltage obtained by the long short-term memory network prediction model is rewritten as:
[0237]
[0238] Among them, λ iis the weighted correction coefficient of the predicted voltage obtained by the prediction model, where i = 1, 2, ..., H. The smaller λ is, the slower the system reduces the error, and vice versa, the faster the system reduces the error. However, the larger λ is, the more likely it is to cause system oscillation.
[0239] During implementation, a bistable characteristic of the mechanical part of the wave energy uses a predictive control based on the long short-term memory network to control the output voltage of the dual-active bridge converter, which solves the problems of inaccurate modeling results and neglect of harmonics caused by traditional fundamental wave modeling, improves the shortcomings of slow PI control response speed, and improves the performance of the algorithm in solving complex optimization problems.
[0240] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.
[0241] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0242] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.
[0243] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.
[0244] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0245] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.
[0246] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A DC / DC predictive control method for bistable probabilistic resonance wave energy generation, characterized in that: The specific steps include: S1. Calculate the nonlinear equations of motion and perform numerical analysis using the Runge-Kutta method to verify the bistable characteristics of the mechanical system. S2, based on the bistable characteristics in S1, the instantaneous power tracking of the three-phase rectifier is realized through direct power control; S3, based on the control of the input current of the three-phase rectifier in S2, a long short-term memory network predictive control model is established; S4, based on the long short-term memory network predictive control model in S3, perform offline and online training, predict the time domain, and fit the long short-term memory network of the dual active bridge converter; S5. Perform rolling optimization and feedback correction based on the long short-term memory network predictive control model.
2. The DC / DC predictive control method for bistable probabilistic resonance wave energy generation according to claim 1 is characterized in that: The step S1 specifically includes: S11. Establish the nonlinear motion equation of the mass block in the bistable system: For the wave energy power generation system, the mechanical part is equivalent to a bistable system to capture the energy of the waves. The wave energy power generation system consists of a mass block, a track and an elastic spring. The motion equation of the mass block is expressed as: Where m is the mass, x is d is the displacement starting from the central axis of symmetry, F is the elastic restoring force, and θ is the angle between the axis of the elastic spring and the horizontal direction; Considering the spring restoring force, the equation of motion can be rewritten as: Where K is the spring constant, h is the vertical distance from the support point to the center of the block, l0 is the initial length of the elastic spring, and x t is the displacement starting from the symmetric center axis of the support; S12. Calculate elastic potential energy: Among them, U represents the elastic potential energy of the spring, and x represents the relative displacement between the mass block and the support point; The displacement equation formed by differentials can be expressed as: The root formula is as follows: Among them, x1, x2, and x3 are the extreme values of the elastic potential energy corresponding to the three static positions of the mass block in the motion model; the elastic potential energy at points x1 and x3 is the minimum value, so the motion model only corresponds to the stop position of the block mass block in the static state; S13, the control equation is: Among them, Δ, φ, and ψ represent the coefficient relationship. The potential energy calculation formula is Taylor expanded to obtain the following formula: S14. Numerical analysis using the Runge-Kutta method After establishing the motion equation of the bistable motion model, considering the damping effect of friction, the motion equation is expressed as: Where c is the damping coefficient, Substitute the elastic force calculation and the elastic spring geometry into the excitation acceleration of the basic excitation: By analyzing the elastic potential energy, the bistable characteristics of the mechanical system are verified.
3. The DC / DC predictive control method for bistable probabilistic resonance wave energy generation according to claim 1 is characterized in that: The step S2 specifically includes: S21. Write the equation at the topological N node of the three-phase rectifier in the form of a state equation: Among them, s a 、s b 、s c are the switching functions of phases abc respectively, e a 、e b 、e c is a three-phase AC power supply, L and C are filter inductance and filter capacitor respectively, R L is the load resistance; R represents the equivalent resistance of the filter inductor; S22, DQ stationary coordinate transformation: Among them, i α 、i β is the component of the current in the DQ stationary coordinate system, e α 、e β is the component of voltage in the DQ stationary coordinate system; S23. The calculation formula of instantaneous power in three-phase abc stationary coordinate system is: Among them, p represents instantaneous active power, q represents instantaneous reactive power; Instantaneous mathematical model of DQ stationary coordinate system: Where, ω represents the angular frequency of the three-phase rectifier system, u o Indicates the output voltage of the three-phase rectifier system.
4. The DC / DC predictive control method for bistable probabilistic resonance wave energy generation according to claim 1 is characterized in that: The step S3 specifically includes: S31. Perform Fourier modeling on the dual active bridge and perform Fourier decomposition on the primary and secondary voltages of the high-frequency transformer: Among them, v H1 、v H2 is the transformer primary and secondary voltages, V1 and V2 represent the input and output voltages of the dual active bridge converter, respectively. hs represents half of a switching cycle, d f It indicates the duty cycle of the phase shift time between the primary and secondary sides within half a switching cycle; S32, high frequency link current expression is: Among them, L DAB Represents filter capacitor; The dual active bridge inductor current is symmetrical within one cycle: Where n represents the harmonic order; ω o Indicates the angular frequency of the DAB system at the fundamental frequency; S33. Based on S31 and S32, the expression of high-frequency link transmission power in a switching cycle is obtained: S34, using a nonlinear autoregressive sliding average model to establish the input-output relationship of the dual active bridge converter system, as shown below: V o (k+1)=f[V o (k),···,V o (k-g+1),d f (k-d),···,d f (k-d-j+1)] Where f is the nonlinear autoregressive moving average model function; V o is the output voltage of the dual active bridge converter; d f is the transformer primary-secondary shift ratio; k is the current moment; g is the output dimension of the system; j is the input dimension of the system; d is the delay time between the control input and the system response; Use long short-term memory network to fit and approximate f: V o (k+1)=f LSTM [V o (k),···,V o (k-g+1),d f (k-d),···,d f (k-d-j+1)] Among them, f LSTM is the long short-term memory network model function; V o (k+1) is the predicted value at time k+1 predicted by the system based on the information at the current time and past times.
5. The DC / DC predictive control method for bistable probabilistic resonance wave energy generation according to claim 1 is characterized in that: The step S4 specifically includes: S41. Offline and online training based on long short-term memory network model: The output voltage of the three-phase rectifier in S3 is used as the input power supply of the dual active bridge. The predictive control of the dual active bridge uses the original secondary side phase shift angle duty cycle as input and the secondary side output voltage as output. After the open-loop operation of the system is stable, different phase shift duty cycles and corresponding output voltages are sampled. After the sampling is completed, the data is sent to the long short-term memory network for offline training. When the system is running in closed loop, the duty ratio of the primary and secondary side phase shift angle is directly output by the controller. For different inputs and different loads, the duty ratio of the primary and secondary side phase shift angle of the transformer is obtained by sampling the closed loop controller output. f , the output voltage under different operating conditions is compared with d f Input the long short-term memory network and train it to achieve predictive control under different working conditions; S42, calculation prediction time domain and control time domain: In predictive control, the number of steps used to predict the future output or state of the system in each operation cycle is defined as the prediction time domain, and the predicted output voltage is defined as V H , which is expressed as follows: The system requires the size of the prediction time domain H to be: H≥d+M+1 Among them, M is the control time domain; S43, Long short-term memory network fitting of dual active bridge converter: Use historical data to train the long short-term memory network model to capture the law of changes in system output over time, verify the predictive performance of the network model on the reserved test set, and evaluate accuracy and generalization ability; Assuming the output voltage of the dual active bridge converter is V o Depending on the historical input and output information of the converter, the fitting formula of the long short-term memory network for the dual active bridge converter is: V o (k+1)=f LSTM [z(k),…,z(k-T+1)] Where T represents the time step of the LSTM input vector; z(k) represents the input and output information of the LSTM at time k, and the expression of z(k) is as follows: z(k)=[V o (k),d f (k-d)] When the system predicts the next H steps, the expression is: Here, i represents the time step.
6. The DC / DC predictive control method for bistable probabilistic resonance wave energy generation according to claim 1 is characterized in that: The step S5 specifically includes: S51. Rolling Optimization Based on Long Short-Term Memory Network Predictive Control: Construct a value function to evaluate the effect of predictive control and optimize the response speed of the converter. Define the value function J as: Among them, V ref (k+i|k) represents the output voltage reference value at time k+i, r j is the suppression weight coefficient of the control quantity; V ref (k+i|k) constitutes the reference trajectory sequence, which is expressed as: V r (k)=[V ref (k+1k),…,V ref (k+H|k)] Let R r =[r0,…r M-1 ], then the value function is written in matrix form as: J=[V r (k)-V H (k)] T [V r (k)-V H (k)]+[Δd f (k)] T R r [Δd f (k)] The reference trajectory of the system output is defined as: For a given output voltage reference value V ref ,The goal of predictive control is to minimize the error between the output voltage and the reference value in the prediction time domain; S52. Feedback correction based on long short-term memory network predictive control: At time k+1, the actual output voltage of the system is V o (k+1), and the output voltage V predicted at time k o (k+1), the error e(k+1) of the long short-term memory network prediction model at time k+1 is: e(k+1)=V o (k+1)-V o (k+1) The error is used to correct the long short-term memory network prediction model, and the predicted voltage obtained by the long short-term memory network prediction model is rewritten as: Among them, λ i is the weighted correction coefficient of the predicted voltage obtained by the prediction model, where i = 1, 2, …, H.
Citation Information
Cited By
CLLC resonant converter soft start method based on edge neural network prediction
CN119891731A
