Gradient iteration based data-driven modeling method for complex medium UWPT system

Abstract

The application relates to the technical field of underwater wireless power transfer (UWPT), and particularly discloses a complex medium UWPT system data-driven modeling method based on gradient iteration, which comprises the following steps: firstly, determining a circuit model of an underwater wireless power transfer system (UWPT system); then, establishing a controlled autoregressive model of the circuit model; then, defining a parameter vector and an information vector of the UWPT system; and finally, sampling data with a length of L, and performing gradient iteration based on the data and the established controlled autoregressive model; and outputting current system parameters after the iteration is completed. The application adopts data-driven modeling, uses a very simple black box model, directly bases on input and output data of the system, and establishes a controlled autoregressive model of the system by using a gradient iteration method. The application greatly simplifies the modeling steps of the system in an actual industrial environment, significantly reduces the modeling workload of the system, and improves the modeling precision and reliability of the system in a complex medium environment.

Description

Technical Field

[0001] This invention relates to the field of Underwater Wireless Power Transfer (UWPT) technology, and in particular to a data-driven modeling method for complex medium UWPT systems based on gradient iteration. Background Technology

[0002] As land resources are gradually depleted, attention is shifting to the development of marine resources. Currently, underwater engineering commonly uses wired cables for power supply, which still has significant room for improvement in terms of convenience and safety. Underwater wireless power transfer (UWPT) technology offers a promising solution to this problem. A wireless power transfer system mainly consists of the following components:

[0003] ① Primary energy emission terminal

[0004] The transmitter of a wireless power transfer system typically consists of a power supply, a rectifier and filter circuit, a high-frequency inverter circuit, and a resonant circuit. Powered by mains frequency electricity, the system first converts the mains frequency power into the required DC power through a rectifier and filter circuit. Then, it enters a high-frequency inverter circuit composed of a full-bridge or half-bridge inverter, which converts it into a high-frequency square wave signal. Finally, the resonant circuit generates a high-frequency alternating magnetic field to transfer energy. It is important to note that the resonant circuit's frequency should match the inverter's operating frequency.

[0005] ② Coupled magnetic field section

[0006] The function of this part is to transfer energy. When the high-frequency alternating current on the primary side passes through the coil, it generates a high-frequency magnetic field, which is used to realize wireless energy transfer between the transmitter and receiver.

[0007] ③ Secondary side energy receiver

[0008] The receiving end typically consists of a resonant circuit, a rectifier and filter circuit, and a load. The function of the resonant circuit is to improve the energy reception efficiency of the secondary and tertiary sides. The current passing through the resonant circuit is alternating current (AC), which needs to be rectified and filtered to convert it into direct current (DC) to power the load.

[0009] Building a model of the object is the prerequisite and foundation for control and stability analysis. It can be used for simulation, system behavior analysis, analysis of the system's intrinsic mechanisms, control, prediction, monitoring, and fault diagnosis.

[0010] Currently, the traditional method for system modeling is mechanistic analysis (analytical method). This method typically requires analyzing the motion laws of the process and deriving a mathematical model describing the system using the conservation and continuity of matter and energy proposed by various disciplines. This type of modeling is sometimes called white-box modeling. It requires a relatively mature understanding of the system's technology and its mechanisms, and that the theories, theorems, and data involved are relatively accurate. When performing theoretical modeling under complex conditions, the mathematical characteristics of each part must be clear and concise; otherwise, the problem will become overly complicated. Furthermore, the parameters and structure of the actual process are not always fully understood. Under complex media conditions, it is difficult to obtain the system's structure and parameters due to the influence of various parameters. Therefore, it is difficult to perform mechanistic modeling of the system under the conditions discussed in this patent. Summary of the Invention

[0011] This invention provides a data-driven modeling method for UWPT systems in complex media based on gradient iteration. The technical problem it solves is: how to model UWPT systems under complex media conditions and simply estimate the system's structure and parameters.

[0012] To address the above technical problems, this invention provides a data-driven modeling method for complex medium UWPT systems based on gradient iteration, comprising the following steps:

[0013] S1. Determine the circuit model of the underwater wireless power transfer system, i.e., the UWPT system;

[0014] S2. The controlled autoregressive model of the circuit model is established as follows:

[0015]

[0016] Where {v(t)} and {θ(t)} are the input phase shift angle and output voltage sequence of the WPT system, respectively.

[0017] {n(t)} has a mean of 0 and a variance of σ. 2 Unmeasurable white noise sequence, z -1 Let A(z) be the unit shift operator, and let A(z) be the polynomial to be identified for v(t). a It is the order of the polynomial A(z). The first to nth order of polynomial A(z) a The coefficients of order n, B(z) are the polynomial coefficients of θ(t) to be identified, and n b It is the order of the polynomial B(z). The first to nth order of polynomial B(z) b The coefficient of the order;

[0018] S3. Define the parameter vector of the UWPT system. n = n a +nb R represents a real number; the information vector of the UWPT system is also defined. The controlled autoregressive model is expressed as:

[0019]

[0020] S4. Sample data of length L, and perform gradient iteration based on this data and the controlled autoregressive model to obtain parameters.

[0021] Furthermore, step S4 specifically includes the following steps:

[0022] S41. Based on data of length L, define the stacked output vector V(L) and the stacked information matrix Φ(L) as follows:

[0023]

[0024] Where v(1) v(2) ... v(L) are v(t) corresponding to data of length L. For data of length L

[0025] S42. Define about the parameter vector Static criterion function:

[0026]

[0027] Where || represents the norm;

[0028] S43. Let k = 1, 2, 3, ... be the iteration variables, and define the parameter vector. The k-th iteration estimate for:

[0029]

[0030] in, For the k-th iteration For the k-th iteration

[0031] S44, to minimize The negative gradient is used for iterative search. When the iteration termination condition is met, the iteration ends and the parameters at this point are output.

[0032] Furthermore, in step S44, an iterative search is performed using a negative gradient, according to the formula:

[0033]

[0034] Where μ>0 is the step size factor, grad[] represents the gradient, and k-1 represents the k-1th iteration.

[0035] 4. The data-driven modeling method for complex medium UWPT systems based on gradient iteration according to claim 3, characterized in that the step size factor μ must satisfy:

[0036]

[0037] in, This is the upper bound of the value of μ.

[0038] Further, in step S44, the iteration termination condition is:

[0039]

[0040] Where ε represents the accuracy of parameter estimation.

[0041] Further, step S44 specifically includes the following steps:

[0042] S441. Initialization: Set all variables to 0, let the number of iterations k = 1, and give the data length L and the parameter estimation accuracy ε.

[0043] S442. Let k0 = 1:L, collect observation data v(k0Δt) and Where Δt is the sampling interval;

[0044] S443, based on the sampled data v(k0Δt) and Construct the stacked output vector V(L) and the stacked information matrix Φ(L), and determine the iteration step size μ;

[0045] S444, Use Refresh the parameter estimation vector;

[0046] S445, Determine if the condition is met. If so, output the parameters at this time. Otherwise, return to step S444 to proceed to the next iteration.

[0047] Furthermore, the circuit model of the UWPT system includes a primary-side energy transmitter and a secondary-side energy receiver. The primary-side energy transmitter includes a DC power supply, an inverter, and a primary resonant structure connected in sequence. The secondary-side energy receiver includes a secondary resonant structure, a rectifier bridge, and a filter capacitor C connected in sequence. f and load R L .

[0048] Furthermore, the inverter includes a first full-bridge inverter and a second full-bridge inverter connected in parallel with the DC power supply.

[0049] Furthermore, the primary resonant structure includes a first primary resonant circuit connected to the first full-bridge inverter and a second primary resonant circuit connected to the second full-bridge inverter.

[0050] Furthermore, the first primary resonant circuit and the second primary resonant circuit have the same structure, both adopting an LCC-type compensation topology; the secondary resonant structure adopts an S-type compensation topology.

[0051] The present invention provides a data-driven modeling method for complex medium UWPT systems based on gradient iteration. First, the circuit model of the underwater wireless power transfer system (UWPT) is determined. Then, a controlled autoregressive model of the circuit model is established. Next, the parameter vector and information vector of the UWPT system are defined. Finally, gradient iteration is performed using data of length L and the established controlled autoregressive model. After the iteration is completed, the current system parameters are output. This invention employs data-driven modeling, using an extremely simple black-box model. It directly establishes a controlled autoregressive model of the system based on the system's input and output data using a gradient iteration method. This greatly simplifies the system modeling steps in actual industrial environments, significantly reduces the workload of system modeling, and improves the modeling accuracy and reliability of the system in complex media environments. Attached Figure Description

[0052] Figure 1 This is a circuit model diagram of the UWPT system provided in the embodiments of the present invention;

[0053] Figure 2 This is provided by the embodiments of the present invention. Figure 1 The equivalent circuit diagram;

[0054] Figure 3 This is a flowchart of parameter identification using collected data provided in an embodiment of the present invention;

[0055] Figure 4 This is a waveform diagram of the input signal during the experiment provided in an embodiment of the present invention;

[0056] Figure 5 This is a comparison diagram of the actual output waveform of the system and the output waveform of the data-driven model provided in the embodiments of the present invention. Detailed Implementation

[0057] The embodiments of the present invention are described in detail below with reference to the accompanying drawings. The embodiments are given for illustrative purposes only and should not be construed as limiting the present invention. The accompanying drawings are for reference and illustration only and do not constitute a limitation on the scope of patent protection of the present invention, because many changes can be made to the present invention without departing from the spirit and scope of the present invention.

[0058] Due to the influence of saltwater conductivity in complex media such as underwater environments, as well as electromagnetic interference between various electrical components, the transfer function model of a wireless power transfer system is extremely difficult to determine using traditional circuit models. To address this, this invention employs data-driven modeling, which simplifies the process of determining the system transfer function. The advantages of the data-driven modeling method are as follows:

[0059] First, since the input and output signals of a system are generally measurable, the dynamic characteristics of the system must be reflected in these input and output data. In the complex medium mentioned in this patent, even if the structure and parameters of the system are unknown, the system's relationship can be deduced from the information provided by the system's operation and experimental data obtained through multiple measurements, thereby establishing the system's mathematical model.

[0060] Second, data-driven fuzzy system modeling does not require prior knowledge of the object, but directly models it based on the object's input and output data, which greatly reduces the knowledge requirements of users and has great potential in solving the control of highly nonlinear and severely uncertain systems.

[0061] Third, this data-driven modeling approach can express the complex relationships between inputs and outputs with fewer rules.

[0062] Specifically, the gradient-iteration-based data-driven modeling method for complex medium UWPT systems provided in this embodiment of the invention includes steps S1 to S4.

[0063] S1. Determine the circuit model of the underwater wireless power transfer system, i.e., the UWPT system.

[0064] by Figure 2 Taking the UWPT system shown as an example, the circuit model of the UWPT system includes a primary-side energy transmitter and a secondary-side energy receiver. The primary-side energy transmitter includes a DC power supply, an inverter, and a primary resonant structure connected in sequence. The secondary-side energy receiver includes a secondary resonant structure, a rectifier bridge, and a filter capacitor C connected in sequence. f and load R L The inverter includes a first full-bridge inverter (inverter a) and a second full-bridge inverter (inverter b) connected in parallel with the DC power supply. The primary resonant structure includes a first primary resonant circuit connected to the first full-bridge inverter and a second primary resonant circuit connected to the second full-bridge inverter. The first and second primary resonant circuits have the same structure, both using an LCC-type compensated topology; the secondary resonant structure uses an S-type compensated topology.

[0065] Figure 2 The equivalent circuit diagram shown is as follows: Figure 3 As shown.

[0066] S2. Establish a controlled autoregressive model for the circuit model.

[0067] for Figure 2 The circuit shown can be used to establish a controlled autoregressive model (CAR) for the dual LCC-WPT system as follows:

[0068]

[0069] Where {v(t)} and {θ(t)} are the input phase shift angle and output voltage sequence of the WPT system, respectively, and {n(t)} is a zero-mean, variance σ. 2 Unmeasurable white noise sequence, z -1 Let A(z) be the unit shift operator, and let A(z) be the polynomial to be identified for v(t). a It is the order of the polynomial A(z). The first to nth order of polynomial A(z) a The coefficients of order n, B(z) are the polynomial coefficients of θ(t) to be identified, and n b It is the order of the polynomial B(z). The first to nth order of polynomial B(z) b The coefficient of the order.

[0070] Determine n a With n b The method is as follows: While ensuring n b Greater than n a Under the given conditions, multiple attempts were made to minimize the fitting error (the cumulative error between the identified model and the simulation model under the step response).

[0071] S3. Define the parameter vector and information vector of the UWPT system and transform the controlled autoregressive model.

[0072] The parameter vector of the UWPT system is defined as follows:

[0073]

[0074] n = n a +n b R represents a real number.

[0075] Define the information vector of the UWPT system for:

[0076]

[0077] Based on the defined parameter vector and information vector The controlled autoregressive model is represented as:

[0078]

[0079] S4. Sample input and output data, and process parameters. To identify.

[0080] This step samples data of length L and performs gradient iteration based on this data and the controlled autoregressive model to obtain the parameters. The specific steps include:

[0081] S41. Based on data of length L, define the stacked output vector V(L) and the stacked information matrix Φ(L) as follows:

[0082]

[0083] Where v(1) v(2) ... v(L) are v(t) corresponding to data of length L. For data of length L

[0084] S42. Define about the parameter vector Static criterion function:

[0085]

[0086] Where || represents the norm;

[0087] S43. Let k = 1, 2, 3, ... be the iteration variables, and define the parameter vector. The k-th iteration estimate for:

[0088]

[0089] in, For the k-th iteration For the k-th iteration

[0090] S44, to minimize The negative gradient is used for iterative search. When the iteration termination condition is met, the iteration ends and the parameters at this point are output.

[0091] In step S44, an iterative search is performed using the negative gradient, according to the formula:

[0092]

[0093] Where μ>0 is the step size factor, grad[] represents the gradient, and k-1 represents the (k-1)th iteration.

[0094] To ensure the convergence of equation (8), the step size factor μ must satisfy:

[0095]

[0096] in, This is the upper bound of the value of μ.

[0097] In step S44, the iteration termination condition is:

[0098]

[0099] Where ε represents the accuracy of parameter estimation.

[0100] Based on the above analysis, and referring to... Figure 3 Step S44 specifically includes the following steps:

[0101] S441. Initialization: Set all variables to 0, let the number of iterations k = 1, and give the data length L and the parameter estimation accuracy ε.

[0102] S442. Let k0 = 1:L, collect observation data v(k0Δt) and Where Δt is the sampling interval;

[0103] S443, based on the sampled data v(k0Δt) and Construct the stacked output vector V(L) and the stacked information matrix Φ(L), and determine the iteration step size μ;

[0104] S444, Use Refresh the parameter estimation vector;

[0105] S445, Determine if the condition is met. If so, output the parameters at this time. Otherwise, return to step S444 to proceed to the next iteration.

[0106] Based on the above parameter identification process, the CAR parameter model of the system can be obtained, and its discrete transfer function form is as follows:

[0107]

[0108] Applying a bilinear transformation to equation (4.23), the continuous transfer function model of the system is obtained as follows:

[0109]

[0110] Where s is the Laplace operator.

[0111] Based on the preceding analysis, a system simulation model was built in Simulink. The same input was given to both the actual system and the system derived through data-driven modeling. Figure 4 Compare the output waveforms of the two, such as Figure 5As shown, the blue waveform is the output waveform of our model, and the orange waveform is the output waveform of data-driven modeling. From Figure 5 It can be seen that the model established by the theory proposed in this invention has a high degree of fit with the output response of the actual system, proving the effectiveness of this method.

[0112] In summary, the gradient-iteration-based data-driven modeling method for complex medium UWPT systems provided by this invention first determines the circuit model of the underwater wireless power transfer system (UWPT system), then establishes a controlled autoregressive model of the circuit model, then defines the parameter vector and information vector of the UWPT system, and finally performs gradient iteration based on data of length L and the established controlled autoregressive model. After the iteration is completed, the current system parameters are output. This invention employs data-driven modeling, using an extremely simple black-box model. It directly establishes a controlled autoregressive model of the system based on the system's input and output data using a gradient iteration method. This greatly simplifies the system modeling steps in actual industrial environments, significantly reduces the workload of system modeling, and improves the modeling accuracy and reliability of the system in complex media environments.

[0113] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A data-driven modeling method for complex medium UWPT systems based on gradient iteration, characterized in that, Including the following steps: S1. Determine the circuit model of the underwater wireless power transfer system, i.e., the UWPT system; S2. The controlled autoregressive model of the circuit model is established as follows: ， Where {v(t)} and {θ(t)} are the input phase shift angle and output voltage sequence of the UWPT system, respectively, and {n(t)} is a zero-mean, variance σ. 2 Unmeasurable white noise sequence, z -1 Let A(z) be the unit shift operator, and let A(z) be the polynomial coefficients of v(t) to be identified. a It is the order of the polynomial A(z). The first to nth order of polynomial A(z) a The coefficients of order n, B(z) are the polynomial coefficients of θ(t) to be identified, and n b It is the order of the polynomial B(z). The first to nth order of polynomial B(z) b The coefficient of the order; S3. Define the parameter vector of the UWPT system. , n=n a +n b R represents a real number; the information vector of the UWPT system is also defined. The controlled autoregressive model is expressed as: ， S4. Sample data of length L, and perform gradient iteration based on this data and the controlled autoregressive model to obtain parameters. .

2. The data-driven modeling method for complex medium UWPT systems based on gradient iteration according to claim 1, characterized in that, Step S4 specifically includes the following steps: S41. Based on data of length L, define the stacked output vector V(L) and the stacked information matrix. as follows: ， in, Let v(t) be the data of length L. For data of length L ; S42. Define about the parameter vector Static criterion function: ， in, Represents the norm; S43. Let k = 1, 2, 3, ... be the iteration variables, and define the parameter vector. The k-th iteration estimate for: ， in, For the k-th iteration , For the k-th iteration ; S44, to minimize The algorithm uses negative gradients for iterative search. When the iteration termination condition is met, the iteration ends and the parameters at that point are output. .

3. The data-driven modeling method for complex medium UWPT systems based on gradient iteration according to claim 2, characterized in that, In step S44, an iterative search is performed using negative gradients, according to the formula: ， Where μ>0 is the step size factor, Let represent the gradient, and k-1 represent the (k-1)th iteration.

4. The data-driven modeling method for complex medium UWPT systems based on gradient iteration according to claim 3, characterized in that, The step size factor μ must satisfy: ， in, This is the upper bound of the value of μ.

5. The data-driven modeling method for complex medium UWPT systems based on gradient iteration according to claim 4, characterized in that, In step S44, the iteration termination condition is: ， in, This indicates the accuracy of the parameter estimation.

6. The data-driven modeling method for complex medium UWPT systems based on gradient iteration according to claim 5, characterized in that, Step S44 specifically includes the following steps: S441. Initialization: Set all variables to 0 initially, let the iteration number k=1, and given the data length L and parameter estimation accuracy. ; S442. Let k0 = 1:L, collect observation data v(k0Δt) and , where Δt is the sampling interval; S443, based on the sampled data v(k0Δt) and Construct the stacked output vector V(L) and the stacked information matrix. And determine the iteration step size μ; S444, Use Refresh the parameter estimation vector; S445, Determine if the condition is met. If so, output the parameters at this time. Otherwise, return to step S444 to proceed to the next iteration.

7. The data-driven modeling method for complex medium UWPT systems based on gradient iteration according to any one of claims 1 to 6, characterized in that: The circuit model of the UWPT system includes a primary-side energy transmitter and a secondary-side energy receiver. The primary-side energy transmitter includes a DC power supply, an inverter, and a primary resonant structure connected in sequence. The secondary-side energy receiver includes a secondary resonant structure, a rectifier bridge, and a filter capacitor C connected in sequence. f and load R L .

8. The data-driven modeling method for complex medium UWPT systems based on gradient iteration according to claim 7, characterized in that: The inverter includes a first full-bridge inverter and a second full-bridge inverter connected in parallel with the DC power supply.

9. The data-driven modeling method for complex medium UWPT systems based on gradient iteration according to claim 8, characterized in that: The primary resonant structure includes a first primary resonant circuit connected to the first full-bridge inverter and a second primary resonant circuit connected to the second full-bridge inverter.

10. The data-driven modeling method for complex medium UWPT systems based on gradient iteration according to claim 9, characterized in that: The first primary resonant circuit and the second primary resonant circuit have the same structure and both adopt an LCC type compensation topology; the secondary resonant structure adopts an S type compensation topology.

Images