A Deep Learning-Based Offset Detection Method and System for Wireless Power Transfer Systems

Abstract

This invention discloses a method and system for offset detection in a wireless power transfer system based on deep learning, comprising the following steps: Step 1: When the output current of the wireless power transfer system is abnormal, the output current on the load is collected; Step 2: The spatiotemporal features of the output current spectrum are extracted and normalized to obtain time-frequency information; Step 3: The time-frequency information obtained in Step 2 is input into the offset detection model to obtain the current offset detection result of the wireless power transfer system; the offset detection result is: the abnormal output current is caused by the offset between the receiving coil and the transmitting coil, or the abnormal output current is not caused by the offset between the receiving coil and the transmitting coil; Step 4: The input voltage is adjusted according to the current offset detection result of the wireless power transfer system.

Description

Technical Field

[0001] This invention belongs to the field of power electronics technology, specifically a method and system for offset detection in a wireless power transmission system based on deep learning. Background Technology

[0002] The ocean contains abundant natural resources, and the development of these resources is receiving increasing attention. Autonomous underwater vehicles (AUVs) play a crucial role in this development. To eliminate the salvage process before recharging, improve safety, and extend the cruising range of AUVs, wireless power transfer technology for AUVs has been continuously developed, enabling power transfer without direct contact.

[0003] However, the underwater environment is extremely complex. When the magnetic core material is in the deep sea, the pressure of the seawater significantly reduces its permeability, thus decreasing the self-inductance and mutual inductance of the coupling mechanism, as well as the coupling coefficient and transmission efficiency. Furthermore, ocean currents are a significant factor affecting power transmission efficiency. Due to these disturbances, the magnetic coupling mechanism experiences frequent mechanical swaying, leading to core misalignment or displacement. This directly impacts leakage flux and the stability of the magnetic circuit, ultimately affecting the overall system performance. Although methods exist to improve the system's resistance to misalignment, its occurrence remains unavoidable.

[0004] Therefore, in order to ensure the efficient transmission of wireless power transmission systems, when transmission efficiency is compromised, corresponding measures are taken to determine whether the system is affected by offset, and the primary voltage regulation is adjusted to maintain efficient power transmission. Summary of the Invention

[0005] Purpose of the invention: To address the problem of not being able to accurately determine whether the power transmission efficiency is impaired due to offset, and to solve the problem of how to maintain efficient power transmission when offset occurs, this invention proposes an offset detection method and system for wireless power transmission systems based on deep learning.

[0006] Technical Solution: A deep learning-based offset detection method for wireless power transfer systems, applied to wireless power transfer systems including a transmitting coil, a receiving coil, and a load; the transmitting coil transmits energy to the receiving coil according to the input voltage, and the receiving coil receives the energy from the transmitting coil, generating an output current on the load; including the following steps:

[0007] Step 1: When the output current of the wireless power transmission system is abnormal, collect the output current on the load;

[0008] Step 2: Extract and normalize the spatiotemporal features of the output current spectrum to obtain time-frequency information;

[0009] Step 3: Input the time-frequency information obtained in Step 2 into the offset detection model to obtain the current offset detection result of the wireless power transmission system; the offset detection result is: the abnormal output current is caused by the offset between the receiving coil and the transmitting coil, or the abnormal output current is not caused by the offset between the receiving coil and the transmitting coil;

[0010] Step 4: Adjust the input voltage based on the current offset detection results of the wireless power transmission system.

[0011] Furthermore, the offset detection model is a trained LSTM neural network, which consists of LSTM layers, fully connected layers, softmax layers, and classification layers.

[0012] Furthermore, the LSTM neural network is trained using the following steps to obtain the offset detection model:

[0013] The output current of the wireless power transfer system is collected when the receiving coil and the transmitting coil are offset at different times, and the output current of the wireless power transfer system is collected under different load resistances; the collected output signals are divided to obtain training set and test set;

[0014] The spatiotemporal features of the output signals in the training and test sets are extracted and normalized sequentially to obtain time-frequency information;

[0015] Tag each time-frequency information, including: the time-frequency information of the output current of the wireless power transfer system when the receiving coil and the transmitting coil are offset by different values ​​is tagged as true, and the time-frequency information of the output current of the wireless power transfer system under different load resistances is tagged as false;

[0016] The LSTM neural network was trained using the training set to obtain the offset detection model.

[0017] Furthermore, the misalignment between the receiving coil and the transmitting coil includes lateral misalignment and angular misalignment.

[0018] Furthermore, spatiotemporal features are extracted from the spectrum of the current signal, including short-time Fourier transform and spectral entropy analysis of the current signal spectrum.

[0019] This invention discloses a deep learning-based offset detection system for a wireless power transfer system, applied to a wireless power transfer system. The wireless power transfer system includes a transmitting coil, a receiving coil, and a load. The transmitting coil transmits energy to the receiving coil according to the input voltage, and the receiving coil receives the energy from the transmitting coil, generating an output current on the load. The system includes:

[0020] The current signal acquisition module is used to acquire the output current on the load when the output current of the wireless power transmission system is abnormal.

[0021] The spatiotemporal feature extraction module is used to extract and normalize the spatiotemporal features of the output current spectrum to obtain time-frequency information.

[0022] An offset detection model is used to obtain the current offset detection result of the wireless power transmission system based on the input time-frequency information. The offset detection result is: the abnormal output current is caused by the offset between the receiving coil and the transmitting coil, or the abnormal output current is not caused by the offset between the receiving coil and the transmitting coil. The offset between the receiving coil and the transmitting coil includes lateral offset and angular offset.

[0023] The voltage adjustment module is used to adjust the input voltage based on the current offset detection results of the wireless power transmission system.

[0024] Furthermore, the offset detection model is an LSTM neural network trained according to the following steps:

[0025] The output current of the wireless power transfer system is collected when the receiving coil and the transmitting coil are offset at different times, and the output current of the wireless power transfer system is collected under different load resistances; the collected output signals are divided to obtain training set and test set;

[0026] The spatiotemporal features of the output signals in the training and test sets are extracted and normalized sequentially to obtain time-frequency information;

[0027] Tag each time-frequency information, including: the time-frequency information of the output current of the wireless power transfer system when the receiving coil and the transmitting coil are offset by different values ​​is tagged as true, and the time-frequency information of the output current of the wireless power transfer system under different load resistances is tagged as false;

[0028] The LSTM neural network is trained using the training set to obtain the offset detection model;

[0029] The LSTM neural network consists of LSTM layers, fully connected layers, softmax layers, and classification layers.

[0030] Furthermore, spatiotemporal features are extracted from the spectrum of the current signal, including short-time Fourier transform and spectral entropy analysis of the current signal spectrum.

[0031] The present invention discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the aforementioned deep learning-based wireless power transmission system offset detection method.

[0032] The present invention discloses a storage medium storing a program for an offset detection method, wherein when the program for the offset detection method is executed by at least one processor, it implements the steps of the aforementioned offset detection method for a deep learning-based wireless power transmission system.

[0033] Beneficial effects: Compared with the prior art, the method / device of this invention can quickly determine whether the change in current transmission efficiency is caused by the displacement of the charging plates, and quickly take action to maintain the efficient transmission of electrical energy. Under the influence of multiple factors underwater, it can also ensure the reliability of the wireless power transmission system. Attached Figure Description

[0034] Figure 1 This is a flowchart of a deep learning-based offset detection method for wireless power transfer systems.

[0035] Figure 2 Circuit topology diagram for a wireless charging power transmission system;

[0036] Figure 3 This is a spatial model of two coils with lateral offset.

[0037] Figure 4 This is a spatial model of the angular offset between two coils.

[0038] Figure 5 The equivalent model of leakage inductance for LCC-LCC compensation topology;

[0039] Figure 6 The curves show the effects of offset angle and transmission distance on transmission efficiency. Detailed Implementation

[0040] The technical solution of the present invention will now be further described in conjunction with the accompanying drawings and embodiments.

[0041] like Figure 2As shown, the wireless power transfer system of the present invention includes: an inverter, a transmitting LCC circuit, a transmitting coil L1, a receiving LCC circuit, a receiving coil L2, and a rectifier. The transmitting coil and the transmitting LCC circuit constitute the transmitting end of the wireless power transfer system. The transmitting coil adopts a solenoid-like layout to generate a relatively uniform magnetic field. The transmitting LCC circuit consists of an inductor L... f1 Capacitor C f1 It consists of capacitor C1; the receiving end LCC circuit consists of inductor L f2 Capacitor C f2 It consists of capacitor C2; R L Indicates load, U R and I R These represent the output voltage and output current of the IPT converter, respectively. The inverter is used to drive the input DC voltage U. d It consists of four field-effect transistors S1, S2, S3, and S4, and the rectifier consists of four Schottky diodes (D1, D2, D3, and D4). The receiving coil is a solenoid-like coil used to couple the transmitting coil to receive energy.

[0042] Example 1:

[0043] like Figure 1 As shown, this embodiment discloses a method for offset detection in a wireless power transfer system based on deep learning, which mainly includes the following steps:

[0044] Step 1: Acquire the output current of the wireless power transfer system under different offsets between the receiving and transmitting coils using an oscilloscope. Also acquire the output current under different load resistances using an oscilloscope. The purpose of acquiring the output current under different load resistances is to simulate an underwater environment, where other factors can also cause changes in load resistance. Specifically: First, build the circuit topology of the wireless power transfer system in the simulation software and set the parameters of the corresponding components. Use an oscilloscope to acquire the output current at the load end under normal conditions. Manually adjust the coupling coefficient in the topology to simulate a certain lateral / angular offset between the receiving and transmitting coils, and record the output current at the load end at this time. By continuously adjusting the coupling coefficient and changing the mutual inductance M, the output current is acquired. Because other factors can also affect the system underwater, the load resistance in the topology model is continuously changed, and the acquired output current signal is used as false information.

[0045] Step 2: Process the output current collected in Step 1 to obtain the training set and the test set; specific operations include:

[0046] The output current of the wireless power transfer system is collected when the receiving coil and the transmitting coil are offset at different times, and the output current of the wireless power transfer system is collected under different load resistances.

[0047] The output current of the wireless power transfer system when the receiving coil and the transmitting coil are offset at different times and the output current of the wireless power transfer system under different load resistances are segmented to make the number of samples for each current signal the same.

[0048] The output current after data segmentation is divided into training and test sets. In this embodiment, the training and test sets are divided in a 9:1 ratio. The division of the training and test sets needs to balance the distribution of different samples. In this embodiment, the repmat function is used to balance the number of samples of the two types of signals.

[0049] The spatiotemporal features of the output current in the training set and the test set are extracted and normalized sequentially to obtain time-frequency information;

[0050] Tag each time-frequency information, including: the time-frequency information of the output current of the wireless power transfer system when the receiving coil and the transmitting coil are offset by different values ​​is tagged as true, and the time-frequency information of the output current of the wireless power transfer system under different load resistances is tagged as false;

[0051] The LSTM neural network was trained using the training set to obtain the offset detection model.

[0052] During signal acquisition, if a sudden disturbance occurs, it cannot be well captured by simply using Fourier transform in the frequency domain. It is merely considered as a superposition of a series of low-frequency components, failing to accurately reflect the changes brought about by the disturbance. Therefore, Fourier transform is used to decompose the entire time-domain process into numerous small processes of equal length, each of which is approximately stationary. Then, Fourier transform is performed, allowing the acquisition of non-stationary signals.

[0053] Step 3: Construct a deep learning classifier, train it using the training set, and test it using the test set to obtain a classification model for detecting biases. In this embodiment, the deep learning classifier is an LSTM neural network consisting of LSTM layers, fully connected layers, softmax layers, and classification layers. The size of the LSTM neural network structure is determined by defining Epochs and Batchsize, and the initial weights and biases are set. A loss function is defined, and stochastic gradient descent is used to solve for the LSTM neural network parameters, completing the training of the deep learning classifier.

[0054] Step 4: Obtain the output current of the wireless power transmission system under actual working conditions using an oscilloscope, and input the time-frequency information of the output current into the classification model obtained in Step 3 to obtain the offset detection result; the offset detection result is: the abnormal output current is caused by the offset between the receiving coil and the transmitting coil, or the abnormal output current is not caused by the offset between the receiving coil and the transmitting coil.

[0055] Step 5: Based on the offset detection results, take corresponding measures to ensure the transmission efficiency of the wireless power transmission system. Specifically, when the offset detection result indicates that the abnormal output current is caused by the offset between the receiving coil and the transmitting coil, i.e., by a change in the mutual inductance M, adjust the input voltage to bring the input current back to the expected value, maintain normal charging, and ensure transmission efficiency. Figure 3 As shown, when there is a lateral positional offset in the coil, the mutual inductance formula is:

[0056]

[0057] In the formula, r1 and r2 are the radii of the transmitting and receiving coils, respectively; t is the lateral offset of the transmitting and receiving coils; d is the distance between the centers; n1 is the number of turns of the transmitting coil; n2 is the number of turns of the receiving coil; μ0 represents the permeability; θ, These represent the offset angles of the receiving coil and the transmitting coil relative to the x-axis, respectively.

[0058] like Figure 4 As shown, when the coil has a certain offset angle, the mutual inductance formula is:

[0059]

[0060] In the formula, the center coordinates of the transmitting coil are represented as (0, t, d), and α is the offset angle of the coil relative to the z-axis.

[0061] Figure 5 The equivalent model of leakage inductance in the LCC-LCC compensation topology is shown. According to Kirchhoff's laws, the KVL matrix equation can be obtained, and the input voltage U can be derived. in With output current I o The ratio relationship is as follows:

[0062]

[0063]

[0064] Therefore, when there is a misalignment between the coils, M changes. Figure 6 The curves showing the effects of offset angle and transmission distance on transmission efficiency are illustrated. When offset exists, transmission efficiency decreases and is impaired. To achieve continuous charging, the input voltage U is changed.in To make the output current I o Maintain normal values.

[0065] Example 2:

[0066] This embodiment discloses a deep learning-based offset detection system for a wireless power transfer system. The system includes a transmitting coil, a receiving coil, and a load. The transmitting coil transmits energy to the receiving coil based on the input voltage. The receiving coil receives the energy from the transmitting coil and generates an output current across the load. The system includes:

[0067] The current signal acquisition module is used to acquire the output current on the load when the output current of the wireless power transmission system is abnormal.

[0068] The spatiotemporal feature extraction module is used to extract and normalize the spatiotemporal features of the output current spectrum to obtain time-frequency information.

[0069] An offset detection model is used to obtain the current offset detection result of the wireless power transmission system based on the input time-frequency information. The offset detection result is: the abnormal output current is caused by the offset between the receiving coil and the transmitting coil, or the abnormal output current is not caused by the offset between the receiving coil and the transmitting coil. The offset between the receiving coil and the transmitting coil includes lateral offset and angular offset. The offset detection model in this embodiment is a trained LSTM neural network, which consists of LSTM layers, fully connected layers, softmax layers, and classification layers.

[0070] The voltage adjustment module is used to adjust the input voltage based on the current offset detection results of the wireless power transmission system.

[0071] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0072] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0073] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for detecting offset of a wireless power transmission system based on deep learning, applied to a wireless power transmission system, the wireless power transmission system comprising a transmitting coil, a receiving coil and a load; the transmitting coil transmits energy to the receiving coil according to an input voltage, the receiving coil is used to receive energy from the transmitting coil, and an output current is generated on the load; characterized in that: The method comprises the following steps: Step 1: collecting the output current on the load when the output current of the wireless power transmission system is abnormal; Step 2: performing time-space feature extraction and normalization processing on the frequency spectrum of the output current to obtain time-frequency information; Step 3: inputting the time-frequency information obtained in step 2 into an offset detection model to obtain the current offset detection result of the wireless power transmission system; the offset detection result is that the abnormal output current is caused by the offset between the receiving coil and the transmitting coil or the abnormal output current is not caused by the offset between the receiving coil and the transmitting coil; Step 4: adjusting the input voltage according to the current offset detection result of the wireless power transmission system; The offset detection model is a trained LSTM neural network, which is composed of an LSTM layer, a fully connected layer, a softmax layer and a classification layer; The LSTM neural network is trained by the following steps to obtain the offset detection model: Collect the output current of the wireless power transmission system when the receiving coil and the transmitting coil have different offsets, and collect the output current of the wireless power transmission system under different load resistances; segment the collected output current data; Divide the segmented output current to obtain a training set and a test set; Perform time-space feature extraction and normalization processing on the frequency spectrum of the output current in the training set and the test set in sequence to obtain time-frequency information; Label each time-frequency information, including: the label of the time-frequency information of the output current of the wireless power transmission system when the receiving coil and the transmitting coil have different offsets is true, and the label of the time-frequency information of the output current of the wireless power transmission system under different load resistances is false; Train the LSTM neural network using the training set to obtain the offset detection model.

2. The method of claim 1, wherein the method is based on deep learning. The offset between the receiving coil and the transmitting coil includes horizontal offset and angle offset.

3. The method of claim 1, wherein the method further comprises: The time-space feature extraction on the frequency spectrum of the current signal includes short-time Fourier transform and spectral entropy analysis.

4. A deep learning-based wireless power transmission system offset detection system, applied to a wireless power transmission system, the wireless power transmission system comprising a transmitting coil, a receiving coil and a load; the transmitting coil transmits energy to the receiving coil according to an input voltage, the receiving coil is used to receive energy from the transmitting coil, and an output current is generated on the load; characterized in that: It comprises: A current signal collection module for collecting the output current on the load when the output current of the wireless power transmission system is abnormal; A time-space feature extraction module for performing time-space feature extraction and normalization processing on the frequency spectrum of the output current to obtain time-frequency information; An offset detection model for obtaining the current offset detection result of the wireless power transmission system based on the input time-frequency information; the offset detection result is that the abnormal output current is caused by the offset between the receiving coil and the transmitting coil or the abnormal output current is not caused by the offset between the receiving coil and the transmitting coil; the offset between the receiving coil and the transmitting coil includes horizontal offset and angle offset; A voltage adjustment module for adjusting the input voltage according to the current offset detection result of the wireless power transmission system; The offset detection model is an LSTM neural network trained according to the following steps: Collect the output current of the wireless power transmission system when the receiving coil and the transmitting coil have different offsets, and the output current of the wireless power transmission system under different load resistances; Divide the output current after data segmentation to obtain a training set and a test set; Extract the time-space features of the frequency spectrum of the output current in the training set and the test set in turn and perform normalization processing to obtain time-frequency information; Label each time-frequency information, including: the label of the time-frequency information of the output current of the wireless power transmission system when the receiving coil and the transmitting coil have different offsets is true, and the label of the time-frequency information of the output current of the wireless power transmission system under different load resistances is false; Train the LSTM neural network using the training set to obtain an offset detection model; The LSTM neural network is composed of an LSTM layer, a full connection layer, a softmax layer and a classification layer.

5. The deep learning based wireless power transfer system offset detection system of claim 4, wherein: The time-space features of the frequency spectrum of the current signal are extracted, including: short-time Fourier transform and spectral entropy analysis are performed on the frequency spectrum of the current signal. 6.A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the wireless power transmission system offset detection method based on deep learning according to any one of claims 1 to 3 when executing the computer program. 7.A storage medium, wherein the storage medium stores a program of an offset detection method, and the program of the offset detection method implements the steps of the wireless power transmission system offset detection method based on deep learning according to any one of claims 1 to 3 when executed by at least one processor.

Images