Capacitor health parameter estimation model construction method, state monitoring method and device

Abstract

The application provides a capacitor health parameter estimation model construction method, a state monitoring method and device. The capacitor health parameter estimation model construction method comprises the following steps: constructing a pre-training set based on time domain characteristic information and frequency domain characteristic information; constructing a label training set based on an actual capacitance value and an actual ESR of a capacitor; constructing a stacked auto-encoder (SAE) model comprising multiple auto-encoders (AEs), greedily pre-training each AE layer by layer, and initializing the weight and bias vector of the SAE model with the parameters after training; calculating a reconstruction error and a label output error, combining the reconstruction error and the label output error to form an objective function; performing a backward fine-tuning on the SAE model to update the weight and bias vector in order to minimize the objective function, and outputting a stacked progressive auto-encoder model with optimal parameters as a capacitor health parameter estimation model. The application can improve the estimation accuracy of the capacitor health parameter.

Description

Technical Field

[0001] This application relates to the field of power supply technology, and in particular to a method for constructing a capacitor health parameter estimation model, a condition monitoring method, and a device. Background Technology

[0002] Battery State of Health (SOH) represents the percentage of a battery's full-charge capacity relative to its rated capacity. It describes the degree of capacity degradation and is a key indicator for determining whether a power battery meets obsolescence standards; therefore, the accuracy of SOH calculation is crucial. Many factors influence battery SOH, and many parameters are difficult to measure in real time, making SOH estimation complex and challenging. Current battery SOH estimation methods primarily rely on estimation models to estimate capacitance and equivalent series resistance (ESR), and then assess the battery's health based on the relationship between capacitance, ESR, and SOH.

[0003] In implementing the embodiments of this application, it was found that the prior art has at least the following problems:

[0004] Existing estimation models have low accuracy in estimating capacitor health parameters related to battery health status, which affects the battery health status assessment results. Summary of the Invention

[0005] This application provides a method for constructing a capacitor health parameter estimation model, a state monitoring method, and an apparatus to solve the problem of low accuracy in estimating capacitor health parameters related to battery health status in existing estimation models.

[0006] In a first aspect, embodiments of this application provide a method for constructing a capacitance health parameter estimation model, including:

[0007] Time-domain and frequency-domain feature information are determined based on historical capacitor current and voltage signals, and a pre-training set is constructed based on the time-domain and frequency-domain feature information.

[0008] Obtain the actual capacitance value and actual ESR of the capacitor, and construct a label training set based on the actual capacitance value and actual ESR of the capacitor;

[0009] Construct a Stacked Autoencoder (SAE) model containing multiple autoencoders (AEs), pre-train each AE layer by layer greedily, and initialize the weights and bias vectors of the SAE model with the trained parameters; each AE includes a decoder and an encoder.

[0010] The encoder generates latent vectors using the pre-training set as input, and the decoder generates reconstructed data that approximates the pre-training set using the latent vectors as input.

[0011] Calculate the reconstruction error and the label output error, and combine the reconstruction error and the label output error to form an objective function; wherein, the reconstruction error is determined based on the reconstructed data and the pre-training set; and the label output error is determined based on the latent vector and the label training set.

[0012] The SAE model is back-tuned by minimizing the objective function to update the weights and the bias vector, and the stacked progressive autoencoder model with optimal output parameters is used as the capacitance health parameter estimation model.

[0013] In one possible implementation, determining the time-domain feature information and frequency-domain feature information based on the historical capacitor current signal and the historical capacitor voltage signal includes:

[0014] Acquire historical capacitor current and voltage signals;

[0015] Time-domain feature extraction is performed on the historical capacitor current signal and the historical capacitor voltage signal respectively to obtain time-domain feature information, and the time-domain feature information is then subjected to fast Fourier transform to obtain frequency-domain feature information.

[0016] In one possible implementation, the time-domain feature information includes one or more of the following: the effective value of the switching ripple current, the effective value of the switching ripple voltage, the effective value of the capacitor DC current, the effective value of the capacitor DC voltage, the switching frequency, the rated capacitance value, and the rated ESR of the capacitor.

[0017] The frequency domain feature information includes one or more of the following: power frequency voltage amplitude, power frequency voltage phase, power frequency current amplitude, power frequency current phase, switching frequency ripple voltage amplitude, switching frequency ripple voltage phase, switching frequency ripple current amplitude, and switching frequency ripple current phase.

[0018] In one possible implementation, the objective function is as follows:

[0019]

[0020] in, The reconstruction error is referred to as "the reconstruction error". The label output error; x i The values ​​of neurons in the pre-training set; Indicates reconstructed data values; y i The label values ​​in the labeled training set; is the value of the latent vector; μ is the scaling factor, with a value range of (0, 1); m is the number of label types in the label training set, i = 1, 2, ..., m.

[0021] In one possible implementation, the step of back-tuning the SAE model to update the weights and the bias vector by minimizing the objective function includes:

[0022] The weights and the bias vector are updated using the backpropagation algorithm and gradient descent method.

[0023] Secondly, embodiments of this application provide a capacitor health status monitoring method based on the capacitor health parameter estimation model described in the first aspect or any possible implementation thereof, including:

[0024] Acquire real-time capacitor current and capacitor voltage signals;

[0025] Real-time time-domain feature information and real-time frequency-domain feature information are extracted based on the capacitor current signal and the capacitor voltage signal;

[0026] The real-time time-domain feature information and the real-time frequency-domain feature information are input into the trained capacitor health parameter estimation model, and the capacitor value and ESR are output.

[0027] The capacitor health status is calculated based on the capacitance value and the ESR, so as to monitor the health status of the capacitor.

[0028] In one possible implementation, calculating the capacitor health based on the capacitance value and the ESR includes:

[0029] The capacitor health status in both capacitive and voltage-based forms is calculated using the following formulas:

[0030]

[0031]

[0032] Among them, SOH C For capacitor health; SOH R For ESR capacitor health; C e|c C represents the equivalent capacitance value under the current condition. rated R is the nominal rated capacitance. EOL R0 is the ESR in the End-of-Life (EOL) state; R0 is the initial ESR; Rc is the ESR in the current state.

[0033] The average of the capacitance health in the form of capacitance and the capacitance health in the form of voltage is calculated as the target capacitance health.

[0034] In one possible implementation, the extraction of real-time time-domain feature information and real-time frequency-domain feature information based on the capacitor current signal and the capacitor voltage signal includes:

[0035] Time-domain feature extraction is performed on the capacitor current signal and the capacitor voltage signal respectively to obtain real-time time-domain feature information, and fast Fourier transform is performed on the real-time time-domain feature information to obtain real-time frequency-domain feature information.

[0036] Thirdly, embodiments of this application provide an apparatus for constructing a capacitance health parameter estimation model, comprising:

[0037] The training set construction module is used to determine time-domain and frequency-domain feature information based on historical capacitor current and voltage signals, and to construct a pre-training set based on the time-domain and frequency-domain feature information; to obtain the actual capacitance value and the actual equivalent series resistance (ESR) of the capacitor, and to construct a label training set based on the actual capacitance value and the actual ESR of the capacitor.

[0038] The model building module is used to construct a stacked progressive autoencoder (SAE) model containing multiple autoencoders (AEs). Each AE is pre-trained layer-by-layer greedily, and the weights and bias vectors of the SAE model are initialized with the trained parameters. Each AE includes a decoder and an encoder. The encoder generates latent vectors using the pre-training set as input, and the decoder generates reconstructed data approximating the pre-training set using the latent vectors as input. The reconstruction error and label output error are calculated, and the reconstruction error and label output error are combined to form an objective function. The reconstruction error is determined based on the reconstructed data and the pre-training set; the label output error is determined based on the latent vectors and the label training set. The SAE model is then fine-tuned by minimizing the objective function to update the weights and bias vectors, and the stacked progressive autoencoder model with optimal output parameters is used as a capacitance health parameter estimation model.

[0039] Fourthly, embodiments of this application provide a capacitor health status monitoring device, comprising:

[0040] The acquisition module is used to acquire the real-time capacitor current signal and capacitor voltage signal;

[0041] The extraction module is used to extract real-time time-domain feature information and real-time frequency-domain feature information based on the capacitor current signal and the capacitor voltage signal;

[0042] The calculation module is used to input the real-time time-domain feature information and the real-time frequency-domain feature information into the trained capacitor health parameter estimation model, output the capacitance value and ESR; and calculate the capacitor health degree based on the capacitance value and the ESR, so as to realize the health status monitoring of the capacitor based on the capacitor health degree.

[0043] Fifthly, embodiments of this application provide an electronic 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 method described in the first aspect, the second aspect, any possible implementation of the first aspect, or any possible implementation of the second aspect.

[0044] Sixthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described in the first aspect, the second aspect, any possible implementation of the first aspect, or any possible implementation of the second aspect.

[0045] This application provides a method for constructing a capacitor health parameter estimation model, a state monitoring method, and an apparatus. A pre-training set is constructed by extracting time-domain and frequency-domain feature information from historical capacitor current and voltage signals, and a labeled training set is constructed using the actual capacitance value and actual ESR of the capacitor. Introducing the actual capacitance value and actual ESR as label information during model training provides an efficient and clear framework for pre-training, establishing a correspondence between the model output and the capacitance C and ESR. The pre-training set and the label training set are input into an SAE model composed of multiple autoencoders (AEs). Each AE is greedily pre-trained layer by layer, and the weights and bias vectors of the SAE model are initialized with the trained parameters. The reconstruction error and the label output error are calculated. The reconstruction error and the label output error are combined to form an objective function. The SAE model is then fine-tuned by minimizing the objective function to update the weights and bias vectors. The stacked progressive autoencoder model with the optimal output parameters is used as the capacitance health parameter estimation model. The objective function, which combines the reconstruction error and the label output error, can achieve lossless information and stable state. At the same time, it can clarify the characteristic significance of capacitance parameter information in the layer-by-layer extraction process, improve the accuracy of capacitance health parameter estimation, and thus improve the monitoring efficiency of capacitance health status. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart illustrating the implementation of a capacitance health parameter estimation model construction method provided in an embodiment of this application.

[0048] Figure 2 This is a schematic diagram of an improved AE structure provided in an embodiment of this application;

[0049] Figure 3 This is a schematic diagram illustrating the working principle of the improved AE structure provided in one embodiment of this application;

[0050] Figure 4 This is a schematic diagram of an improved SAE structure provided in an embodiment of this application;

[0051] Figure 5 This is a flowchart illustrating the implementation of a capacitor health status monitoring method according to another embodiment of this application;

[0052] Figure 6 This is a schematic diagram of the structure of a device for constructing a capacitance health parameter estimation model according to an embodiment of this application;

[0053] Figure 7 This is a schematic diagram of the structure of a capacitor health status monitoring device provided in one embodiment of this application;

[0054] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0055] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0056] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.

[0057] Unless otherwise stated, the term "multiple" means two or more. The character " / " indicates that the preceding and following objects are in an "or" relationship. For example, A / B means: A or B. The term "and / or" describes an association between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or, A and B.

[0058] The terms used in this application are for describing embodiments only and are not intended to limit the claims. As used in the description of embodiments and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. Similarly, the term “and / or” as used herein means including one or more of the associated listed elements and all possible combinations thereof. Additionally, when used in this application, the terms “comprise” and its variations “comprises” and / or “comprising” refer to the presence of stated features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof. Without further limitation, an element defined by the phrase “comprising an…” does not exclude the presence of additional identical elements in the process, method, or apparatus that includes said element.

[0059] In this application, each embodiment focuses on describing the differences from other embodiments, and similar or identical parts between embodiments can be referred to mutually. For methods, products, etc., disclosed in the embodiments, if they correspond to the method section disclosed in the embodiments, then the relevant parts can be referred to the description of the method section.

[0060] To make the objectives, technical solutions, and advantages of this application clearer, the following description will be provided in conjunction with the accompanying drawings and specific embodiments.

[0061] Figure 1 This is a flowchart illustrating the implementation of a capacitance health parameter estimation model construction method provided in an embodiment of this application, as follows: Figure 1 As shown, the method includes the following steps:

[0062] S101, determine time-domain and frequency-domain feature information based on historical capacitor current and voltage signals, and construct a pre-training set based on the time-domain and frequency-domain feature information.

[0063] In this embodiment, capacitor health parameters are parameters used to characterize the health state of the capacitor or the overall health state of the battery. In this embodiment, capacitor health parameters include capacitance value and ESR (Electrostatic Displacement).

[0064] In other possible implementations, capacitor health parameters include not only capacitance value and ESR, but also other parameters such as operating temperature. The first two are essential, while temperature depends on the model or whether the capacitor has a corresponding sensor.

[0065] S102, obtain the actual capacitance value and actual ESR of the capacitor, and construct a label training set based on the actual capacitance value and actual ESR of the capacitor.

[0066] In this process, capacitance parameter monitoring focuses on the capacitor itself, specifically its capacitance value and ESR (Electronic Stability Reduction). Choosing accurate values ​​for these two parameters as labels enables supervised learning in what was originally unsupervised model training. The actual capacitance value and actual ESR need to be calibrated, for example, by measuring the actual capacitance value using a bridge circuit. Once these values ​​are fixed as labels, they remain unchanged during calculations, guiding the direction of model parameter updates during training.

[0067] In other possible implementations, the data in the labeled training set may also include parameters other than capacitance values ​​and ESR. Specifically, the types of parameters in the labeled training set may be adjusted according to the accuracy requirements of the training error or the monitoring requirements of capacitance health parameters.

[0068] S103, construct an SAE model containing multiple AEs, pre-train each AE layer by layer greedily, and initialize the weights and bias vectors of the SAE model with the trained parameters; each AE includes a decoder and an encoder.

[0069] S104, the encoder generates latent vectors using the pre-training set as input, and the decoder generates reconstructed data that approximates the pre-training set using the latent vectors as input.

[0070] S105, calculate the reconstruction error and the label output error, and combine the reconstruction error and the label output error to form the objective function; wherein, the reconstruction error is determined based on the reconstructed data and the pre-training set; the label output error is determined based on the latent vector and the label training set.

[0071] S106 uses the minimization objective function to fine-tune the SAE model in reverse, updating the weights and bias vectors, and outputs the stacked progressive autoencoder model with optimal parameters as the capacitance health parameter estimation model.

[0072] Minimizing the objective function includes: repeatedly iterating the training until the objective function is less than a preset value, or, the number of training iterations reaches a preset number.

[0073] In this embodiment, a capacitance health parameter estimation model is constructed based on AE (Advanced Image Processing). AE is essentially a lossless transformation of input data; however, simply pursuing lossless information and state stability with the objective function of minimizing reconstruction error may lead to fuzzification of capacitance parameter information during the layer-by-layer extraction process, resulting in unclear meaning of the extracted features.

[0074] Therefore, to address this issue, this application embodiment optimizes the AE by introducing label information. AE is an unsupervised neural network model, and feature extraction during the pre-training phase lacks guidance. Label information, however, can clearly show the differences between various training samples. Introducing label information into AE provides an efficient and clear approach for pre-training, establishing a correspondence between the final extracted two-dimensional results and the capacitance value and ESR. The two-dimensional results are the two quantities extracted by AE through matrix operations based on the time-domain and frequency-domain characteristics of the capacitor voltage and current signals.

[0075] In practice, the correspondence between the two-dimensional result and the capacitance and ESR is quantitative. Ideally, the two quantities corresponding to the two-dimensional result are numerically equal to C and ESR, but in reality, there will be an error within 1%.

[0076] Figure 2 This is a schematic diagram of the improved AE structure. A single AE itself has a two-layer architecture, but it's drawn as a three-layer structure to illustrate the reconstruction process. The dashed boxes represent introduced label information. Each neuron carries a bias, and each horizontal line represents the computational weights between neurons. The values ​​for the next layer are calculated using matrix operations as follows:

[0077] Y = XW + B

[0078] Where Y is the numerical vector of the next layer, W is the weight matrix, B is the bias vector, and X is the data vector of the current layer.

[0079] In addition, such as Figure 2 As shown, during model training, the label information of the training set data, namely the capacitance value and ESR, is known. This information is combined with the capacitance time-domain feature data and frequency-domain feature data, and then input into the model. Specifically, the data is processed side-by-side. For example, if the original label of a data point is [3, 4], and the other features are [5, 6, 7, 8, ..., N], resulting in two vectors, a single vector of [3, 4, 5, 6, ..., N] is synthesized as a sample. The capacitance value, ESR, capacitance time-domain feature data, and frequency-domain feature data are then combined into a single vector and input into the AE.

[0080] The original Advanced Effects (AE) is bidirectional, meaning that hidden layer values ​​can be calculated from the visible layer, and visible layer values ​​can be calculated from the hidden layer. The connection between the hidden layer and the label neurons is unidirectional, meaning that only the label values ​​can be calculated from the hidden layer. Therefore, the structure of the AE during training can be further subdivided as follows: Figure 3 As shown.

[0081] like Figure 3 The improved working principle of AE is as follows:

[0082] The training process is broken down into two steps: encoding and decoding.

[0083] During encoding, the input capacitance time-domain features and capacitance frequency-domain features are converted into data (e.g., a 5-dimensional input [1, 2, 3, 4, 5] is converted to 4 dimensions, or the dimensions are increased or kept unchanged) to obtain the intermediate hidden layer values, i.e., the latent vectors. In this embodiment, the labels are 2-dimensional arrays; therefore, the final output of the intermediate hidden layer must be 2-dimensional data. Therefore, a one-step data conversion (e.g., 5-dimensional → 2-dimensional) or a multi-layer data conversion (e.g., 5-dimensional → 4-dimensional → 2-dimensional, etc.) can be chosen to construct different AE individuals.

[0084] The decoding process is the reverse of the encoding process, gradually mapping from 2D back to 5D. The reconstruction error of the 5D dimension itself, as well as the error between the 2D label data and the output, are calculated to form the objective function. Then, the weights and biases are updated.

[0085] During training, one encoding and decoding cycle constitutes one training round. The training is repeated iteratively until the objective function is less than a preset value (e.g., less than 0.001) or the number of training iterations reaches a preset value (e.g., 500 training rounds).

[0086] The embodiments of this application are as follows: Figure 3 The improved AE is stacked as shown. Figure 4 The improved SAE network is shown. During inverse fine-tuning, the connection between label information and hidden layers is removed, and the global fitting result is used as the optimization objective.

[0087] In the specific implementation process, such as Figure 4 The improved SAE network model shown employs a layer-by-layer greedy pre-training method during training (i.e., it is split into multiple AEs, and each AE is trained according to...). Figure 3 (Training is completed in this way). Then combined into Figure 4 The SAE network model was improved by removing the label neurons (i.e., the connections between the neurons and the label information in the diagram), initializing the SAE with the parameters from the pre-training set, and then fine-tuning it in reverse.

[0088] In this embodiment, a pre-training set is constructed by extracting the time-domain and frequency-domain features of historical capacitor current and voltage signals, while a label training set is constructed using the actual capacitance value and actual ESR of the capacitor. Introducing the actual capacitance value and actual ESR as label information during model training provides an efficient and clear framework for pre-training, establishing a correspondence between the model output and the capacitance (C) and ESR. The pre-training set and label training set are input into an SAE model composed of multiple autoencoders (AEs). Each AE is greedily pre-trained layer by layer, and the weights and bias vectors of the SAE model are initialized with the trained parameters. The reconstruction error and label output error are calculated, and the reconstruction error and label output error are combined to form an objective function. Minimizing this objective function is used to fine-tune the SAE model by updating the weights and bias vectors. The stacked progressive autoencoder model with optimal output parameters serves as the capacitor health parameter estimation model. Constructing the objective function by combining the reconstruction error and label output error achieves lossless information and stable state. Furthermore, it clarifies the characteristic significance of capacitor parameter information during the layer-by-layer extraction process, improving the accuracy of capacitor health parameter estimation and thus enhancing the efficiency of capacitor health status monitoring.

[0089] In one possible implementation, step S101, which determines the time-domain and frequency-domain feature information based on the historical capacitor current signal and the historical capacitor voltage signal, includes the following steps:

[0090] S1011, acquire historical capacitor current signal and historical capacitor voltage signal;

[0091] S1012 extracts time-domain features from historical capacitor current signals and historical capacitor voltage signals to obtain time-domain feature information, and performs fast Fourier transform on the time-domain feature information to obtain frequency-domain feature information.

[0092] In one possible implementation, the time-domain feature information includes one or more of the following: the RMS value of the switching ripple current, the RMS value of the switching ripple voltage, the RMS value of the capacitor DC current, the RMS value of the capacitor DC voltage, the switching frequency, the rated capacitance value, and the rated ESR of the capacitor.

[0093] Frequency domain characteristic information includes one or more of the following: power frequency voltage amplitude, power frequency voltage phase, power frequency current amplitude, power frequency current phase, switching frequency ripple voltage amplitude, switching frequency ripple voltage phase, switching frequency ripple current amplitude, and switching frequency ripple current phase.

[0094] In one possible implementation, the objective function is as follows:

[0095]

[0096] in, The reconstruction error is referred to as "the reconstruction error". The label output error; x i The values ​​of neurons in the pre-training set; Indicates reconstructed data values; y i The label values ​​in the labeled training set; is the value of the latent vector; μ is the scaling factor, with a value range of (0, 1); m is the number of label types in the label training set, i = 1, 2, ..., m.

[0097] In this embodiment, the label type corresponding to m includes the actual capacitance value and the actual ESR of the capacitor, and correspondingly, the value of i includes 1 or 2.

[0098] In one possible implementation, step S106 involves back-tuning the SAE model by minimizing the objective function to update the weights and bias vectors, including:

[0099] The weights and bias vectors are updated using the backpropagation algorithm and gradient descent method.

[0100] Figure 5 This is a flowchart illustrating the implementation of a capacitor health status monitoring method according to another embodiment of this application, as shown below. Figure 5 As shown, the method includes the following steps:

[0101] S501 acquires real-time capacitor current and capacitor voltage signals.

[0102] In this embodiment, the controller performing the capacitor health status monitoring method is either a controller connected to the capacitor, or the controller of the battery to which the capacitor belongs, or the controller of a device equipped with a battery. The acquired capacitor current and voltage signals include signals within at least one cycle to facilitate the extraction of time-domain and frequency-domain features.

[0103] S502 extracts real-time time-domain and real-time frequency-domain feature information based on capacitor current and capacitor voltage signals.

[0104] In one possible implementation, step S502, which extracts real-time time-domain feature information and real-time frequency-domain feature information based on the capacitor current signal and capacitor voltage signal, includes:

[0105] Time-domain feature extraction is performed on the capacitor current signal and capacitor voltage signal respectively to obtain real-time time-domain feature information, and fast Fourier transform is performed on the real-time time-domain feature information to obtain real-time frequency-domain feature information.

[0106] In one possible implementation, the time-domain feature information includes one or more of the following: the RMS value of the switching ripple current, the RMS value of the switching ripple voltage, the RMS value of the capacitor DC current, the RMS value of the capacitor DC voltage, the switching frequency, the rated capacitance value, and the rated ESR of the capacitor.

[0107] Frequency domain characteristic information includes one or more of the following: power frequency voltage amplitude, power frequency voltage phase, power frequency current amplitude, power frequency current phase, switching frequency ripple voltage amplitude, switching frequency ripple voltage phase, switching frequency ripple current amplitude, and switching frequency ripple current phase.

[0108] S503 inputs real-time time-domain and real-time frequency-domain feature information into the trained capacitor health parameter estimation model and outputs the capacitance value and ESR.

[0109] S504 calculates the capacitor health status based on the capacitance value and ESR, and uses the capacitor health status to monitor the health status of the capacitor.

[0110] In one possible implementation, step S504, calculating the capacitor health based on the capacitance value and ESR, includes:

[0111] The capacitor health status in both capacitive and voltage-based forms is calculated using the following formulas:

[0112]

[0113]

[0114] Among them, SOH C For capacitor health; SOH R For ESR capacitor health; C e|c C represents the equivalent capacitance value under the current condition. rated R is the nominal rated capacitance. EOL R0 is the ESR in the End-of-Life (EOL) state; R0 is the initial ESR; Rc is the ESR in the current state.

[0115] The average of the capacitance health in both capacitive and voltage-based forms is used as the target capacitance health.

[0116] In this embodiment, the final capacitor health is determined by averaging the capacitor health in both capacitive and voltage-based forms. This averages the error in calculating the capacitor health in a single form, thereby reducing individual deviations between the capacitive and voltage-based capacitor health and improving the accuracy and reliability of the target capacitor health.

[0117] In other possible implementations, depending on the different capacitor health requirements, the capacitor health in terms of capacitance form or voltage form can be calculated separately as the target capacitor health, thereby improving computational efficiency.

[0118] In specific implementation, among other possible implementation methods, after step S504, the following method is also included: outputting a monitoring report based on the capacitor health value and capacitor parameter information, so as to intuitively view the capacitor status.

[0119] Optionally, capacitor parameter information includes one or more of the following: capacitance value, ESR, temperature, and operating conditions.

[0120] In one specific embodiment, the monitoring report format is as follows:

[0121]

[0122] As shown in the table, the monitoring report records the following items: recording time, capacitor health, capacitor value, ESR, temperature, and operating status.

[0123] In one possible implementation, after step S504, the method further includes: generating a health alert message when the capacitor health value exceeds the safe range, so as to promptly remind staff to inspect or maintain the capacitor.

[0124] In this embodiment, the capacitor health parameter estimation is performed based on the capacitor health parameter estimation model provided in the above embodiment, which can improve the accuracy of capacitor health parameter estimation and thus improve the monitoring efficiency of capacitor health status.

[0125] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0126] The following are device embodiments of this application. For details not described in detail, please refer to the corresponding method embodiments described above.

[0127] Figure 6 This is a schematic diagram of the structure of a device for constructing a capacitance health parameter estimation model according to an embodiment of this application, as shown below. Figure 6 As shown, for ease of explanation, only the parts related to the embodiments of this application are shown, such as... Figure 6 As shown, the device includes:

[0128] The training set construction module 601 is used to determine time-domain and frequency-domain feature information based on historical capacitor current signals and historical capacitor voltage signals, and to construct a pre-training set based on the time-domain and frequency-domain feature information; to obtain the actual capacitance value and the actual equivalent series resistance (ESR) of the capacitor, and to construct a label training set based on the actual capacitance value and the actual ESR of the capacitor.

[0129] The model building module 602 is used to construct a stacked progressive autoencoder (SAE) model containing multiple autoencoders (AEs). Each AE is pre-trained layer-by-layer greedily, and the weights and bias vectors of the SAE model are initialized with the trained parameters. Each AE includes a decoder and an encoder. The encoder generates latent vectors using the pre-training set as input, and the decoder generates reconstructed data approximating the pre-training set using the latent vectors as input. The reconstruction error and label output error are calculated, and combined to form an objective function. The reconstruction error is determined based on the reconstructed data and the pre-training set; the label output error is determined based on the latent vectors and the label training set. The SAE model is then fine-tuned by minimizing the objective function to update the weights and bias vectors. The stacked progressive autoencoder model with optimal output parameters serves as the capacitance health parameter estimation model.

[0130] In one possible implementation, the training set construction module 601 is specifically used for:

[0131] Acquire historical capacitor current and voltage signals;

[0132] Time-domain feature extraction is performed on historical capacitor current signals and historical capacitor voltage signals to obtain time-domain feature information, and then fast Fourier transform is performed on the time-domain feature information to obtain frequency-domain feature information.

[0133] In one possible implementation, the model building module 602 is specifically used for:

[0134] The weights and bias vectors are updated using the backpropagation algorithm and gradient descent method.

[0135] Figure 7 This is a schematic diagram of the structure of a capacitor health status monitoring device provided in one embodiment of this application, as shown below. Figure 7 As shown, for ease of explanation, only the parts related to the embodiments of this application are shown, such as... Figure 7 As shown, the device includes:

[0136] The acquisition module 701 is used to acquire the real-time capacitor current signal and capacitor voltage signal.

[0137] Extraction module 702 is used to extract real-time time-domain feature information and real-time frequency-domain feature information based on capacitor current signal and capacitor voltage signal.

[0138] The calculation module 703 is used to input real-time time-domain feature information and real-time frequency-domain feature information into the trained capacitor health parameter estimation model, output the capacitance value and ESR; and calculate the capacitor health degree based on the capacitance value and ESR, so as to realize the health status monitoring of the capacitor based on the capacitor health degree.

[0139] In one possible implementation, the computing module 703 is specifically used for:

[0140] The capacitor health status in both capacitive and voltage-based forms is calculated using the following formulas:

[0141]

[0142]

[0143] Among them, SOH C For capacitor health; SOH R For ESR capacitor health; C e|c C represents the equivalent capacitance value under the current condition. rated R is the nominal rated capacitance. EOL R0 is the ESR in the End-of-Life (EOL) state; R0 is the initial ESR; Rc is the ESR in the current state.

[0144] The average of the capacitance health in both capacitive and voltage-based forms is used as the target capacitance health.

[0145] In one possible implementation, the extraction module 702 is specifically used for:

[0146] Time-domain feature extraction is performed on the capacitor current signal and capacitor voltage signal respectively to obtain real-time time-domain feature information, and fast Fourier transform is performed on the real-time time-domain feature information to obtain real-time frequency-domain feature information.

[0147] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. For example... Figure 8 As shown, the electronic device 8 of this embodiment includes: a processor 80, a memory 81, and a computer program 82 stored in the memory 81 and executable on the processor 80. When the processor 80 executes the computer program 82, it implements the steps in the various method embodiments described above, for example... Figure 1 Steps S101 to S106 shown, or, Figure 5 Steps S501 to S504 are shown. Alternatively, when the processor 80 executes the computer program 82, it implements the functions of each module / unit in the above-described device embodiments, for example... Figure 6 Modules 601 to 602 shown, or, Figure 7 The functions of modules 701 to 703 are shown.

[0148] For example, the computer program 82 can be divided into one or more modules / units, which are stored in the memory 81 and executed by the processor 80 to complete this application. The one or more modules / units can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program 82 in the electronic device 8. For example, the computer program 82 can be divided into... Figure 3 Modules 301 to 303 are shown.

[0149] The electronic device 8 can be a desktop computer, laptop, handheld computer, or cloud server, etc. The electronic device 8 may include, but is not limited to, a processor 80 and a memory 81. Those skilled in the art will understand that... Figure 8 This is merely an example of electronic device 8 and does not constitute a limitation on electronic device 8. It may include more or fewer components than shown, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0150] The processor 80 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0151] The memory 81 can be an internal storage unit of the electronic device 8, such as a hard disk or memory. The memory 81 can also be an external storage device of the electronic device 8, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory 81 can include both internal and external storage units of the electronic device 8. The memory 81 is used to store the computer program and other programs and data required by the electronic device. The memory 81 can also be used to temporarily store data that has been output or will be output.

[0152] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0153] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0154] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0155] In the embodiments provided in this application, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings or direct couplings or communication connections may be through some interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0156] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0157] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0158] If the integrated module / unit is implemented as 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, all or part of the processes in the above-described embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of the above-described embodiments of the capacitor health parameter estimation model construction method or capacitor health status monitoring method. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0159] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for constructing a capacitance health parameter estimation model, characterized in that, include: Time-domain and frequency-domain feature information are determined based on historical capacitor current and voltage signals, and a pre-training set is constructed based on the time-domain and frequency-domain feature information. Obtain the actual capacitance value and the actual equivalent series resistance (ESR) of the capacitor, and construct a label training set based on the actual capacitance value and the actual ESR of the capacitor. Construct a stacked progressive autoencoder (SAE) model containing multiple autoencoders (AEs), greedily pre-train each AE layer by layer, and initialize the weights and bias vectors of the stacked progressive autoencoder SAE model with the trained parameters; each AE includes a decoder and an encoder. The encoder generates latent vectors using the pre-training set as input, and the decoder generates reconstructed data that approximates the pre-training set using the latent vectors as input. Calculate the reconstruction error and the label output error, and combine the reconstruction error and the label output error to form an objective function; wherein, the reconstruction error is determined based on the reconstructed data and the pre-training set; and the label output error is determined based on the latent vector and the label training set. The weights and bias vectors of the stacked progressive autoencoder (SAE) model are back-tuned to minimize the objective function, and the stacked progressive autoencoder model with optimal output parameters is used as the capacitance health parameter estimation model. The step of determining time-domain and frequency-domain feature information based on historical capacitor current and voltage signals includes: Acquire historical capacitor current and voltage signals; Time-domain feature extraction is performed on the historical capacitor current signal and the historical capacitor voltage signal respectively to obtain time-domain feature information, and fast Fourier transform is performed on the time-domain feature information to obtain frequency-domain feature information; The time-domain feature information includes one or more of the following: the effective value of the switching ripple current, the effective value of the switching ripple voltage, the effective value of the capacitor DC current, the effective value of the capacitor DC voltage, the switching frequency, the rated capacitance value, and the rated ESR of the capacitor; the frequency-domain feature information includes one or more of the following: the amplitude of the power frequency voltage, the phase of the power frequency voltage, the amplitude of the power frequency current, the phase of the power frequency current, the amplitude of the switching frequency ripple voltage, the phase of the switching frequency ripple voltage, the amplitude of the switching frequency ripple current, and the phase of the switching frequency ripple current.

2. The method for constructing a capacitance health parameter estimation model according to claim 1, characterized in that, The objective function is as follows: in, The reconstruction error is referred to as "the reconstruction error". The label output error; The values ​​of neurons in the pre-training set; ; The label values ​​in the labeled training set; The value is the implicit vector value; is the scaling factor, with a value range of (0, 1); m is the number of label types in the label training set, i = 1, 2, ..., m.

3. The method for constructing a capacitance health parameter estimation model according to claim 1, characterized in that, The step of back-tuning and updating the weights and bias vector of the stacked progressive autoencoder (SAE) model by minimizing the objective function includes: The weights and the bias vector are updated using the backpropagation algorithm and gradient descent method.

4. A method for monitoring capacitor health status based on the capacitor health parameter estimation model construction method according to any one of claims 1 to 3, characterized in that, include: Acquire real-time capacitor current and capacitor voltage signals; Real-time time-domain feature information and real-time frequency-domain feature information are extracted based on the capacitor current signal and the capacitor voltage signal; The real-time time-domain feature information and the real-time frequency-domain feature information are input into the trained capacitor health parameter estimation model, and the capacitor value and ESR are output. The capacitor health status is calculated based on the capacitance value and the ESR, so as to monitor the health status of the capacitor.

5. The capacitor health status monitoring method according to claim 4, characterized in that, The calculation of capacitor health based on the capacitance value and the ESR includes: The capacitor health in terms of capacitance and ESR are calculated based on the following formulas: in, For capacitor health; For ESR capacitor health; This represents the equivalent capacitance value under the current conditions. The nominal rated capacitance; ESR in the end-of-life (EOL) state; R 0 For the initial ESR; Rc The ESR in the current state; The average of the capacitor health in the form of capacitance and the capacitor health in the form of ESR is calculated as the target capacitor health.

6. The capacitor health status monitoring method according to claim 4, characterized in that, The extraction of real-time time-domain feature information and real-time frequency-domain feature information based on the capacitor current signal and the capacitor voltage signal includes: Time-domain feature extraction is performed on the capacitor current signal and the capacitor voltage signal respectively to obtain real-time time-domain feature information, and fast Fourier transform is performed on the real-time time-domain feature information to obtain real-time frequency-domain feature information.

7. A device for constructing a capacitance health parameter estimation model, characterized in that, include: The training set construction module is used to determine time-domain feature information and frequency-domain feature information based on historical capacitor current signals and historical capacitor voltage signals, and to construct a pre-training set based on the time-domain feature information and the frequency-domain feature information; Obtain the actual capacitance value and the actual equivalent series resistance (ESR) of the capacitor, and construct a label training set based on the actual capacitance value and the actual ESR of the capacitor. The model building module is used to build a stacked progressive autoencoder (SAE) model containing multiple autoencoders (AEs), greedily pre-training each AE layer by layer, and initializing the weights and bias vectors of the stacked progressive autoencoder SAE model with the trained parameters; each AE includes a decoder and an encoder; the encoder generates latent vectors with the pre-training set as input, and the decoder generates reconstructed data that approximates the pre-training set with the latent vectors as input. Calculate the reconstruction error and the label output error, and combine the reconstruction error and the label output error to form an objective function; wherein, the reconstruction error is determined based on the reconstruction data and the pre-training set; the label output error is determined based on the latent vector and the label training set; the weights and the bias vector of the stacked progressive autoencoder (SAE) model are back-tuned to minimize the objective function, and the stacked progressive autoencoder model with optimal output parameters is used as the capacitance health parameter estimation model; The training set construction module is specifically used for: Acquire historical capacitor current and voltage signals; Time-domain feature extraction is performed on the historical capacitor current signal and the historical capacitor voltage signal respectively to obtain time-domain feature information, and fast Fourier transform is performed on the time-domain feature information to obtain frequency-domain feature information; The time-domain feature information includes one or more of the following: the effective value of the switching ripple current, the effective value of the switching ripple voltage, the effective value of the capacitor DC current, the effective value of the capacitor DC voltage, the switching frequency, the rated capacitance value, and the rated ESR of the capacitor; the frequency-domain feature information includes one or more of the following: the amplitude of the power frequency voltage, the phase of the power frequency voltage, the amplitude of the power frequency current, the phase of the power frequency current, the amplitude of the switching frequency ripple voltage, the phase of the switching frequency ripple voltage, the amplitude of the switching frequency ripple current, and the phase of the switching frequency ripple current.

Images