Lithium ion battery connection fault diagnosis method for energy storage equipment

Abstract

The invention discloses an energy storage equipment lithium ion battery connection fault diagnosis method, relates to the technical field of energy storage equipment lithium ion battery management systems, and is used for solving the problem of battery connection faults. According to the method, battery voltage and current data are obtained in real time. And a dynamic multi-modal filtering mechanism is designed, and different filtering methods are adopted for different environmental noise characteristics. And the noise is further accurately identified and filtered by calculating the ambient temperature and humidity and the spectrum characteristics in the battery signal. A data forgetting mechanism is introduced, a self-adaptive window optimization mechanism is adopted, the window length is automatically adjusted according to the voltage fluctuation real-time change rate, the battery state change is quickly responded, and the diagnosis sensitivity and adaptability are improved by dynamically adjusting the time window and the data weight through a weighted updating mechanism. And based on correlation coefficient analysis, the battery connection state is judged, and accurate fault diagnosis is realized. According to the method, the safety and the stability of the energy storage equipment are effectively improved, and the method has high practical application value.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium-ion battery management systems for energy storage devices, and more specifically, to a method for diagnosing connection faults of lithium-ion batteries for energy storage devices. Background Art

[0002] With the development of energy storage technology, lithium-ion batteries are widely used in various application scenarios, especially in the fields of power storage and electric vehicles. Lithium-ion batteries have become the core components of energy storage devices due to their high energy density and long service life. However, battery connection failures, such as poor contact or loose wiring, may lead to degradation of battery performance and even cause safety issues. Traditional fault diagnosis methods mostly rely on simple monitoring of static voltage and current, which makes it difficult to accurately and promptly identify battery connection problems. Existing diagnostic systems generally have defects such as low sensitivity, slow diagnostic speed, and inability to cope with complex environmental interference.

[0003] Most algorithms in the prior art fail to effectively combine the real-time data changes of the battery and external environmental factors, resulting in the inability to provide accurate fault diagnosis results. Therefore, how to design an efficient and accurate fault diagnosis method to quickly respond to battery connection failures and improve diagnostic accuracy has become the focus of current research. In order to solve these problems, the present invention proposes a fault diagnosis method based on real-time monitoring of battery voltage and current, noise suppression, dynamic filtering and adaptive window optimization mechanism, aiming to improve the safety and stability of energy storage equipment.

[0004] In view of the above problems, the present invention proposes a solution. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method for diagnosing connection faults of lithium-ion batteries of energy storage devices to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] In a preferred embodiment, it comprises:

[0008] Step 1: Obtain real-time battery status data during the connection process of the lithium-ion battery of the energy storage device;

[0009] Step 2: Process the real-time status data of the battery during the entire process to remove the noise caused by sensors, electronic interference, environmental factors, and communications that may affect the fault diagnosis algorithm.

[0010] Step 3: Introduction of forgetting mechanism: Introduce calculation window rolling diagnosis as data forgetting mechanism;

[0011] Step 4: After obtaining high-quality real-time battery status data, the correlation coefficient is calculated using the product difference method. The product of the two deviations is used as the basis for the deviations of different variables from their respective means to reflect the degree of correlation between the two battery cell conditions.

[0012] In a preferred embodiment, in step 1, a dedicated lithium battery monitoring IC with a multi-channel voltage detection function that supports synchronous voltage sampling of multiple batteries is used to acquire battery voltage data; a Hall effect sensor with the characteristic of non-contact current measurement and suitable for high-precision, non-destructive testing is used to acquire battery pack current data.

[0013] In a preferred embodiment, in step 2, the EKF algorithm, a filtering technique based on recursive optimal estimation, is used to perform first-order linearization on the nonlinearity, and the physical model and observation data are used to continuously correct the state estimation and filter out the influence of noise;

[0014] The specific formula for removing noise using the EKF algorithm is:

[0015]

[0016] Where f(x k ,u k ) is the nonlinear state equation; g(x k ,u k ) is the nonlinear output equation; w k is the system noise, with a mean of 0 and a covariance of Q k ,v k is the system observation noise, with a mean of 0; the covariance R k , v k And w k Unrelated to each other;

[0017] Furthermore, in order to deal with nonlinear discrete systems, we use the state function f(xk,uk) and the measurement function g(xk,uk); at time k, by performing a first-order Taylor expansion on these nonlinear functions:

[0018]

[0019] definition is the state transfer matrix, is the observation matrix, rewritten as:

[0020]

[0021] Furthermore, the complete recursive process of the extended Kalman filter algorithm is as follows:

[0022]

[0023] Furthermore, the prior estimates of the state variables are:

[0024] x k / k-1 =f(x k ,u k-1 )

[0025] Furthermore, the error covariance matrix of the prediction k time is:

[0026]

[0027] Furthermore, the gain matrix Kk is calculated as:

[0028]

[0029] Furthermore, the state estimate is updated according to the observed value Uk:

[0030] x k =x k / k-1 +K k [y k -C k x k / k-1 ]

[0031] Furthermore, the error covariance matrix is ​​updated as:

[0032] P k =(ik k C k ) k / k-1

[0033] Furthermore, the generated error is:

[0034] e k =v exp -v mdl

[0035] Design a dynamic multimodal filtering mechanism: By combining multiple filtering methods, the most appropriate filtering mode is dynamically selected according to different environmental data and signal characteristics.

[0036] The temperature and humidity sensor is used to measure the ambient temperature and humidity to provide environmental data when the battery is running. The time domain signal is converted to the frequency domain through fast Fourier transform to analyze the spectrum characteristics in the battery voltage and current signals.

[0037] The noise characteristic coefficient is determined by weighted calculation of environmental data and signal characteristic data, according to the formula: ZXS=Qw×W+Qs×S+Qg×G+Qd×D+Tw×W 2 +Ts×S 2, where; ZXS represents the noise characteristic coefficient, W represents temperature, S represents humidity, G represents high-frequency noise, D represents low-frequency noise, Qw, Qs, Qg, Qd, Tw, Ts are weighting coefficients;

[0038] The noise coefficient threshold Yz is set. When the noise characteristic coefficient is greater than the threshold Yz, the noise characteristic of the battery real-time status data is high noise data, and a high-pass filter and a band-pass filter are used for this type of noise; when the noise characteristic coefficient is less than or equal to the threshold Yz, the noise characteristic of the battery real-time status data is low noise data, and a low-pass filter is used for this type of noise.

[0039] In a preferred embodiment, in step 3, a data forgetting mechanism is introduced into the diagnosis algorithm, and a window rolling diagnosis is calculated as the data forgetting mechanism;

[0040] The formula is as follows:

[0041] P i =P i-1 +x 1+sp-i y 1+sp-i

[0042] Q i =Q i-1 +x 1+sp-i

[0043] R i =R i-1 +y 1+sp-i

[0044]

[0045] Where P1 = x sp y sp Q1=x sp R1=y sp i takes the value of 2≤i≤d, d is the calculation window length, sp is the position of the starting point in the sequence, and increases during the calculation process; and x 1+sp-i ,y 1+sp-i is the voltage value at 1+sp-i of the x,y sampling sequence; at the same time (r x,y ) k is the correlation coefficient of the x, y sampling series from the initial sampling point k point;

[0046] Design an adaptive window optimization mechanism: automatically adjust the window length according to the real-time change rate of voltage fluctuation, and add a weighted update mechanism for the window content. The real-time change rate of voltage fluctuation is calculated based on the formula: Wherein, Vt represents the voltage value at the tth moment, Tt represents the timestamp at the tth moment, Vt-1 represents the voltage value at the previous moment, and Tt-1 represents the timestamp at the previous moment;

[0047] Set the change rate threshold Yb. When the real-time change rate of voltage fluctuation is greater than or equal to the change rate threshold Yb, the battery is considered to be in an abnormal state and the length of the calculation window is shortened to Lmin. When the real-time change rate of voltage fluctuation is less than the change rate threshold Yb, the battery is considered to be stable and the length of the calculation window is extended to Lmax.

[0048] Different weights are assigned to the voltage fluctuation real-time change rate data at each moment in the time window T, according to the formula: Where Wi is the weight coefficient, which is calculated based on the time distance of the real-time change rate data of voltage fluctuation, using the exponential weighting method: w i =α T-i Among them, α is a weighting coefficient, which is approximately a constant between 0 and 1; the weighting coefficient and the window length are dynamically adjusted according to the change rate of the battery voltage fluctuation.

[0049] In a preferred embodiment, in step 4, the correlation coefficient is calculated by the product difference method, based on the deviation of different variables from their respective means, and the product of the two deviations reflects the correlation between the two battery cell conditions.

[0050] The specific calculation formula is as follows:

[0051]

[0052] Among them, r x,y is the correlation coefficient of the two series x and y; σ xy is the covariance of the two series x and y; σ x σ y is the variance of the two series x and y;

[0053] The value of the correlation coefficient is between -1 and +1, when 0≤|r x,y |≤1, indicating that there is a certain degree of linear correlation between the variables in the two series x and y; and |r x,y The closer | is to 1, the closer the linear relationship between the two series x and y is; |r x,y The closer | is to 0, the weaker the linear relationship between the two series x and y;

[0054] Considering that the deviation between the variable and its mean in the calculation formula of the correlation coefficient must be calculated separately, the identity transformation is performed on it:

[0055]

[0056] where x i ,y i are the elements of the two sequences x and y respectively; n is the length of the two sequences x and y.

[0057] The invention discloses a method for diagnosing a connection fault of a lithium-ion battery of an energy storage device, which relates to the technical field of lithium-ion battery management systems for energy storage devices and is used to solve the problem of battery connection faults; the method obtains battery voltage and current data in real time. A dynamic multimodal filtering mechanism is designed, and different filtering methods are adopted for different environmental noise characteristics. By calculating the ambient temperature and humidity and the spectrum characteristics in the battery signal, the noise is further accurately identified and filtered out. A data forgetting mechanism is introduced, and an adaptive window optimization mechanism is adopted. The window length is automatically adjusted according to the real-time change rate of voltage fluctuation, and the battery state changes are quickly responded to. Through a weighted update mechanism, the time window and data weight are dynamically adjusted to improve the diagnostic sensitivity and adaptability. Based on the correlation coefficient analysis, the battery connection state is judged to achieve accurate fault diagnosis. The method effectively improves the safety and stability of energy storage equipment and has a strong practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is a schematic diagram of the operation of a method for diagnosing connection faults of lithium-ion batteries of energy storage devices according to the present invention. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] The present invention is collected.

[0061] Example

[0062] The present invention discloses a method for diagnosing a connection fault of a lithium-ion battery of an energy storage device, comprising:

[0063] Step 1: Obtain the real-time status data of the lithium-ion battery of the energy storage device during connection, including: battery voltage, battery pack current;

[0064] Specifically, a dedicated lithium battery monitoring IC (such as TI's BQ series) with multi-channel voltage detection function that supports synchronous voltage sampling of multiple batteries is used to obtain battery voltage data; a Hall effect sensor (such as Allegro ACS712) with the characteristics of non-contact current measurement and suitable for high-precision, non-destructive testing is used to obtain battery pack current data; for example, the sample data obtained is as follows:

[0065] First second:

[0066] Battery 1 voltage: 3.65V;

[0067] Battery 2 voltage: 3.64V;

[0068] Battery pack current: 12.3A;

[0069] Second 2:

[0070] Battery 1 voltage: 3.66V;

[0071] Battery 2 voltage: 3.65V;

[0072] Battery pack current: 12.1A;

[0073] Step 2: Process the real-time status data of the battery during the entire process to remove noise problems caused by sensors, electronic interference, environmental factors, and communications that affect the fault diagnosis algorithm.

[0074] Specifically, the operating state (such as voltage and current) of a lithium-ion battery is closely related to its internal chemical process and external connection conditions, and is usually manifested as a nonlinear system. In this embodiment, the EKF algorithm, a filtering technology based on recursive optimal estimation, is used to perform first-order linearization of the nonlinearity, and the physical model and observation data are used to continuously correct the state estimate to filter out the influence of noise.

[0075] The specific formula for removing noise using the EKF algorithm is:

[0076]

[0077] Where f(x k ,u k ) is the nonlinear state equation; g(x k ,u k ) is the nonlinear output equation; w k is the system noise, with a mean of 0 and a covariance of Q k ,v k is the system observation noise, with a mean of 0; the covariance R k , v k And w k Not related to each other.

[0078] Furthermore, in order to deal with nonlinear discrete systems, we use the state function f(xk,uk) and the measurement function g(xk,uk). At time k, by performing a first-order Taylor expansion on these nonlinear functions:

[0079]

[0080] definition is the state transfer matrix, is the observation matrix, rewritten as:

[0081]

[0082] Furthermore, the complete recursive process of the extended Kalman filter algorithm is as follows:

[0083]

[0084] Furthermore, the prior estimates of the state variables are:

[0085] x k / k-1 =f(x k ,u k-1 )

[0086] Furthermore, the error covariance matrix of the prediction k time is:

[0087]

[0088] Furthermore, the gain matrix Kk is calculated as:

[0089]

[0090] Furthermore, the state estimate is updated according to the observed value Uk:

[0091] x k =x k / k-1 +K k [y k -C k x k / k-1 ]

[0092] Furthermore, the error covariance matrix is ​​updated as:

[0093] P k =(ik k C k ) k / k-1

[0094] Furthermore, the generated error is:

[0095] e k =v exp -v mdl

[0096] In this way, the EKF algorithm can gradually adjust the state estimate and remove noise from the data.

[0097] It should be noted that the EKF algorithm is suitable for processing single noise characteristics; however, in the actual operation of the battery, the noise sources are diverse, and the EKF algorithm cannot effectively process all types of noise. Different types of noise require different processing strategies. Since the type and intensity of noise are closely related to environmental factors and signal characteristics, in this embodiment, a dynamic multimodal filtering mechanism is designed: by combining multiple filtering methods, the most appropriate filtering mode is dynamically selected according to different environments and signal characteristics to achieve more accurate data processing and noise suppression.

[0098] Specifically, temperature changes will directly affect the battery's voltage, current, and internal impedance. In a high-temperature environment, the battery's internal chemical reaction rate is accelerated, which may lead to unstable battery voltage and more obvious low-frequency noise. In a low-temperature environment, the battery's internal resistance increases, the voltage fluctuation during charging and discharging increases, and the noise characteristics tend to be low-frequency and slowly changing. In a high-humidity environment, corrosion may occur on the battery surface or external wiring, resulting in increased signal fluctuations and noise. The insulation of the battery pack may be reduced, which may cause current pulsation or voltage instability. This interference is usually low-frequency and continuous. At the same time, the battery's current and voltage data are interfered with by external signals (such as electromagnetic interference and radio frequency interference), which will also affect the noise. For example, high-frequency interference with noise usually above 200Hz will affect the accuracy of the battery monitoring signal; low-frequency interference with noise usually below 50Hz will cause periodic fluctuations in the battery voltage or current signal, causing signal distortion. Therefore, in this embodiment, a temperature and humidity sensor (such as DHT22) is used to provide environmental information during battery operation by measuring ambient temperature and humidity; and a fast Fourier transform (FFT) is used to convert time domain signals into frequency domain to analyze the frequency spectrum characteristics in the battery voltage and current signals; the following sample data is obtained:

[0099] 1st time:

[0100] Battery 1 voltage: 3.65V

[0101] Battery 2 voltage: 3.64V

[0102] Battery pack current: 12.3A

[0103] Ambient temperature: 30℃

[0104] Ambient humidity: 75%

[0105] High frequency noise: 250Hz

[0106] Low frequency noise: 40Hz

[0107] 2nd time:

[0108] Battery 1 voltage: 3.66V

[0109] Battery 2 voltage: 3.65V

[0110] Battery pack current: 12.1A

[0111] Ambient temperature: 45℃

[0112] Ambient humidity: 60%

[0113] High frequency noise: 264Hz

[0114] Low frequency noise: 35Hz

[0115] Furthermore, the noise characteristic coefficient is determined by weighted calculation of environmental data and signal characteristic data, specifically according to the formula: ZXS = Qw×W + Qs×S + Qg×G + Qd×D + Tw×W 2 +Ts×S 2 , where ZXS represents the noise characteristic coefficient, W represents temperature, S represents humidity, G represents high-frequency noise, D represents low-frequency noise, Qw, Qs, Qg, Qd, Tw, and Ts are weighting coefficients, which are set according to experimental data or historical data; for example: by calculation, Qw is 0.02, Qs is 0.1, Qg is 1, Qd is 0.5, Tw is 0.0001, and Ts is 0.002. Substituting the sample number into the noise characteristic coefficient of the first battery real-time status data is 279.315, and the noise characteristic coefficient of the second battery real-time status data is 295.803;

[0116] Set the noise coefficient threshold Yz. When the noise characteristic coefficient is greater than the threshold Yz, the noise characteristic of the battery real-time status data is high noise data. For this type of noise, a high-pass filter and a band-pass filter are used to remove high-frequency noise and ensure the stability of the battery data. When the noise characteristic coefficient is less than or equal to the threshold Yz, the noise characteristic of the battery real-time status data is low noise data. For this type of noise, a low-pass filter is used to remove lower frequency noise. For example: Set the noise coefficient threshold to 288. The noise characteristic coefficient of the first battery real-time status data obtained through the example data is 279.315 less than or equal to 288. The first battery real-time status data is low noise data. A low-pass filter is used to remove lower frequency noise; the noise characteristic coefficient of the second battery real-time status data is 295.803 greater than 288. The second battery real-time status data is high noise data. A high-pass filter and a band-pass filter are used to remove high-frequency noise and ensure the stability of the battery data.

[0117] Step 3: Introducing the forgetting mechanism

[0118] During operation, the battery management system will continuously sample the voltage of the single cell. Since the diagnostic algorithm in this article is calculated based on the correlation coefficient, when the voltage sampling series is too long, the abnormal jump of the sampled voltage at a certain place due to a loose connection failure may not be reflected in the calculation result of the correlation coefficient, that is, the sensitivity of the algorithm and the effectiveness of diagnosis will decrease as the data increases. Therefore, in this embodiment, a data forgetting mechanism is introduced in the diagnostic algorithm to ensure that the sensitivity and effectiveness of the algorithm will not decrease as the system operation time increases. The calculation window rolling diagnosis is introduced as a data forgetting mechanism.

[0119] Furthermore, the specific formula is as follows:

[0120] P i =P i-1 +x 1+sp-i y 1+sp-i

[0121] Q i =Q i-1 +x 1+sp-i

[0122] R i =R i-1 +y 1+sp-i

[0123]

[0124] Among them, i is 2≤i≤d, d is the calculation window length, P1=x sp y sp ; Q1 = x sp ; R1=y sp ; sp is the position of the starting point in the sequence, which increases during the calculation process; x 1+sp-i ,y 1+sp-i is the voltage value at 1+sp-i of the x, y sampling sequence; (r x,y ) k is the correlation coefficient of the x and y sampling series from the kth point of the initial sampling point.

[0125] It should be noted that a fixed rolling window may be unable to adapt to the different frequency characteristics of faults. In this embodiment, an adaptive window optimization mechanism is designed: the window length is automatically adjusted according to the real-time change rate of voltage fluctuations, and a weighted update mechanism for the window content is added. By weighting data from different periods, more accurate and efficient diagnosis can be achieved, further improving the sensitivity and adaptability of the diagnosis.

[0126] The calculation of the real-time change rate of voltage fluctuation is based on the formula: Among them, Vt represents the voltage value at the tth moment, Tt represents the timestamp at the tth moment, Vt-1 represents the voltage value at the previous moment, and Tt-1 represents the timestamp at the previous moment. The following example:

[0127] The real-time data of the monitored battery voltage are as follows:

[0128] 1st second: Battery voltage: 3.65V

[0129] Second 2: Battery voltage: 3.64V

[0130] 3rd second: Battery voltage: 3.75V

[0131] 4th second: Battery voltage: 3.70V

[0132] 5th second: Battery voltage: 3.80V

[0133] Substituting the sample data into the formula, we get:

[0134] Second sec: rate of change = (3.64-3.65) / (2-1) = -0.01 V / s

[0135] 3rd second: rate of change = (3.75-3.64) / (3-2) = 0.11 V / s

[0136] 4th second: rate of change = (3.70-3.75) / (4-3) = -0.05 V / s

[0137] 5th second: rate of change = (3.80-3.70) / (5-4) = 0.10 V / s

[0138] The real-time change rate of voltage fluctuation is used as the trigger condition for window adjustment, and the size of the window is determined by the real-time change rate of voltage fluctuation. Specifically, the change rate threshold Yb is set. When the real-time change rate of voltage fluctuation is greater than or equal to the change rate threshold Yb, it is considered that the battery is in an abnormal state and needs to respond quickly, so the length of the calculation window is shortened to Lmin; when the real-time change rate of voltage fluctuation is less than the change rate threshold Yb, it is considered that the battery state is stable, and a longer calculation window can be used to reduce the impact of noise on data analysis, so the length of the calculation window is extended to Lmax; for example: setting the change rate threshold Yb to 0.1V / s, the fluctuation rate in the 3rd and 5th seconds exceeds the 0.1V / s threshold, the window length should be shortened to Lmin, and in the 2nd and 4th seconds, the fluctuation rate is small, and the window length should be extended to Lmax.

[0139] Furthermore, a weighted update mechanism is introduced to more accurately reflect changes in battery status by assigning different weights to data in different time periods. This improves diagnostic sensitivity and adaptability, ensuring that the battery status detection algorithm does not miss short-term fluctuation anomalies, but can also smooth long-term stable signals, thereby improving diagnostic accuracy.

[0140] Specifically, different weights are assigned to the data at each moment in the time window T, according to the formula: Among them, Wi is the weight coefficient, which is calculated according to the time distance of the data, using the exponential weighting method: w i =α T-i Among them, α is the weighting coefficient, which is approximately a constant between 0 and 1. If the data point is closer, it is given a higher weight to ensure that the system responds to real-time data first; if the data point is farther away, the weight is lower to reduce its impact on the current state.

[0141] The weighting coefficient and window length are dynamically adjusted according to the rate of change of battery voltage fluctuations. When the battery voltage changes greatly, it means that the battery state has changed rapidly. The weighted update window should be shortened so that the latest data has a greater weight. When the battery voltage changes tend to be stable, the window length can be appropriately increased to introduce more historical data and improve the assessment of the long-term stability of the battery state.

[0142] Step 4: After obtaining high-quality data, the correlation coefficient needs to be calculated using the product difference method. The product of the two deviations reflects the degree of correlation between the two battery cell conditions based on the deviations between the different variables and their respective means to perform fault diagnosis.

[0143] Furthermore, the specific calculation formula is as follows:

[0144]

[0145] Among them, r x,y is the correlation coefficient of the two series x and y; σ xy is the covariance of the two series x and y; σ x , σ y is the variance of the two series x and y.

[0146] The value of the correlation coefficient is between -1 and +1. When 0≤|r x,y When |≤1, it indicates that there is a certain degree of linear correlation between the variables in the two series x and y. x,y |The closer it is to 1, the closer the linear relationship between the two series x and y is; |r x,y The closer to 0, the weaker the linear relationship between the two series x and y.

[0147] Furthermore, considering that the deviation of the variable from its mean in the calculation formula of the correlation coefficient must be calculated separately, it cannot meet the real-time requirements of the loose connection fault diagnosis algorithm, so an identity transformation is performed on it:

[0148]

[0149] Among them, x i ,y i are the elements of the two sequences x and y respectively; n is the length of the two sequences x and y.

[0150] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0151] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0152] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application of the technical solution and the invention constraints. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0153] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0154] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0155] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for diagnosing connection faults of lithium-ion batteries in energy storage equipment. It is characterized by; including: Step 1: Obtain real-time battery status data during the connection process of the lithium-ion battery of the energy storage device; Step 2: Process the real-time status data of the battery during the entire process to remove the noise caused by sensors, electronic interference, environmental factors, and communications that may affect the fault diagnosis algorithm. Step 3: Introduction of forgetting mechanism: Introduce calculation window rolling diagnosis as data forgetting mechanism; Step 4: After obtaining high-quality real-time battery status data, the correlation coefficient is calculated using the product difference method. The product of the two deviations is used as the basis for the deviations of different variables from their respective means to reflect the degree of correlation between the two battery cell conditions.

2. A method for diagnosing connection faults of lithium-ion batteries of energy storage devices according to claim 1, characterized in that: In step 1, a dedicated lithium battery monitoring IC with multi-channel voltage detection function that supports synchronous voltage sampling of multiple batteries is used to acquire battery voltage data; a Hall effect sensor with the characteristic of non-contact current measurement and suitable for high-precision, non-destructive testing is used to acquire battery pack current data.

3. A method for diagnosing connection faults of lithium-ion batteries of energy storage equipment according to claim 2, characterized in that: In step 2, the EKF algorithm, a filtering technique based on recursive optimal estimation, is used to perform first-order linearization on the nonlinearity, and the physical model and observation data are used to continuously correct the state estimation and filter out the influence of noise; The specific formula for removing noise using the EKF algorithm is: Where f(x k ,u k ) is the nonlinear state equation; g(x k ,u k ) is the nonlinear output equation; w k is the system noise, with a mean of 0 and a covariance of Q k ,v k is the system observation noise, with a mean of 0; the covariance R k , v k And w k Unrelated to each other; Furthermore, in order to deal with nonlinear discrete systems, we use the state function f(xk,uk) and the measurement function g(xk,uk); at time k, by performing a first-order Taylor expansion on these nonlinear functions: definition is the state transfer matrix, is the observation matrix, rewritten as: Furthermore, the complete recursive process of the extended Kalman filter algorithm is as follows: Furthermore, the prior estimates of the state variables are: x k / k-1 =f(x k ,u k-1 ) Furthermore, the error covariance matrix of the prediction k time is: Furthermore, the gain matrix Kk is calculated as: Furthermore, the state estimate is updated according to the observed value Uk: x k =x k / k-1 +K k [y k -C k x k / k-1 ] Furthermore, the error covariance matrix is ​​updated as: P k =(i-k k C k )P k / k-1 Furthermore, the generated error is: yes k =v exp -v mdl Design a dynamic multimodal filtering mechanism: By combining multiple filtering methods, the most appropriate filtering mode is dynamically selected according to different environmental data and signal characteristics. The temperature and humidity sensor is used to measure the ambient temperature and humidity to provide environmental data when the battery is running. The time domain signal is converted to the frequency domain through fast Fourier transform to analyze the spectrum characteristics in the battery voltage and current signals. The noise characteristic coefficient is determined by weighted calculation of environmental data and signal characteristic data, specifically according to the formula: ZXS = Qw×W + Qs×S + Qg×G + Qd×D + Tw×W + Ts×S2, where; ZXS represents the noise characteristic coefficient, W represents temperature, S represents humidity, G represents high-frequency noise, D represents low-frequency noise, Qw, Qs, Qg, Qd, Tw, Ts are weighting coefficients; Set the noise coefficient threshold Yz. When the noise characteristic coefficient is greater than the threshold Yz, the noise characteristic of the real-time battery status data is high noise data. A high-pass filter and a band-pass filter are used for this type of noise. When the noise characteristic coefficient is less than or equal to the threshold value Yz, the noise characteristic of the real-time battery status data is low-noise data, and a low-pass filter is used for this type of noise.

4. A method for diagnosing connection faults of lithium-ion batteries of energy storage equipment according to claim 3, characterized in that; In step 3, a data forgetting mechanism is introduced into the diagnosis algorithm, and a window rolling diagnosis is calculated as the data forgetting mechanism; The formula is as follows: P i =P i-1 +x 1+sp-i y 1+sp-i Q i =Q i-1 +x 1+sp-i R i =R i-1 +y 1+sp-i Where P1 = x sp y sp Q1=x sp R1=y sp i takes the value of 2≤i≤d, d is the calculation window length, sp is the position of the starting point in the sequence, and increases during the calculation process; and x 1+sp-i ,y 1+sp-i is the voltage value at 1+sp-i of the x,y sampling sequence; at the same time (r x,y ) k is the correlation coefficient of the x, y sampling series from the initial sampling point k point; Design an adaptive window optimization mechanism: automatically adjust the window length according to the real-time change rate of voltage fluctuation, and add a weighted update mechanism for the window content. The real-time change rate of voltage fluctuation is calculated based on the formula: Wherein, Vt represents the voltage value at the tth moment, Tt represents the timestamp at the tth moment, Vt-1 represents the voltage value at the previous moment, and Tt-1 represents the timestamp at the previous moment; Set the change rate threshold Yb. When the real-time change rate of voltage fluctuation is greater than or equal to the change rate threshold Yb, the battery is considered to be in an abnormal state and the length of the calculation window is shortened to Lmin. When the real-time change rate of voltage fluctuation is less than the change rate threshold Yb, the battery is considered to be stable and the length of the calculation window is extended to Lmax. Different weights are assigned to the voltage fluctuation real-time change rate data at each moment in the time window T, according to the formula: Where Wi is the weight coefficient, which is calculated based on the time distance of the real-time change rate data of voltage fluctuation, using the exponential weighting method: w i =α T-i Among them, α is a weighting coefficient, which is approximately a constant between 0 and 1; the weighting coefficient and the window length are dynamically adjusted according to the change rate of the battery voltage fluctuation.

5. A method for diagnosing connection faults of lithium-ion batteries of energy storage equipment according to claim 4, characterized in that: In step 4, the correlation coefficient is calculated by the product difference method, based on the deviation of different variables from their respective means, and the product of the two deviations reflects the degree of correlation between the two battery cell conditions. The specific calculation formula is as follows: Among them, r x,y is the correlation coefficient of the two series x and y; σ xy is the covariance of the two series x and y; σ x σ y is the variance of the two series x and y; The value of the correlation coefficient is between -1 and +1. When 0≤|r x,y |≤1, indicating that there is a certain degree of linear correlation between the variables in the two series x and y; and |r x,y The closer | is to 1, the closer the linear relationship between the two series x and y is; |r x,y The closer | is to 0, the weaker the linear relationship between the two series x and y; Considering that the deviation between the variable and its mean in the calculation formula of the correlation coefficient must be calculated separately, the identity transformation is performed on it: where x i ,y i are the elements of the two sequences x and y respectively; n is the length of the two sequences x and y.

Images