Fault diagnosis methods for lithium battery voltage and current sensors

Abstract

This invention belongs to the field of battery fault diagnosis, and particularly relates to a fault diagnosis method for lithium battery voltage and current sensors. The method includes: constructing an equivalent circuit model of a lithium-ion battery and identifying its parameters; constructing an augmented system model with a singular structure, treating the faults of the voltage and current sensors as augmented states; constructing an augmented state observer; designing the gain of the augmented state observer so that the poles of its error dynamic system are located within a preset disk domain; running the augmented state observer to simultaneously output a first estimate of the battery system state and a second estimate of the sensor fault; and achieving fault diagnosis for the sensor. This invention solves the problems of unstable and lagging sensor fault identification and inability to accurately estimate the specific shape of faults under dynamic operating conditions using traditional methods, and can achieve high-precision real-time estimation of various complex faults such as bias, periodicity, and intermittency.

Description

Technical Field

[0001] This invention belongs to the field of battery fault diagnosis, and particularly relates to a fault diagnosis method for lithium battery voltage and current sensors. Background Technology

[0002] Battery management systems (BMS) rely on accurate measurements from voltage and current sensors for State of Charge (SOC) estimation and safety protection. Sensor malfunctions such as bias drift, intermittent loss of signal, and EMI interference can lead to misjudgments of SOC, increased risk of overcharging / over-discharging, and even thermal runaway. Commonly used battery sensor fault diagnosis methods include thresholding, data-driven approaches, and model-based methods. This patent employs a model-based approach, which offers stronger physical interpretability compared to thresholding and data-driven methods, and does not require extensive labeled data or training on datasets. Furthermore, the proposed observer method demonstrates superior performance in estimating complex faults. Summary of the Invention

[0003] This invention aims to overcome the shortcomings of existing technologies by providing a fault diagnosis method for lithium battery voltage and current sensors. This method can solve the problems of instability and lag in identifying sensor faults such as bias-type, periodic disturbance-type, and random noise-type faults under dynamic operating conditions, as well as the problem that when estimating low-frequency fluctuation-type and intermittent faults of voltage sensors, i.e., bias faults, it can only determine the time of fault occurrence but not the specific shape of the fault.

[0004] To solve the above-mentioned technical problems, the present invention is implemented as follows:

[0005] A fault diagnosis method for a lithium battery voltage and current sensor includes the following steps:

[0006] Step 1: Construct an equivalent circuit model of a lithium-ion battery and identify its parameters;

[0007] Step 2: Treat the faults of the voltage sensor and current sensor as augmented states, and construct an augmented system model with a singular structure;

[0008] Step 3: Based on the augmented system model described in Step 2, construct an augmented state observer; design the gain of the augmented state observer based on the linear matrix inequality method so that the poles of its error dynamic system are located within a preset disk domain, thereby ensuring the convergence speed and robustness of fault estimation.

[0009] Step 4: Run the augmented state observer described in Step 3 to synchronously output a first estimate of the battery system state and a second estimate of the faults of the voltage and current sensors; based on the second estimate, diagnose the sensor faults.

[0010] Furthermore, the step 4, which involves diagnosing faults in the voltage sensor and current sensor based on the second estimated value, includes: performing fault detection based on whether the second estimated value is zero; and / or identifying and estimating the fault type based on the magnitude or waveform of the second estimated value.

[0011] Furthermore, the equivalent circuit model described in step 1 is a second-order RC model, and its state equation is:

[0012]

[0013] in Open circuit voltage, For the terminal voltage, For ohmic internal resistance, and Polarization voltage, , For polarization internal resistance, , Polarizing capacitor; It represents the current; it is positive when discharging and negative when charging.

[0014] Furthermore, in step 1, the five variables R0, R... are tested using a hybrid pulse power characteristic (HPPC) experiment on the equivalent circuit model parameters. p1 C p1 R p2 C p2 Identify and determine the open circuit voltage. The relationship between the battery state of charge (SOC) and the state of charge (SOC) is processed by piecewise linearization.

[0015] Furthermore, in step 2, the augmented model with the singular structure satisfies the following condition:

[0016]

[0017] It is a state vector. It is the output vector of the measurement. It is the system's input vector. It's a sensor malfunction.

[0018] Furthermore, the construction of the augmented state observer described in step 3 specifically involves constructing an observer of the following form:

[0019]

[0020] in , This is the observer gain matrix to be designed.

[0021] Furthermore, in step 3, given the disk region If there exists a symmetric matrix The following inequalities must be satisfied:

[0022]

[0023] Then the system matrix All eigenvalues ​​are located at Inside;

[0024] Substitute the system matrix and use Linearization yields the equivalence criterion.

[0025]

[0026] Solve , Later generations This ensures that the error poles fall within the preset disk region, thus guaranteeing the convergence of the fault.

[0027] Furthermore, the sensor fault includes at least one of bias fault, periodic disturbance fault, random noise fault, low-frequency fluctuation fault, and intermittent loss fault.

[0028] A battery management system includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the above-described method.

[0029] This invention addresses two core issues in the estimation and diagnosis of voltage and current sensor faults in electric vehicle lithium-ion batteries: first, how to solve the problems of instability and lag in identifying bias-type, periodic disturbance-type, and random noise-type sensor faults under dynamic operating conditions (such as UDDS); second, how to solve the problem that existing methods can only determine the time of occurrence of low-frequency fluctuation-type and intermittent faults, i.e., bias faults, but not the specific shape of the fault; it can achieve high-precision real-time estimation of various complex faults such as bias, periodic, and intermittent faults. Figure 2 As shown in Figure 3, this model can estimate the shape of the fault quite well, solving the problem that previous augmented observers could only determine the occurrence time but not the specific shape when observing low-frequency faults in voltage sensors. In the absence of a fault, the error in the terminal voltage output by the system is within... This also proves the accuracy of the parameter identification method. Attached Figure Description

[0030] Figure 1 This is the estimation result of the voltage sensor of the present invention when intermittent faults occur;

[0031] Figure 2 This is the estimation result of the current sensor of the present invention when intermittent faults occur;

[0032] Figure 3 For the present invention and Linearization methods between them;

[0033] Figure 4 The voltage curve is measured for this invention;

[0034] Figure 5 The comparison results and error rate between the output voltages after parameter identification in this invention;

[0035] Figure 6 For the fault-free condition of this invention Estimate the situation;

[0036] Figure 7 This is an estimate of the terminal voltage of the present invention;

[0037] Figure 8 This is the battery testing platform for the present invention;

[0038] Figure 9 A comparison chart showing the results with the traditional extended-dimensional system PD observer;

[0039] Figure 10 This is a comparison chart with the traditional extended-dimensional system PD observer. Specific Implementation

[0040] The present invention will now be described in detail through specific embodiments. These embodiments are provided to enable a more thorough understanding of the invention and to fully convey the scope of the invention to those skilled in the art. As used throughout the specification and claims, the terms "comprising" or "including" are open-ended and are interpreted as "comprising but not limited to". The following description is a preferred embodiment for carrying out the invention; however, this description is intended to illustrate the general principles of the specification and is not intended to limit the scope of the invention. The scope of protection of the present invention is determined by the appended claims.

[0041] A fault diagnosis method for a lithium battery voltage and current sensor, wherein the sensor fault includes at least one of bias fault, periodic disturbance fault, random noise fault, low-frequency fluctuation fault, and intermittent loss fault, and the specific diagnosis method includes the following steps:

[0042] Step 1: Construct an equivalent circuit model of a lithium-ion battery and identify its parameters;

[0043] Step 2: Treat the faults of the voltage sensor and current sensor as augmented states, and construct an augmented system model with a singular structure;

[0044] Step 3: Based on the augmented system model described in Step 2, construct an augmented state observer; design the gain of the augmented state observer based on the linear matrix inequality method so that the poles of its error dynamic system are located within a preset disk domain, thereby ensuring the convergence speed and robustness of fault estimation.

[0045] Step 4: Run the augmented state observer described in Step 3 to synchronously output a first estimate of the battery system state and a second estimate of the sensor fault; based on the second estimate, diagnose the sensor fault.

[0046] The equivalent circuit model described in step 1 of this invention is a second-order RC model, and its state equation is:

[0047]

[0048] in Open circuit voltage, For the terminal voltage, For ohmic internal resistance, and Polarization voltage, , For polarization internal resistance, , Polarizing capacitor; This represents the current, which is positive during discharge and negative during charging. This invention uses a hybrid pulse power characteristic (HPPC) experiment to measure the five variables R0, R... p1 C p1 R p2 C p2 Identify and determine the open circuit voltage. The relationship between the battery state of charge (SOC) and the state of charge (SOC) is processed by piecewise linearization.

[0049] In step 2 of this invention, the augmented model with a singular structure satisfies the following condition:

[0050] ;

[0051] It is a state vector. It is the output vector of the measurement. It is the system's input vector. It's a sensor malfunction.

[0052] The construction of the augmented state observer in step 3 of this invention specifically involves constructing an observer of the following form:

[0053]

[0054] in , This is the observer gain matrix to be designed.

[0055] Given disk region If there exists a symmetric matrix The following inequalities must be satisfied:

[0056]

[0057] Then the system matrix All eigenvalues ​​are located at Inside;

[0058] Substitute the system matrix and use Linearization yields the equivalence criterion.

[0059]

[0060] Solve , Later generations This ensures that the error poles fall within the preset disk region, thus guaranteeing the convergence of the fault.

[0061] The step 4 of this invention, which involves diagnosing sensor faults based on the second estimated value, includes: performing fault detection based on whether the second estimated value is zero; and / or identifying and estimating the fault type based on the numerical value or waveform shape of the second estimated value.

[0062] In this example, the center of the disk domain is set to o=-2, and the radius is r=2. By solving LMI(11), matrices P and Y are obtained, and then the observer gain L is calculated.

[0063]

[0064]

[0065]

[0066] A battery management system includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the above-described method.

[0067] Specifically, the fault diagnosis method for lithium-ion battery sensors of the present invention includes the following steps: modeling and parameter identification of the equivalent circuit model; fault augmentation modeling; solving the Ω pseudo-inverse matrix; disk pole domain LMI gain design module and example verification module.

[0068] Step 1: Modeling and parameter identification of the equivalent circuit model

[0069] This invention employs a second-order RC model, which balances parameter identifiability with computational complexity and result accuracy; therefore, it is widely used for battery state and fault estimation. The equivalent circuit equation of this second-order RC model is as follows:

[0070]

[0071] in Open circuit voltage, For the terminal voltage, For ohmic internal resistance, and Polarization voltage, , For polarization internal resistance, , Polarizing capacitor; It represents the current; it is positive when discharging and negative when charging.

[0072] In the parameter identification process, piecewise linearization of the open-circuit voltage is adopted to facilitate real-time calculation. and battery remaining charge state Relationship Based on this, the terminal voltage It can be rewritten as

[0073]

[0074] To verify the accuracy of the parameter identification, we built a battery charge-discharge test platform, as shown in the figure below. We then used an industry-standard Hybrid Pulse Power Characterization (HPPC) experiment to identify the five variables required for subsequent testing. .

[0075] By discretizing the terminal voltage expression, we can obtain the variables that need to be identified from the following formula.

[0076] in , Open circuit voltage, For ohmic internal resistance, , For polarization internal resistance, Represents current.

[0077] The error rate between the estimated voltage and the actual voltage obtained by this method can be controlled within a certain range. Within this range, it is evident that the identified model can accurately estimate the internal dynamic changes of the battery during charging and discharging.

[0078] Step 2: Fault Augmentation Modeling

[0079] Voltage and current sensor faults are augmented into the state as additive faults, forming an augmented model with a singular structure.

[0080]

[0081] It is a state vector. It is the output vector of the measurement. This is the system's input vector. Further augmentation is performed using fs(t) as an additional state variable. The system then needs to be rewritten as follows:

[0082]

[0083] in , , This augmented model preserves the original system dynamics while also improving the matrix... The singular structure reveals the characteristics of the fault, which facilitates the subsequent design of an observer to simultaneously estimate the fault and the state.

[0084] Define a matrix , can find the matrix and Make

[0085]

[0086] Its general solution can be expressed as

[0087]

[0088] For the augmented state-space equations described above, the following augmented observer is designed.

[0089]

[0090] The result of the derivation of the error system is as follows

[0091]

[0092] Step 3: Disk Pole Domain LMI Design

[0093] Given disk region If there exists a symmetric matrix The following inequalities must be satisfied:

[0094]

[0095] Then the system matrix All eigenvalues ​​are located at Inside.

[0096] Substitute the system matrix and use Linearization yields the equivalence criterion.

[0097]

[0098] Solve , Later generations This ensures that the error poles fall within the preset disk region, thus guaranteeing the convergence of the fault.

[0099] Step 4: Instance Verification

[0100] To verify the rationality of the proposed model, this patent conducted a UDDS test on a 2.8Ah lithium-ion battery cell at 25 degrees Celsius. This test aims to better simulate the real operating conditions of lithium-ion batteries in electric vehicles. The changes in current and voltage, as well as the estimation effect of the battery's remaining state of charge (SOC), were observed by applying the following segmented faults. The fault applied to the voltage sensor is...

[0101]

[0102] The fault applied to the current sensor is

[0103]

[0104] Figure 9 and Figure 10 The comparison results with the traditional extended dimension system PD observer show that the problem of insufficient estimation of low-frequency fault signals by this observer has been solved.

[0105] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A fault diagnosis method for a lithium battery voltage and current sensor, characterized in that, Includes the following steps: Step 1: Construct an equivalent circuit model of a lithium-ion battery and identify its parameters; Step 2: Treat the faults of the voltage sensor and current sensor as augmented states, and construct an augmented system model with a singular structure; Step 3: Based on the augmented system model described in Step 2, construct an augmented state observer; The gain of the augmented state observer is designed based on the linear matrix inequality method so that the poles of its error dynamic system are located within a preset disk domain. Step 4: Run the augmented state observer described in Step 3 to simultaneously output a first estimate of the battery system state and a second estimate of the voltage and current sensor states; Based on the second estimate, the fault diagnosis of the voltage and current sensor is realized.

2. The fault diagnosis method for the lithium battery voltage and current sensor according to claim 1, characterized in that, Step 4, which involves diagnosing faults in the voltage and current sensors based on the second estimated value, includes: detecting faults based on whether the second estimated value is zero; and / or identifying and estimating the fault type based on the magnitude or waveform of the second estimated value.

3. The fault diagnosis method for the lithium battery voltage and current sensor according to claim 1, characterized in that, The equivalent circuit model described in step 1 is a second-order RC model, and its state equation is: ； in Open circuit voltage, For the terminal voltage, For ohmic internal resistance, and Polarization voltage, , For polarization internal resistance, , Polarizing capacitor; It represents the current; it is positive when discharging and negative when charging.

4. The fault diagnosis method for the lithium battery voltage and current sensor according to claim 3, characterized in that, In step 1, the five variables R0, R1, R2, and R3 of the equivalent circuit model parameters are tested using a hybrid pulse power characteristic (HPPC) experiment. p1 C p1 R p2 C p2 Identify and determine the open circuit voltage. The relationship between the battery state of charge (SOC) and the state of charge (SOC) is processed by piecewise linearization.

5. The fault diagnosis method for the lithium battery voltage and current sensor according to claim 3, characterized in that, In step 2, the augmented model with a singular structure satisfies the following condition: ； It is a state vector. It is the output vector of the measurement. It is the system's input vector. It's a sensor malfunction.

6. The fault diagnosis method for the lithium battery voltage and current sensor according to claim 3, characterized in that, The construction of the augmented state observer in step 3 specifically involves constructing an observer of the following form: ； in = , This is the observer gain matrix to be designed.

7. The fault diagnosis method for the lithium battery voltage and current sensor according to claim 6, characterized in that, In step 3, the preset disk region If there exists a symmetric matrix The following inequalities must be satisfied: ； Then the system matrix All eigenvalues ​​are located at Inside; Substitute the system matrix and use Linearization yields the equivalence criterion. ； Solve , Later generations This ensures that the error poles fall within the preset disk region, thus guaranteeing the convergence of the fault.

8. The fault diagnosis method for the lithium battery voltage and current sensor according to any one of claims 1 to 7, characterized in that, The sensor faults include at least one of the following: bias fault, periodic disturbance fault, random noise fault, low-frequency fluctuation fault, and intermittent loss fault.

9. A battery management system, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method as described in any one of claims 1 to 7.

Images