Electric vehicle battery system high-voltage accessory fault diagnosis device and method

Abstract

The invention discloses an electric vehicle battery system high-voltage accessory fault diagnosis device and method. A fault diagnosis system and a high-voltage load are connected in parallel to a high-voltage bus; the fault diagnosis system comprises a signal generation module and a data acquisition module, and a response current signal is obtained through the data acquisition module; decomposing the response current signal into a plurality of time sequence prototype features, calculating the prototype matching degree of each prototype feature, combining the prototype matching degree into a matching degree vector as a fault symptom fuzzy vector, obtaining a fuzzy diagnosis matrix, and further establishing a relationship between the fault symptom fuzzy vector and a fault mode fuzzy vector; and selecting the mode corresponding to the value with the maximum membership degree from the fault mode fuzzy vector to obtain the high-voltage accessory fault of the battery system of the electric vehicle. According to the invention, the fault diagnosis of the high-voltage accessory key part of the electric vehicle battery system is realized, the safety and reliability of the electric vehicle battery system can be improved, and the safety of the vehicle and the vehicle-mounted personnel is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric vehicles, and in particular relates to a device and method for diagnosing faults of high-voltage accessories of a battery system of an electric vehicle. Background Art

[0002] The high-voltage accessories of the electric vehicle battery system are mainly composed of battery heating management devices, battery cooling management devices and high-low voltage conversion devices DC-DC. The battery heating management device and cooling management device provide a suitable temperature range and a good operating environment for the battery system, and the high-low voltage conversion device DC-DC provides power for low-voltage loads such as the vehicle water pump, fan and controller. When the high-voltage accessories fail, it will directly affect the performance of the battery system, and even cause major safety accidents such as internal battery short circuit and fire, which seriously affects the safe operation of electric heavy trucks.

[0003] In actual engineering applications, in order to avoid the accumulation and expansion of faults, it is necessary to promptly discover and locate the faulty components in the high-voltage accessories of the battery system, and take immediate maintenance measures to improve the safety of the battery system and ensure the normal operation of the vehicle. Therefore, it is necessary to carry out fault diagnosis on multiple components of the system's high-voltage accessories at the same time; at the same time, considering the diversity of failure modes of each component of the high-voltage accessories, the fault diagnosis method should have good robustness to multiple failure modes of specific components; in addition, due to the compact structure of the battery system, the high protection level, and the difficulty of disassembly and modification, the fault diagnosis method should fully consider the convenience of the deployment of the diagnostic system.

[0004] At present, the research on high-voltage load fault diagnosis of electric vehicles mainly focuses on the fault diagnosis of a single component. The component-level fault diagnosis method requires a large number of measurement points and cannot provide system-level fault information, which makes it difficult to meet the core needs of high-voltage accessory fault diagnosis of battery systems. At the same time, the existing component-level fault diagnosis methods are mostly limited to specific fault modes, and the robustness of the fault diagnosis methods is greatly limited. Summary of the invention

[0005] Purpose of the invention: In order to overcome the deficiencies in the prior art, the present invention provides a method for diagnosing faults of high-voltage accessories in the battery system of an electric vehicle. This method only requires the deployment of a single measuring point on the high-voltage busbar of the accessory system to achieve fault diagnosis of multiple parallel components in the accessory system. At the same time, based on the cognitive prototype theory in cognitive psychology, this chapter proposes a mathematical modeling method for cognitive prototype formation and cognitive prototype matching, and applies it to response signal analysis. The present invention combines the impulse response method with the cognitive prototype theory to achieve system-level fault diagnosis with good diagnostic robustness and deployment convenience, thereby ensuring the timeliness of fault handling of high-voltage accessories in the battery system and improving the safety and operational efficiency of electric heavy trucks.

[0006] Technical solution: To achieve the above purpose, the technical solution adopted by the present invention is:

[0007] A method for diagnosing a fault in a high-voltage accessory of an electric vehicle battery system comprises the following steps:

[0008] Step 1: connect the fault diagnosis system and the high-voltage load in parallel to the high-voltage bus. The fault diagnosis system includes a signal generation module and a data acquisition module. When the electric vehicle is not in operation, the signal generation module controls the relay to generate a standard rising edge impulse in the high-voltage bus and inject it into the high-voltage bus. After that, the load in the high-voltage accessory system responds to the impulse to form a response current in the bus. The response current signal is obtained through the data acquisition module.

[0009] Step 2, decompose the response current signal into multiple time-series prototype features, and calculate the prototype matching degree of each prototype feature, wherein the multiple time-series prototype features are prototype feature 1, prototype feature 2, prototype feature 3, prototype feature 4, and prototype feature 5, wherein prototype feature 1 is the high pulse of the first pulse signal P1, prototype feature 2 is the neutrality of the first stable segment signal S1, prototype feature 3 is the wide pulse of the second pulse signal P2, prototype feature 4 is the difference between the lower envelope of the second stable segment and the low value of the first stable segment, and prototype feature 5 is the neutrality of the overall RMS value. A prototype / faulty component correlation matrix is ​​established based on the prototype features and the faulty component situation.

[0010] Step 3: Combine the prototype matching degrees of each prototype feature into a matching degree vector and use it as the fault symptom fuzzy vector. A fuzzy diagnosis matrix is ​​obtained based on the fault symptom fuzzy vector and the prototype / fault component correlation matrix.

[0011] Step 4: Establish the relationship between the fault symptom fuzzy vector and the fault mode fuzzy vector according to the fuzzy diagnosis matrix. Select the mode corresponding to the value with the largest membership degree from the fault mode fuzzy vector to obtain the high-voltage accessory fault of the electric vehicle battery system.

[0012] Preferably: the first pulse signal P1 high pulse is identified and extracted from the response current signal by the characteristic-frequency theory, and the first pulse signal P1 high pulse prototype matching degree is calculated by the matching degree calculation method based on the similarity law, and the P1 high pulse boundary line Th is P1 As the dividing line λ, the calculated P1 high pulse prototype matching degree is recorded as θ P1 .

[0013] Preferably: the first stable segment signal S1 tends to the center by identifying and extracting the upper limit and lower limit of S1 tending to the center from the response current signal through the tendency to the center theory, and calculating the S1 tendency to the center prototype matching degree by using the matching degree calculation method based on the proximity law, and setting the S1 tendency to the center prototype upper boundary Ceil S1 As the upper boundary line λCeil , the lower boundary of S1 centripetal prototype Floor S1 As the lower boundary line λ Floor Considering that S1 is a sequence feature, the calculated S1 centripetal prototype matching degree is recorded as θ S1 .

[0014] Preferably: the second pulse signal P2 wide pulse is identified and extracted from the response current signal by the characteristic-frequency theory, and the P2 wide pulse prototype matching degree is calculated by the matching degree calculation method based on the similarity law, and the P2 wide pulse boundary line Th is P2 As the dividing line λ, the calculated P2 wide pulse prototype matching degree is recorded as θ P2 .

[0015] Preferably: the low difference dividing line between the lower envelope of the second stable segment and the low difference of the first stable segment is identified and extracted from the response current signal through the characteristic-frequency theory, and the matching degree calculation method based on the similarity law is used to calculate the matching degree between the lower envelope of the second stable segment and the low difference prototype of the first stable segment, and the Δi low difference dividing line Th Δi As the dividing line λ, the calculated Δi low difference prototype matching degree is recorded as θ Δi .

[0016] Preferably: the centrifugal tendency of the overall RMS value is identified and extracted from the response current signal by the centrifugal tendency theory, and the centrifugal tendency prototype matching degree of the overall RMS value is calculated by the matching degree calculation method based on the proximity law, and the upper boundary Ceil of the overall RMS centrifugal tendency prototype is calculated. RMS As the upper boundary line λ Ceil , the overall RMS tends to the lower boundary of the prototype Floor RMS As the lower boundary line λ Floor Considering that the overall RMS value is a single-value feature, the prototype matching degree is calculated, and the calculated prototype matching degree of the overall RMS value is recorded as θ RMS .

[0017] Preferably, a cosine distance operator is used as the fuzzy logic operator of the relationship between the fault symptom fuzzy vector and the fault mode fuzzy vector.

[0018] Another object of the present invention is to provide a high-voltage accessory fault diagnosis device for an electric vehicle battery system, comprising a fault diagnosis system, a prototype matching unit, a fuzzy diagnosis matrix unit, and a fuzzy diagnosis model unit, wherein:

[0019] The fault diagnosis system and the high-voltage load are connected in parallel to the high-voltage bus. The fault diagnosis system includes a signal generation module and a data acquisition module. When the electric vehicle is not in operation, the signal generation module controls the relay to generate a standard rising edge impulse in the high-voltage bus and inject it into the high-voltage bus. After that, the load in the high-voltage accessory system responds to the impulse to form a response current in the bus. The response current signal is obtained through the data acquisition module.

[0020] The prototype matching unit decomposes the response current signal into multiple time-series prototype features and calculates the prototype matching degree of each prototype feature. The multiple time-series prototype features are prototype feature one, prototype feature two, prototype feature three, prototype feature four, and prototype feature five, respectively. Among them, prototype feature one is the high pulse of the first pulse signal P1, prototype feature two is the neutralization of the first stable segment signal S1, prototype feature three is the wide pulse of the second pulse signal P2, prototype feature four is the difference between the lower envelope of the second stable segment and the low value of the first stable segment, and prototype feature five is the neutralization of the overall RMS value.

[0021] The fuzzy diagnosis matrix unit is used to combine the prototype matching degrees of each prototype feature into a matching degree vector as a fault symptom fuzzy vector. A fuzzy diagnosis matrix is ​​obtained according to the fault symptom fuzzy vector and the prototype / fault component correlation matrix.

[0022] The fuzzy diagnosis model unit establishes the relationship between the fault symptom fuzzy vector and the fault mode fuzzy vector according to the fuzzy diagnosis matrix. The mode corresponding to the value with the largest membership degree is selected from the fault mode fuzzy vector to obtain the high voltage accessory fault of the electric vehicle battery system.

[0023] Preferably: including switch 1 K1, switch 2 K2, switch 3 K3, switch 4 K4, switch 5 K5, switch 6 K6, wherein switch 1 K1 is arranged on the positive electrode of the battery pack system and the positive electrode connection line of the battery box assembly auxiliary system, the negative electrode of the battery pack system, the manual quick disconnector, switch 2 K2, and the negative electrode connection line of the battery box assembly auxiliary system, one end of switch 3 K3 is connected to the connection line between the positive electrode of the battery pack system and switch 1 K1, and the other end is connected to the fault diagnosis system. One end of switch 5 K5 is connected to the connection line between the positive electrode of the battery box assembly auxiliary system and switch 1 K1, and the other end is connected to the fault diagnosis system. One end of switch 4 K4 is connected to the connection line between the negative electrode of the battery pack system and switch 2 K2, and the other end is connected to the fault diagnosis system. One end of switch 6 K6 is connected to the connection line between the negative electrode of the battery box assembly auxiliary system and switch 2 K2, and the other end is connected to the fault diagnosis system.

[0024] Preferably: when the electric vehicle is working normally, switches K1 and K2 are closed, K3 / K4 / K5 / K6 are disconnected, and the fault diagnosis system is in an offline state. When the electric vehicle is not working, switches K1 and K2 are disconnected, K3 / K4 / K5 / K6 are closed, the fault diagnosis system is connected to the high-voltage bus, and offline fault diagnosis is started.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] (1) A method for diagnosing high-voltage accessory faults in an electric vehicle battery system is proposed. This method can achieve system-level fault diagnosis by deploying only a single measurement point on the high-voltage busbar. The impulse signal is directly controlled and output by the power battery system, which is easy to deploy.

[0027] (2) The timing characteristics of the response signal of the high-voltage accessory system under standard impulse were analyzed, and the corresponding relationship between each timing characteristic and the faulty component of the high-voltage accessory system was determined. Based on this, the relationship matrix between the prototype characteristics and the faulty components was obtained.

[0028] (3) The mathematical modeling methods of "central tendency theory" and "feature-frequency theory" in cognitive psychology were proposed, and based on the established mathematical models, the prototypes of various time series features in the impulse response signal were defined.

[0029] (4) A mathematical modeling method of prototype matching theory in cognitive psychology was proposed. Based on the established mathematical model, the matching degree between the time series features and the corresponding prototypes was calculated, and the quantitative characterization of the time series features was achieved. Combined with the feature prototype definition, the robustness of fault diagnosis was greatly improved.

[0030] In summary, the present invention not only realizes the fault diagnosis of key components of high-voltage accessories of the electric vehicle battery system, but also can improve the safety and reliability of the electric vehicle battery system, and ensure the safety of the vehicle and the people on board. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Get a schematic for the impulse response signal.

[0032] Figure 2 Identify schematic diagram for response signal analysis.

[0033] Figure 3 This is a schematic diagram of the deployment method of the fault diagnosis system.

[0034] Figure 4 Schematic diagram of the timing structure of the current response signal.

[0035] Figure 5 Schematic diagram for defining the prototype of sequence characteristics based on the theory of central tendency.

[0036] Figure 6 Schematic diagram of the steps for determining prototype parameters based on characteristic-frequency theory.

[0037] Figure 7 Schematic diagram for determining the prototype characteristic parameters based on the characteristic-frequency theory.

[0038] Figure 8 Schematic diagram of prototype matching calculation based on the proximity law.

[0039] Fig. 9 Schematic diagram of prototype matching calculation based on similarity law.

[0040] Fig.10 This is the circuit diagram of the experimental bench.

[0041] Fig.11 is the matching degree of the test sample set.

[0042] Fig.12 Schematic diagram of the fuzzy vector of the failure mode of the test sample set. DETAILED DESCRIPTION

[0043] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these examples are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0044] A method for diagnosing faults of high-voltage accessories of an electric vehicle battery system mainly includes two parts: impulse response signal acquisition and response signal analysis and identification. Among them, impulse response signal acquisition mainly depends on the hardware system, including standard impulse signal generation and response signal acquisition. Its basic structure is as follows: Figure 1 As shown in the figure; Based on the cognitive prototype theory, the response signal analysis and identification includes the temporal feature prototype definition and temporal feature quantitative representation, the calculation method of the core prototype index and the fault diagnosis process are shown in the figure. Figure 2 shown.

[0045] When the impulse response signal is acquired, the standard impulse signal battery system provides it. The standard impulse signal directly acts on the bus of the high-voltage accessories. At the same time, the bus current is collected by the response signal acquisition system. The core components of the high-voltage accessories are the battery heating film, the air-conditioning bus capacitor and the DC-DC, among which the DC-DC output load is mainly various types of low-voltage motors. This method does not require the deployment of nodes in each load loop, but only requires the deployment of a single measuring point on the high-voltage bus, so it has good deployment convenience.

[0046] After obtaining the response signal, based on the cognitive prototype theory, the five time series features of the response signal, namely "first pulse", "first steady state", "second pulse", "steady state difference" and "overall effective value" are defined as feature prototypes, and then the prototype formation mathematical model established by the present invention is used to define its prototype; on this basis, the prototype matching mathematical model proposed by the present invention is used to calculate the prototype matching degree of each time series feature in the test sample, and finally merge them into a feature vector, and realize the fault diagnosis of the high-voltage accessories of the battery box assembly based on fuzzy logic reasoning. The specific steps are as follows:

[0047] 1. Impulse response signal acquisition and timing analysis

[0048] (1)Basic composition of high-voltage accessories of battery system

[0049] The high-voltage accessories of the battery system mainly include battery heating management devices, battery cooling management devices and high-low voltage conversion devices, etc.; among them, the battery heating management device mainly provides heating function for the battery, the battery cooling management device mainly provides cooling function for the battery, and the high-low voltage conversion device mainly provides low-voltage power supply for the low-voltage load part including cooling water pump, cooling fan and control system.

[0050] (2) Impulse signal generation and response signal acquisition

[0051] The battery system high voltage accessory fault diagnosis device is deployed inside the battery box assembly and connected in parallel with the high voltage load to the high voltage bus, such as Figure 3 When the electric vehicle is working normally, switches K1 and K2 are closed, K3 / K4 / K5 / K6 are disconnected, and the fault diagnosis system is in an offline state; when the electric vehicle is not working, switches K1 and K2 are disconnected, K3 / K4 / K5 / K6 are closed, the fault diagnosis system is connected to the high-voltage bus, and offline fault diagnosis is started.

[0052] The fault diagnosis system consists of a signal generation module and a data acquisition module. When the electric vehicle is not in operation, the signal generation module controls the relay to generate a standard rising edge impulse on the bus and inject it into the high-voltage bus. After that, the load in the high-voltage accessory system responds to the impulse and forms a response current in the bus. The response current signal is obtained by the data acquisition system, and the high-voltage accessory system fault diagnosis is realized by analyzing the response signal.

[0053] In order to make the hardware structure of impulse signal acquisition simple enough, minimize the impact of fault diagnosis on the system, and shorten the fault diagnosis time, the standard impulse signal consists of a rising edge pulse and lasts for 3 seconds, where the voltage is 0 from 0 to 1s, the battery pack output voltage U from 1 to 3s, and the voltage is 0 after 3 seconds. Among them, the effective signal stages are stage 1 and stage 2. In order to ensure that the detailed components of the response signal are captured, the data acquisition frequency should be above 10kHz.

[0054] (3) Analysis of the timing structure of impulse response signals

[0055] Due to the differences in the characteristics of the load components of the high-voltage accessories of the battery system and the starting current characteristics of the low-voltage load motor connected to the DC-DC output terminal, the load current components contained in the response current signal at different times are different. The present invention analyzes the load current components contained in each time interval according to the time sequence. Figure 4 According to the timing of the response signal, the five timing characteristics of the response signal are defined as "the first pulse P1", "the first steady state S1", "the second pulse P2", "the steady-state difference Δi is the difference between the currents of the S2 and S1 segments", and "the overall effective value is the effective value of the entire current signal".

[0056] 2. Research on mathematical modeling methods of cognitive prototype formation theory

[0057] In cognitive psychology, one of the classic theories of visual pattern recognition is the cognitive prototype theory. This theory holds that for specific things, a certain abstract pattern is stored in long-term memory as a prototype. When performing visual pattern recognition, it can be checked against the prototype. If similarities occur, the pattern is recognized. Prototype refers to the intrinsic representation of a type of pattern, that is, the generalized representation of all samples of a category or category, which represents the form abstracted from the basic components of a certain type of pattern. The prototype mentioned here has certain robust characteristics. The prototype is not only an abstract object of a group of objective things, but also a typical or "optimal" representation of the pattern. Considering the robustness requirements of the high-voltage accessory system-level fault diagnosis of the battery box assembly, the present invention mainly studies the prototype theory and prototype matching theory in detail, and proposes a mathematical modeling method based on its basic principles. Under the framework of the prototype matching theory, the element benchmark is the "prototype" abstracted from each feature. The establishment of a prototype is the basis of the prototype matching theory. For the formation of prototypes, the current mainstream theoretical models are the "central tendency theory" and the "feature-frequency theory".

[0058] (1) Central-tendency theory and its mathematical modeling

[0059] The theory of tendency to the center holds that the prototype represents the mean of a group of samples, that is, the prototype can be mathematically represented as a hypothetical point in a multidimensional space where the average values ​​of various features converge. Therefore, the prototype is an abstract object stored in the memory, which represents the tendency of a category to move toward the center and has a certain robustness. Establishing a feature prototype based on the theory of tendency to the center is suitable for establishing a prototype based on a set of feature samples extracted under a single mode. The prototype is used to determine whether the corresponding features of the test sample belong to the mode. Taking into account that features include two types: single-value features and sequence features, when using the theory of tendency to the center to establish a prototype, the present invention divides it into two cases and performs mathematical modeling respectively.

[0060] If the extracted feature is a single-value feature, assume that the feature extracted from the i-th sample is F i =α i , and there are n samples in the sample set, then the sample set of features is composed of:

[0061] F=[F 1 F 2 … F n ] T =[α 1 α 2 … α n ] T

[0062] According to the principle of the central tendency model, the mean of the sample set is:

[0063]

[0064] In order to ensure the robustness of the prototype, the statistical standard deviation of the sample set is:

[0065]

[0066] Based on the triple standard deviation principle in mathematical statistics, the prototype distribution of single-valued features is:

[0067] P CT =α j ∈[λ Floor ,λ Ceil ]

[0068] λ Floor =μ F -3σ F

[0069] λ Ceil =μ F +3σ F

[0070] If the extracted feature is a sequence feature, the steps for obtaining the prototype parameters are as follows: Figure 5 shown.

[0071] Assume that the length of the sequence feature extracted from each sample used to define the prototype is m, and the sequence feature of the i-th sample is expressed as:

[0072] F i =[β i1 β i2 … β ij … β im ]

[0073] And there are n samples in the sample set, and the sample set is composed of:

[0074] F=[F 1 F 2 … F n ] T

[0075] For each sample F i Find their maximum values ​​β imax and the minimum value β imin , then the maximum value sequence and minimum value sequence of the sample set are:

[0076] F max =[β 1max β 2max … β i(max) … β n(max) ] T

[0077] F min =[β 1min β 2min … β i(min) … β n(min) ] T

[0078] Calculate the mean and standard deviation of the maximum and minimum value sequences respectively:

[0079]

[0080]

[0081] Based on the triple standard deviation principle in mathematical statistics, the prototype distribution of single-valued features is:

[0082] P CT =β ij ∈[λ Floor ,λ Ceil ]

[0083] λ Floor =μ Fmin -3σ Fmin

[0084] λ Ceil =μ Fmax +3σ Fmax

[0085] (2) Attribute-frequency theory and its mathematical modeling

[0086] The feature-frequency theory believes that the prototype represents the majority or most common feature combination in a certain mode. In this model, the meaning of the prototype is a set of "best examples" with a specific pattern, which is a pattern that brings together the most frequently occurring features in a series of samples. For a specific feature, the feature-frequency theory is mainly used to determine the dividing line between any two modes. Based on the dividing line, the specific feature is used to divide the samples of the two modes so that the number of correctly classified samples in each mode is maximized, which conforms to the "highest frequency" and "best example" principles in the feature-frequency theory. Taking a one-dimensional single-valued feature as an example, the steps to obtain the prototype parameters are as follows: Figure 6 shown.

[0087] The method of building a prototype based on the feature-frequency theory is as follows:

[0088] Assume that there are two modes, Mode A and Mode B, where Mode A contains n A samples, and the i-th sample is F A(i) =α i ; Pattern B contains n B samples, and the jth sample is F B(j) =β j , then the one-dimensional single-valued feature F of mode A and mode B A and F B The sample set consists of:

[0089]

[0090] Assume that the sample set F A The mean value is μ FA , sample set F B The mean value is μ FB ,and Now we need to determine the dividing line λ so that α i ≥λ and β j The number of elements ≤ λ is as large as possible.

[0091] Let F A The minimum value is α min , F B The maximum value is β max , if α min ≥β max, then according to the characteristic-frequency theory, the dividing line λ is:

[0092] λ=(α min +β max ) / 2

[0093] Under this condition, the schematic diagram of obtaining the key parameters of the feature prototype is as follows Figure 7 As shown in (a), the prototype dividing line is the mean of the minimum value of the high-value feature and the maximum value of the low-value feature.

[0094] If α min <β max , assuming F A Medium less than β max With F B Greater than α min The union of the elements of is:

[0095] F′=[γ 1 γ 2 …γ n ]

[0096] Assume that λ=γ k When α is satisfied i ≥γ k ,i=1,2,…,n A With β j ≤γ k ,j=1,2,…,n B The sum of the number of elements is the largest, then the dividing line determined based on the characteristic-frequency theory is:

[0097] λ=γ k

[0098] Under this condition, the schematic diagram of obtaining the key parameters of the characteristic prototype is as follows Figure 7 (b) as shown.

[0099] 3. Impulse response signal prototype feature decomposition

[0100] From the analysis of the timing structure of the current response signal, it can be seen that the signals in each time period of the current response signal represent the status of each component in the high-voltage accessories of the battery system. In order to realize the fault diagnosis of the high-voltage accessory system, that is, to locate the fault to the component level, and at the same time, to have a certain robustness to the various fault modes of each component, the present invention decomposes the current response signal into multiple timing prototype features based on the prototype theory in cognitive psychology. That is:

[0101] (1) Prototype feature 1: Pulse height of the first pulse signal (P1)

[0102] The P1 pulse height represents the component performance status of the air conditioner DC bus capacitor. When the bus capacitor is not faulty, the P1 pulse height is relatively high. When the P1 pulse height drops below a certain dividing line, it means that the air conditioner DC bus capacitor is faulty, affecting the normal use of the system.

[0103] (2) Prototype feature 2: Current distribution of the first stable segment signal (S1)

[0104] The current distribution of S1 characterizes the performance status of the battery heating film. When the battery heating film works normally, the current distribution of S1 approaches a certain value and fluctuates slightly around the certain value, showing a "central tendency" characteristic; when the battery heating film has a local short circuit or some branches are open, the central tendency of the S1 current will be destroyed and deviate from the preset current range.

[0105] (3) Prototype feature 3: Pulse width of the second pulse signal (P2)

[0106] The P2 pulse width characterizes the performance status of the DC-DC component. When the DC-DC works normally, the P2 width is greater than a certain dividing line; if the DC-DC has various faults, such as coil open circuit, coil virtual connection, MOSFET open circuit, etc., the P2 pulse will be severely narrowed or even missing.

[0107] (4) Prototype feature 4: Difference between the lower envelope of the second stable segment and the first stable segment (Δi value)

[0108] Δi is the current difference between the lower envelope of S2 and S1, and its value represents the stable working state of the load motor at the DC-DC output end; if the load motor cannot enter a stable working state due to a DC-DC fault, the Δi value is lower than a certain dividing line. Therefore, the Δi value mainly depends on whether the DC-DC is faulty.

[0109] (5) Prototype feature 5: overall RMS value distribution

[0110] In the high-voltage accessories of the battery system, the air-conditioning DC bus capacitor is a reactive device, and the active power is mainly consumed by the battery heating film and the DC-DC output load. Therefore, when the battery heating film and the DC-DC work normally, the overall RMS value distribution tends to a certain value and fluctuates slightly around the certain value; when the overall RMS value tends to change neutrally, that is, deviates from the preset certain value and its range, it indicates that the battery heating film is working abnormally, or the DC-DC and its load are working abnormally.

[0111] The abnormal characteristics of the components represented by the five prototype features described above are summarized, and the prototype / faulty component correlation matrix is ​​shown in Table 1, where "1" indicates strong correlation under theoretical conditions and "0" indicates no correlation under theoretical conditions.

[0112] Table 1 Correlation matrix between prototype features and faulty components

[0113]

[0114] 4. Impulse response signal prototype definition

[0115] Based on the above prototype formation theory and mathematical modeling method, the present invention defines the five prototypes in Table 1. According to the basic characteristics of the prototype and the completeness of the prototype definition sample, the method shown in Table 2 is used to define each prototype and obtain its key parameters.

[0116] Table 2 Prototype definition method and key parameters

[0117]

[0118] 5. Prototype matching of impulse response signals

[0119] According to the prototype features in the impulse response signal defined in the present invention and the prototype defined in Table 2, it is necessary to identify the corresponding prototype features from the impulse response signal and extract the key parameters corresponding to the prototype for matching. Calculate the prototype matching degree.

[0120] (1) Research on mathematical modeling methods of cognitive prototype matching theory

[0121] Prototype matching theory is one of the mainstream theories of visual pattern recognition in cognitive psychology. When performing visual pattern recognition, a pattern can be tested against the prototype. If similarities are found, the pattern is recognized; if the degree of mismatch is significant, a search is then conducted for a prototype that better meets the target.

[0122] Since the essence of the prototype matching process is to test the similarity between the target and the prototype, and during the testing process, human visual cognition is affected by the Gestalt law, the calculation of matching degree also takes the Gestalt law as one of the basic theories.

[0123] a) Matching degree calculation based on proximity law

[0124] When the proximity law is used for prototype matching, the basic principle is: when the target features are close enough to the features represented by the prototype, the target is considered to belong to the same mode as the prototype. Since it is necessary to determine whether the target features are close enough to the features represented by the prototype, the prototype matching degree calculation based on the proximity law is generally coordinated with the prototype determined by the central tendency theory.

[0125] The prototype determined by the central tendency theory is a specific interval. Considering that in the prototype matching process, it is necessary to make a preliminary judgment on whether it meets the prototype before further matching calculation, the calculation strategy is mainly based on interval judgment, and the Sigmoid function is used to fine-tune the matching degree. Corresponding to the process of determining the prototype based on the central tendency theory, the prototype matching degree is calculated for single-value features and sequence features respectively. The basic process is as follows: Figure 8 shown.

[0126] For single-valued features, based on the prototype distribution determined by the central tendency theory, it is assumed that the characterization parameter of the feature sample to be verified is α j If the sample is close enough to the prototype, that is, it is within the prototype distribution range, then the sample is considered to be consistent with the prototype mode, and the matching degree is calculated as follows:

[0127]

[0128] Among them, ε is the matching scale parameter, μ F is the mean of the prototype sample set.

[0129] If the sample characterization parameter α j Outside the prototype interval, the sample is considered not to belong to the mode defined by the corresponding prototype. j <μ F -3σ F , then the matching degree is calculated as:

[0130]

[0131] If α j >μ F +3σ F , then the matching degree is calculated as:

[0132]

[0133] For the sequence characteristics, based on the prototype distribution determined by the central tendency theory, it is assumed that the representation sequence of the feature sample to be verified is F k =[β k1 β k2 … β kj … β km ], and the maximum value of each element is β k(max) , the minimum value is β k(min) , if F k If all elements in satisfy the above conditions, the sample is considered to be close enough to the prototype and the sample conforms to the pattern defined by the prototype. The matching degree is calculated as follows:

[0134]

[0135] If the minimum value β in the sequence samplek(min) Less than the lower limit of the prototype distribution, that is, β k(min) <λ Floor , then it is considered that the sample does not belong to the pattern defined by the prototype, and the matching degree is:

[0136]

[0137] If the maximum value β in the sequence sample k(max) Greater than the upper limit of the prototype distribution, that is, β k(max) >λ Ceil , then the sample is also considered to be inconsistent with the pattern defined by the prototype, and the matching degree is calculated as follows:

[0138]

[0139] b) Matching degree calculation based on similarity law

[0140] Based on the principle of similarity law, object similarity is mainly defined by the consistency of a certain attribute definition parameter, for example, higher pulse, smaller difference, larger RMS value, etc., and the description of "higher", "smaller", and "larger" is mainly realized in the algorithm through threshold or dividing line judgment. Considering that the prototype core parameter defined based on the "feature-frequency" theory in the previous article is the dividing line, the prototype matching degree calculation method based on similarity is suitable for determining whether a sample conforms to the prototype defined by the feature-frequency theory. Its basic process is as follows: Fig. 9 shown.

[0141] Assume that for a single-valued feature sample α j , when α j When α is greater than the dividing line λ, the sample is considered to meet the characteristics defined by the prototype. Under this condition, if α j >λ, that is, the sample meets the pattern characteristics defined by the prototype, and the matching degree is calculated as follows:

[0142]

[0143] On the contrary, when α j <λ, the sample does not conform to the pattern defined by the prototype. In this case, the matching degree is calculated as:

[0144]

[0145] 6. Impulse response signal cognitive prototype matching calculation

[0146] For each sample in the test sample set, calculate the key characteristic parameters, and use the method in step 5 to calculate the prototype matching degree of each test sample. The matching degree to be calculated is as follows:

[0147] (1) “P1 High Pulse” Prototype Matching Calculation

[0148] Since the "P1 high pulse" prototype is obtained by the "characteristic-frequency theory", the matching degree calculation method based on the similarity law is used when calculating the matching degree of the characteristic prototype. P1 As the dividing line λ, the calculated prototype matching degree is recorded as θ P1 .

[0149] (2) Calculation of the prototype matching degree of “S1 central tendency”

[0150] Since the "S1 central tendency" prototype is obtained from the "central tendency theory", the matching degree calculation method based on the proximity law is used when calculating the matching degree of the feature prototype. S1 As the upper boundary line λ Ceil , the lower boundary of S1 centripetal prototype Floor S1 As the lower boundary line λ Floor ; Considering that S1 is a sequence feature, the calculated prototype matching degree is recorded as θ S1 .

[0151] (3) “P2 wide pulse” prototype matching calculation

[0152] Since the "P2 wide pulse" prototype is obtained by the "characteristic-frequency theory", when calculating the matching degree of the characteristic prototype, similar to the "P1 high pulse", the matching degree calculation method based on the similarity law is adopted. P2 As the dividing line λ, the calculated prototype matching degree is recorded as θ P2 .

[0153] (4) Calculation of “Δi low difference” prototype matching degree

[0154] Since the "Δi low difference" prototype is obtained by the "feature-frequency theory", when calculating the matching degree of the feature prototype, similar to the "P1 high pulse", the matching degree calculation method based on the similarity law is adopted. Δi As the dividing line λ, the calculated prototype matching degree is recorded as θ Δi .

[0155] (5) Calculation of prototype matching degree of “overall RMS tendency to neutrality”

[0156] Since the prototype of "overall RMS tendency to centrality" is obtained from the "centration theory", the matching degree calculation method based on the proximity law is used when calculating the matching degree of the feature prototype. RMS As the upper boundary line λ Ceil , the overall RMS tends to the lower boundary of the prototype FloorRMS As the lower boundary line λ Floor ; Considering that the overall RMS value is a single-value feature, the prototype matching degree is calculated, and the calculated prototype matching degree is recorded as θ RMS .

[0157] 7. Construction of fuzzy diagnosis matrix and fault diagnosis output

[0158] After obtaining the prototype matching degrees of all features for the response current test sample, the prototype matching degrees of all features are merged into a matching degree vector and used as the fault symptom fuzzy vector, denoted as x:

[0159] x=[θ P1 θ S1 θ P2 θ Δi θ RMS ]

[0160] Based on the analysis of the equipment and component composition and fault impact of the battery high-voltage accessory system, as well as the analysis of the timing composition of the response signal and the decomposition of the prototype features, the relationship between each prototype feature and the fault mode can be obtained. Starting from the fault diagnosis based on fuzzy logic, the prototype / fault component correlation matrix in step 1 is the fuzzy diagnosis matrix (fuzzy relationship matrix) in fuzzy fault diagnosis.

[0161]

[0162] Assume that the failure mode fuzzy vector of the high-voltage accessory system is

[0163]

[0164] According to the basic principle of fuzzy mathematics, the fault symptom fuzzy vector x and the fault mode fuzzy vector The relationship between them is:

[0165]

[0166] in, is a fuzzy logic operator. Considering that the matching degree itself represents the matching degree between each prototype feature in the response signal and the prototype, the higher the similarity between the fault symptom vector and the corresponding column vector in the fuzzy relationship matrix, the higher the membership degree of the test sample to the mode corresponding to the column vector. Therefore, this paper adopts the cosine distance operator as the fuzzy logic operator, and the calculation method of the operator is:

[0167] y NM = x·r 1 / (||x||·||r 1 ||)

[0168] y ETF = x·r 2 / (||x||·||r 2 ||)

[0169] y DC = x·r 3 / (||x||·||r 3 ||)

[0170] y C = x·r 4 / (||x||·||r 4 ||)

[0171] Considering that in this study, the premise of fault diagnosis is a single fault, the mode corresponding to the value with the largest membership degree is selected from the fault mode fuzzy vector, which is the mode of the test sample, and the high-voltage accessory fault of the electric vehicle battery system is obtained.

[0172] In another embodiment, a high-voltage accessory fault diagnosis device for an electric vehicle battery system is provided, comprising a fault diagnosis system, a prototype matching unit, a fuzzy diagnosis matrix unit, and a fuzzy diagnosis model unit, wherein:

[0173] The fault diagnosis system and the high-voltage load are connected in parallel to the high-voltage bus. The fault diagnosis system includes a signal generation module and a data acquisition module. When the electric vehicle is not in operation, the signal generation module controls the relay to generate a standard rising edge impulse in the high-voltage bus and inject it into the high-voltage bus. After that, the load in the high-voltage accessory system responds to the impulse to form a response current in the bus. The response current signal is obtained through the data acquisition module.

[0174] It includes switch 1 K1, switch 2 K2, switch 3 K3, switch 4 K4, switch 5 K5, and switch 6 K6, wherein switch 1 K1 is set on the positive electrode of the battery pack system and the positive electrode connection line of the battery box assembly auxiliary system, the negative electrode of the battery pack system, the manual quick disconnector, switch 2 K2, and the negative electrode connection line of the battery box assembly auxiliary system, one end of switch 3 K3 is connected to the positive electrode of the battery pack system and the connection line of switch 1 K1, and the other end is connected to the fault diagnosis system. One end of switch 5 K5 is connected to the positive electrode of the battery box assembly auxiliary system and the connection line of switch 1 K1, and the other end is connected to the fault diagnosis system. One end of switch 4 K4 is connected to the negative electrode of the battery pack system and the connection line of switch 2 K2, and the other end is connected to the fault diagnosis system. One end of switch 6 K6 is connected to the negative electrode of the battery box assembly auxiliary system and the connection line of switch 2 K2, and the other end is connected to the fault diagnosis system.

[0175] When the electric vehicle is working normally, switches K1 and K2 are closed, K3 / K4 / K5 / K6 are disconnected, and the fault diagnosis system is in an offline state. When the electric vehicle is not working, switches K1 and K2 are disconnected, K3 / K4 / K5 / K6 are closed, the fault diagnosis system is connected to the high-voltage bus, and offline fault diagnosis is started.

[0176] The prototype matching unit decomposes the response current signal into multiple time-series prototype features and calculates the prototype matching degree of each prototype feature. The multiple time-series prototype features are prototype feature one, prototype feature two, prototype feature three, prototype feature four, and prototype feature five, respectively. Among them, prototype feature one is the high pulse of the first pulse signal P1, prototype feature two is the neutralization of the first stable segment signal S1, prototype feature three is the wide pulse of the second pulse signal P2, prototype feature four is the difference between the lower envelope of the second stable segment and the low value of the first stable segment, and prototype feature five is the neutralization of the overall RMS value.

[0177] The fuzzy diagnosis matrix unit is used to combine the prototype matching degrees of each prototype feature into a matching degree vector as a fault symptom fuzzy vector. A fuzzy diagnosis matrix is ​​obtained according to the fault symptom fuzzy vector and the prototype / fault component correlation matrix.

[0178] The fuzzy diagnosis model unit establishes the relationship between the fault symptom fuzzy vector and the fault mode fuzzy vector according to the fuzzy diagnosis matrix. The mode corresponding to the value with the largest membership degree is selected from the fault mode fuzzy vector to obtain the high voltage accessory fault of the electric vehicle battery system.

[0179] simulation

[0180] In order to verify the fault diagnosis method of high-voltage accessories of the battery box assembly proposed in the present invention, this section designs an experimental bench and performs typical fault injection based on the basic structure of the high-voltage accessories of the battery system. The standard impulse signal is injected into the experimental system, and the response current signal is collected at the same time. A data set is constructed to verify the proposed response signal processing method.

[0181] The key components of the high-voltage accessories of the battery system mainly include the battery heating film, the air-conditioning DC bus capacitor, the DC-DC and its load. The DC-DC output power is mainly consumed by the DC motor load. Therefore, when designing and building the experimental bench, the main loads involved include: battery heating film, capacitor, DC-DC and its corresponding DC motor load; at the same time, a battery module composed of 21700 battery cells is used to simulate the battery system to provide a reference power supply for the standard impulse signal. The circuit diagram of the experimental bench is as follows Fig.10 As shown:

[0182] For the faulty components in Table 3, a total of 20 groups of data were collected under each fault injection mode, namely: 120 groups of battery heating film fault data, 140 groups of DC-DC fault data, and 120 groups of capacitor fault data; in addition, 120 groups of data samples under normal working conditions were collected, and the samples collected under normal working conditions are referred to as NM.

[0183] Data collection starts at t = 0s and ends at t = 3s. The data sampling frequency is f S =32kHz, that is, each response current data sample contains 96,000 data points, and the data sampling accuracy is 18 bits. Using the known sampling resistor value, the collected voltage data is converted into current data.

[0184] In order to test the robustness of the algorithm proposed in this paper, the prototype definition samples and test samples of some faulty components adopt different failure modes to verify the algorithm's ability to diagnose unknown faults. The composition of the prototype definition samples and test samples of the algorithm is shown in Table 4, that is, the prototype definition sample set has a total of 240 samples, and the test sample set has a total of 260 samples.

[0185] Table 3 Key component fault injection methods

[0186]

[0187]

[0188] Table 4 Prototype definition and test sample set composition

[0189]

[0190] Based on the method of the present invention, for the response signal, the key parameters defined by each prototype are: the dividing line of the P1 high pulse (used to characterize the DC bus capacitor fault), the upper and lower limits of the S1 current tendency to neutrality (used to characterize the fault of the battery heating film), the dividing line of the P2 wide pulse (used to characterize the DC-DC fault), the dividing line of the Δi low difference (used to characterize the DC-DC fault), and the upper and lower limits of the overall RMS tendency to neutrality (used to characterize the fault of the battery heating film and the DC bus capacitor). The prototype name obtained from the prototype definition sample and the key parameters of the prototype definition are shown in Table 5. Based on the dividing lines in the table, it can be used to judge the matching degree between the test sample and each prototype.

[0191] Table 5 Key parameters of prototype definition

[0192]

[0193]

[0194] Based on the prototype definition key parameters in Table 5 and the prototype matching calculation method, the matching degree of each prototype feature is calculated for all test samples. When calculating the matching degree, all matching degree scale parameters are ε = 0.05. When calculating the prototype matching degree of each feature of all test samples, when the feature belongs to the prototype range and is close enough to the prototype, the matching degree is higher, and when the feature does not belong to the prototype range and is farther from the prototype, the matching degree is lower.

[0195] Based on the determined prototype characteristic parameters, feature recognition is performed on all test samples, and the matching degree between each test sample and each prototype characteristic is calculated, namely, the matching degree of P1 high pulse, the matching degree of S1 neutralization, the matching degree of P2 wide pulse, the matching degree of Δi low difference, and the matching degree of overall RMS value neutralization. The calculated matching degrees are as follows: Fig.11 As shown in the test results, it can be seen that the defined prototype features can effectively identify different patterns of the test samples.

[0196] Finally, this section uses fuzzy logic reasoning to determine the final fault diagnosis results. The fault symptom vectors of all test sample sets are as follows: Fig.11 Based on the fuzzy diagnosis matrix defined in the present invention, the cosine distance operator is used to calculate the fault mode fuzzy vector of all samples. Since the test sample set contains 260 test samples in total, the fault mode fuzzy vectors of all test samples form a matrix of size 260×4, which is as follows: Fig.12 As shown in the figure, the fault modes of all test samples can be accurately identified based on the membership degree, thus achieving fault diagnosis for all test samples.

[0197] Specific implementation analysis shows that the method proposed in the present invention has the characteristics of high accuracy and strong robustness.

[0198] Aiming at the system-level fault diagnosis requirements of convenient device deployment and robustness of diagnostic methods when diagnosing faults of high-voltage accessories of electric vehicle battery systems, the present invention proposes a fault diagnosis method based on impulse response analysis. The method makes full use of the different effects of faults of various components in the high-voltage accessory system on the steady-state and transient responses under standard impulses. On this basis, mathematical modeling is performed on the cognitive prototype formation mechanism and prototype matching mechanism in cognitive psychology, and the response current signal is processed based on the mathematical models, thereby realizing system-level fault diagnosis of battery high-voltage accessories.

[0199] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for diagnosing faults of high-voltage accessories of an electric vehicle battery system, characterized in that: The following steps are involved: Step 1, connecting a fault diagnosis system and a high-voltage load in parallel to a high-voltage bus; the fault diagnosis system includes a signal generation module and a data acquisition module. When the electric vehicle is not in operation, the signal generation module controls the relay to generate a standard rising edge impulse in the high-voltage bus and inject it into the high-voltage bus. After that, the load in the high-voltage accessory system responds to the impulse to form a response current in the bus, and the response current signal is obtained through the data acquisition module; Step 2, decomposing the response current signal into multiple time-series prototype features, and calculating the prototype matching degree of each prototype feature, the multiple time-series prototype features are prototype feature one, prototype feature two, prototype feature three, prototype feature four, and prototype feature five, wherein prototype feature one is the first pulse signal P1 high pulse, prototype feature two is the first stable segment signal S1 neutralization, prototype feature three is the second pulse signal P2 wide pulse, prototype feature four is the difference between the lower envelope of the second stable segment and the low value of the first stable segment, and prototype feature five is the neutralization of the overall RMS value; establish a prototype / faulty component correlation matrix according to the prototype features and the faulty component situation; Step 3, combining the prototype matching degrees of each prototype feature into a matching degree vector and using it as a fault symptom fuzzy vector; obtaining a fuzzy diagnosis matrix according to the fault symptom fuzzy vector and the prototype / fault component correlation matrix; Step 4, establishing the relationship between the fault symptom fuzzy vector and the fault mode fuzzy vector according to the fuzzy diagnosis matrix; selecting the mode corresponding to the value with the largest membership degree from the fault mode fuzzy vector, and obtaining the high-voltage accessory fault of the electric vehicle battery system.

2. The electric vehicle battery system high voltage accessory fault diagnosis method according to claim 1, characterized in that: The first pulse signal P1 high pulse is identified and extracted from the response current signal by the characteristic-frequency theory, and the first pulse signal P1 high pulse prototype matching degree is calculated by the matching degree calculation method based on the similarity law, and the P1 high pulse boundary line Th is P1 As the dividing line λ, the calculated P1 high pulse prototype matching degree is recorded as θ P1 .

3. The electric vehicle battery system high voltage accessory fault diagnosis method according to claim 2 is characterized in that: The first stable segment signal S1 tends to the center by identifying and extracting the upper limit and the lower limit of S1 tending to the center from the response current signal through the tendency to the center theory, and the S1 tends to the center prototype matching degree is calculated by using the matching degree calculation method based on the proximity law, and the S1 tends to the center prototype upper boundary Ceil S1 As the upper boundary line λ Ceil , the lower boundary of S1 centripetal prototype Floor S1 As the lower boundary line λ Floor ; Considering that S1 is a sequence feature, the calculated S1 centripetal prototype matching degree is recorded as θ S1 .

4. The electric vehicle battery system high voltage accessory fault diagnosis method according to claim 3 is characterized in that: The second pulse signal P2 wide pulse is identified and extracted from the response current signal by the characteristic-frequency theory, and the P2 wide pulse prototype matching degree is calculated by the matching degree calculation method based on the similarity law, and the P2 wide pulse dividing line Th is P2 As the dividing line λ, the calculated P2 wide pulse prototype matching degree is recorded as θ P2 .

5. The method for diagnosing faults of high-voltage accessories of an electric vehicle battery system according to claim 4, characterized in that: The second stable segment lower envelope and the first stable segment low difference are identified and extracted from the response current signal through the characteristic-frequency theory, and the matching degree calculation method based on the similarity law is used to calculate the matching degree between the second stable segment lower envelope and the first stable segment low difference prototype, and the Δi low difference boundary line Th is Δi As the dividing line λ, the calculated Δi low difference prototype matching degree is recorded as θ Δi .

6. The electric vehicle battery system high voltage accessory fault diagnosis method according to claim 5, characterized in that: The centrifugal tendency of the overall RMS value is identified and extracted from the response current signal by the centrifugal tendency theory, and the centrifugal tendency prototype matching degree of the overall RMS value is calculated by the matching degree calculation method based on the proximity law, and the upper boundary Ceil of the overall RMS centrifugal tendency prototype is calculated. RMS As the upper boundary line λ Ceil , the overall RMS tends to the lower boundary of the prototype Floor RMS As the lower boundary line λ Floor ; Considering that the overall RMS value is a single-value feature, the prototype matching degree is calculated, and the calculated prototype matching degree of the overall RMS value is recorded as θ RMS .

7. The electric vehicle battery system high voltage accessory fault diagnosis method according to claim 6, characterized in that: The cosine distance operator is used as the fuzzy logic operator for the relationship between the fault symptom fuzzy vector and the fault mode fuzzy vector.

8. A diagnostic device based on the electric vehicle battery system high voltage accessory fault diagnosis method according to claim 1, characterized in that: It includes a fault diagnosis system, a prototype matching unit, a fuzzy diagnosis matrix unit, and a fuzzy diagnosis model unit, wherein: The fault diagnosis system and the high-voltage load are connected in parallel to the high-voltage bus; the fault diagnosis system includes a signal generation module and a data acquisition module. When the electric vehicle is not in operation, the signal generation module controls the relay to generate a standard rising edge impulse in the high-voltage bus and inject it into the high-voltage bus. After that, the load in the high-voltage accessory system responds to the impulse to form a response current in the bus, and the response current signal is obtained through the data acquisition module; The prototype matching unit decomposes the response current signal into a plurality of time-series prototype features, and calculates the prototype matching degree of each prototype feature, wherein the plurality of time-series prototype features are respectively prototype feature one, prototype feature two, prototype feature three, prototype feature four, and prototype feature five, wherein the prototype feature one is a high pulse of the first pulse signal P1, the prototype feature two is a neutral tendency of the first stable segment signal S1, the prototype feature three is a wide pulse of the second pulse signal P2, the prototype feature four is a difference between the lower envelope of the second stable segment and the low value of the first stable segment, and the prototype feature five is a neutral tendency of the overall RMS value; The fuzzy diagnosis matrix unit is used to combine the prototype matching degrees of each prototype feature into a matching degree vector and use it as a fault symptom fuzzy vector; obtain a fuzzy diagnosis matrix according to the fault symptom fuzzy vector and the prototype / fault component correlation matrix; The fuzzy diagnosis model unit establishes the relationship between the fault symptom fuzzy vector and the fault mode fuzzy vector according to the fuzzy diagnosis matrix; selects the mode corresponding to the value with the largest membership degree from the fault mode fuzzy vector, and obtains the high-voltage accessory fault of the electric vehicle battery system.

9. The diagnostic device according to claim 8, characterized in that: It includes switch one K1, switch two K2, switch three K3, switch four K4, switch five K5 and switch six K6, wherein switch one K1 is arranged on the connection line between the positive pole of the battery pack system and the positive pole of the battery box assembly auxiliary system, and the negative pole of the battery pack system, the manual quick disconnector, switch two K2 and the negative pole of the battery box assembly auxiliary system are connected; one end of switch three K3 is connected to the connection line between the positive pole of the battery pack system and switch one K1, and the other end is connected to the fault diagnosis system; one end of switch five K5 is connected to the connection line between the positive pole of the battery box assembly auxiliary system and switch one K1, and the other end is connected to the fault diagnosis system; one end of switch four K4 is connected to the connection line between the negative pole of the battery pack system and switch two K2, and the other end is connected to the fault diagnosis system; one end of switch six K6 is connected to the connection line between the negative pole of the battery box assembly auxiliary system and switch two K2, and the other end is connected to the fault diagnosis system.

10. The diagnostic device according to claim 9, characterized in that: When the electric vehicle is working normally, switches K1 and K2 are closed, K3 / K4 / K5 / K6 are disconnected, and the fault diagnosis system is in an offline state; when the electric vehicle is not working, switches K1 and K2 are disconnected, K3 / K4 / K5 / K6 are closed, the fault diagnosis system is connected to the high-voltage bus, and offline fault diagnosis is started.

Images