A Multi-Level Reliability Prediction Method for Power Systems Based on Fault Prediction Parameters

Abstract

This invention discloses a multi-level reliability prediction method for power systems based on fault precursor parameters, belonging to the field of power system reliability prediction technology. It addresses the problem that existing reliability modeling methods struggle to effectively assess the reliability of multi-level, multi-task power systems. The invention constructs a dynamic fault tree to characterize the system's functional structure according to signal hierarchy relationships. External excitations are transformed into component stresses through a multi-stress field simulation model of sub-circuits. Combined with a failure physics model of key parameters, output parameters containing performance degradation and functional failure characteristics are determined, completing the transfer from the component layer to the sub-circuit layer. Based on the system's dynamic fault tree, a system-level dynamic simulation model is established to obtain the functional unit output parameters under equal stress, completing the transfer from the sub-circuit layer to the functional unit layer. Based on the reliability of the functional unit output parameters, weighting coefficients are used to complete the parameter transfer from the functional unit to the system task profile layer and the overall system task layer, predicting the final reliability.

Description

Technical Field

[0001] This invention belongs to the field of power system reliability prediction technology, specifically relating to a multi-level reliability prediction method for power systems based on fault precursor parameters. Background Technology

[0002] From the perspective of reliability design and health management, the reliability of power system output parameters is crucial information for evaluating system reliability, and the accuracy of its modeling determines the accuracy of system reliability prediction. During operation, power systems perform various power supply, distribution, and transfer tasks. The weighted result of the reliability under each system task profile is used as a comprehensive reliability evaluation parameter throughout the system's lifecycle. However, complex power systems are characterized by multi-level circuits, multiple tasks, numerous components, signal coupling between units, and complex and variable task timings. Existing system-level reliability prediction methods mostly only consider the calculation of unit-level input and output parameters, resulting in insufficient accuracy and depth of analysis in predicting system reliability. Digital model-based system reliability prediction methods rely excessively on low-level component information as input, and research on constructing comprehensive system reliability characterization models for systems with limited low-level information is insufficient. Furthermore, the dynamic timing changes and multi-stress coupling phenomena of the system under operational tasks place higher demands on modeling accuracy. Summary of the Invention

[0003] This invention aims to address the problem that existing reliability modeling methods struggle to effectively assess the reliability of multi-level, multi-task power systems.

[0004] A multi-level reliability prediction method for power systems based on fault precursor parameters includes the following steps:

[0005] S1. Characterizing the functional structure of the system based on the dynamic fault tree of the power system;

[0006] The power system dynamic fault tree includes a component layer, a sub-circuit layer, a functional unit layer, a system task profile layer, and a system overall task layer. Directed edges represent the state parameter transmission method at each level. Component state parameter A... l The sub-circuit state parameter B is transmitted to the sub-circuit layer through a multi-stress field simulation model. m The functional unit state parameter C is passed to the functional unit layer through the system-level dynamic simulation model. p By weight coefficient W i,p The system task profile state parameter D is passed to the system task profile layer. i Through weight coefficient W sys,i Passed to the system's overall task layer;

[0007] S2. For each sub-circuit, establish a multi-stress field simulation model that transforms external excitation into component stress, and then analyze the component state parameters A.l Sensitivity analysis determines key parameter A k A failure physics model for key parameters is established by combining component stress. The degradation curve and failure time sample obtained by sampling are substituted into the multi-stress field simulation model of the sub-circuit to obtain the sub-circuit output parameters that can represent the performance degradation characteristics and functional failure characteristics; thus realizing the transfer from the component layer to the sub-circuit layer.

[0008] S3. Based on the key parameters and output parameters of the sub-circuit, the parameters affecting the failure of the power supply system, i.e., the system fault precursor parameters, are determined through correlation coefficient analysis. A system-level dynamic simulation model is established based on the dynamic fault tree of the power supply system. The fault precursor parameters are substituted into the system-level dynamic simulation model to obtain the functional unit output parameters under equal stress, realizing the transfer from the sub-circuit layer to the functional unit layer. The functional unit output parameters can be represented as a degradation curve with distribution, i.e., the functional unit output parameter degradation curve.

[0009] S4. Determine the functional unit degradation model based on the functional unit output parameter degradation curve under equal stress, and determine the reliability R of the functional unit output parameter according to the functional unit degradation model and failure threshold. out,p (t), and through the weighting coefficient W i,p To enable the transfer of functional units to the system task profile layer;

[0010] S5, System Task Profile Status Parameters D i Through weight coefficient W sys,i It is passed to the overall system task layer, which ultimately determines the reliability characteristics such as the overall system task reliability.

[0011] Furthermore, the specific process of step S2 includes the following steps:

[0012] S21, Establish X sub The simulation model of the multi-stress field corresponding to each sub-circuit transforms the external excitation into the component stress.

[0013] S22. Based on the multi-stress field simulation models corresponding to each of the constructed sub-circuits, the component state parameters A are analyzed for each sub-circuit. l Sensitivity analysis was performed to determine the key parameter A. k ;

[0014] S23. Determine the failure physical model for key parameters, including the characterization of key parameter A. k Degradation model A over time k (t)=G P (t,Τ l (t),E l (t) and characterizing key parameter A k Failure model F based on the time to occurrence of hard fault state k(t)=G F (t,Τ l (t),E l (t)), G P (·) and G F (·) are functions used to represent the degradation model and the failure model, respectively;

[0015] By random sampling, degradation curves and failure time samples corresponding to each key parameter are generated. The data corresponding to the degradation curves and the data corresponding to open circuits and short circuits after reaching the failure time are substituted into the multi-stress field simulation model to obtain the sub-circuit output parameters. The sub-circuit output parameters can represent the performance degradation characteristics and functional failure characteristics.

[0016] Furthermore, the multi-stress field simulation model includes an electrical simulation model E for the components. l (t)=G E (t,S,T) and thermal simulation model T l (t)=G T (t,E,M), and the electrical and thermal simulation models can transfer parameters to each other;

[0017] The electrical simulation model transforms the external electrical stress S and temperature stress T into the component's electrical stress E based on time t. l (t);

[0018] The thermal simulation model converts the ambient temperature stress M and the electrical stress of all components in the sub-circuit into the thermal stress T on the surface of each component l. l (t).

[0019] Further, step S22 includes:

[0020] During the life cycle, A was respectively... l First and second order Sobol sensitivity analyses were performed to analyze the impact of single parameter changes on the sub-circuit output and the coupling effect of any two parameter changes on the sub-circuit output. Parameters with first and second order sensitivity values ​​both higher than the sensitivity threshold were identified as critical parameters A. k Key parameter A k Including the intrinsic parameters R of the components k Key node parameter V k .

[0021] Furthermore, the aforementioned multi-stress field simulation model: [V n (t),I n [(t)]=G M (t,S,A k (t),F k (t)), where G M (·) is a function used to represent the obtained multi-stress field simulation model, Vn (t), I n (t) represents the voltage and current parameters of the sub-circuit output node n as it degenerates with time t, V n (t), I n (t) can represent performance degradation characteristics and functional failure characteristics.

[0022] Furthermore, step S3 specifically includes the following steps:

[0023] S31. Based on the key parameters and output parameters of the sub-circuit, Pearson correlation coefficient analysis is used to screen out parameters affecting power system failure from the key parameters determined in S22, namely, system fault precursor parameters, system fault precursor parameter A. S Including the intrinsic parameters R of sensitive components S Sensitive key node parameter V S ;

[0024] S32. Based on the topology of the dynamic fault tree of the power system, a dynamic logic gate component is composed of simulation components. The simulation components include: sequential switches, programmable switches, registers, basic logic gates, etc. A system-level dynamic simulation model is established based on the dynamic logic gate component to obtain the interface between the sub-circuit level and the functional unit level, thereby realizing the conversion of the qualitative power system dynamic fault tree model into a quantitative simulation model.

[0025] S33. First, obtain the system fault precursor parameter A based on historical data. S This includes: R-values ​​fitted from historical data under equal stress. S V corresponding to historical test data under equal stress S ; The non-uniform stress historical test data V corresponding to the sensitive key node parameters test As a supplement, it is multiplied by the correction factor μ to obtain V′. S =(1-μ)V S +μV test ;

[0026] Will be by V S Corrected to V′ S The corresponding system fault precursor parameter A S Substituting into the system-level dynamic simulation model, the calculation aims to minimize the error between the simulated value and the actual value of the system-level dynamic simulation model. The value of this factor is changed through an optimization algorithm to correct the correction factor μ. The output parameters of the functional unit under constant stress are obtained through simulation. The output parameters of the functional unit can be represented as a degradation curve with a distribution, i.e., the degradation curve of the output parameters of the functional unit.

[0027] Furthermore, the process of determining the functional unit degradation model based on the functional unit output parameter degradation curve under equal stress, as described in step S4, includes:

[0028] The degradation model of the p-th functional unit is represented as follows:

[0029] D p (t)=d(t;Φ)+ξ (2)

[0030] Among them, D p ξ is the cumulative degradation amount, d(t; Φ) is the degradation trajectory function representing the degradation curve, Φ is the coefficient in the function, and ξ is the uncertainty parameter that follows a normal distribution.

[0031] Furthermore, the process of determining the reliability of a functional unit based on the functional unit degradation model and failure threshold, as described in step S4, includes:

[0032] D p (t) > failure threshold D f Failure occurs at time t, and the failure probability function F of the p-th functional unit at time t is given. p (t) is:

[0033]

[0034] Among them G f Let be a standard normal cumulative distribution function with mean μ and standard deviation σ.

[0035] The reliability R of the output parameter of the p-th functional unit is calculated based on the failure probability function. out,p (t)=1-F p (t).

[0036] Furthermore, the weighting coefficients for the transfer from the functional unit to the system task profile layer in step S4 are as follows:

[0037]

[0038] Among them, W i,p The weighting coefficients of the output parameters of the p-th functional unit in the i-th system task profile; t 第p个功能单元时长 t represents the working time of the p-th functional unit in the i-th system task profile; 第i个系统任务剖面总时长 This represents the total duration corresponding to the i-th system task profile.

[0039] Furthermore, the specific process of step S5 includes the following steps:

[0040] Determine the weight coefficients Among them, W sys,i Let t be the weighting coefficient of the total system tasks over the entire lifecycle for the i-th system task profile; 第i个任务剖面出现次数对应的时长 t represents the total duration of the i-th system task profile calculated based on the number of occurrences. 系统总任务 Indicates the total duration of all system tasks;

[0041] Reliability R based on functional unit output parameters out，p Determine the total mission reliability R throughout its entire lifecycle. sys :

[0042]

[0043] Based on the overall task reliability R sys Make a reliability prediction.

[0044] Compared with the prior art, the present invention has the following advantages:

[0045] This invention employs hierarchical dynamic modeling of system-level circuits, combining system failure precursor parameters with comprehensive reliability characterization parameters based on weights of different system task profiles. This addresses the problem that existing reliability modeling methods struggle to effectively assess the reliability of multi-level, multi-task power systems. Sensitivity analysis and correlation analysis are used to extract various key failure precursor parameters, reducing the amount of information required at the component level and improving the ability to predict power system reliability with limited information. Furthermore, historical experimental data is integrated for system-level reliability calculations, ensuring prediction accuracy. This invention also provides a new approach to combining failure physics models with system-level reliability prediction. Attached Figure Description

[0046] Figure 1 This is a flowchart of a multi-level reliability prediction method for power systems based on fault precursor parameters;

[0047] Figure 2 It is a system layering and parameter transfer flow diagram;

[0048] Figure 3 It is a dynamic fault tree of the power system;

[0049] Figure 4 This is a simulation diagram of the multi-stress field of sub-circuit B4;

[0050] Figure 5 These are the results of Sobol's first-order and second-order sensitivity analyses;

[0051] Figure 6 It is a sub-circuit degradation simulation platform based on iSIGHT;

[0052] Figure 7 It is a data sampling method for degradation and failure models;

[0053] Figure 8 It is a system-level dynamic simulation model;

[0054] Figure 9 These are the degradation curves of fault precursor parameters of resistors A7 and chips A8, and the output parameters of functional unit C2.

[0055] Figure 10 This is the reliability curve of functional unit C2;

[0056] Figure 11 It is the overall system reliability curve, which is a comprehensive evaluation parameter of the system. Detailed Implementation

[0057] To address the problems existing in the prior art, this invention provides a multi-level reliability prediction method for power systems based on fault precursor parameters. This method is based on key parameter degradation and failure models, establishes parameter transfer relationships between five functional levels, and extracts system fault precursor parameters through quantitative analysis, enabling reliability prediction of complex electronic system functions with limited information. Furthermore, it improves the ability to predict power system reliability under various task flows by using comprehensive system reliability characterization parameters. Specific implementation methods are described below.

[0058] Specific implementation method one: Combining Figure 1 This implementation method is described below.

[0059] A multi-level reliability prediction method for power systems based on fault precursor parameters includes the following steps:

[0060] Step S1: Divide the power supply system into five levels according to the signal hierarchy: component layer, sub-circuit layer, functional unit layer, system task profile layer, and overall system task layer (output reliability characteristics); and construct a dynamic fault tree to characterize the system's functional structure, determining the state parameter transmission method for each level; specifically including:

[0061] Step S11: The power system is referred to simply as the system. Based on the directed acyclic graph (DAG) construction method, the system is divided into five levels. Each level is considered a network node. Let the system include X... comp +X sub +X func +X sys +1 node: X comp individual components, X sub Individual circuits, X func Functional units, X sys One system task profile, one overall system task. The system network node state set is Ω = {A} l B m C p D i ,R sys}, where A l (l=1,2,…,X comp ) indicates the component status parameter; B m (m=1,2,…,X sub) represents the state parameters of the sub-circuit; C p (p = 1, 2, ..., X) func ) represents the state parameter of the functional unit; D i (i = 1, 2, ..., X) sys ) represents the system task profile state parameters; R sys It represents the overall system reliability characteristics (reliability, failure rate, etc.); the constructed directed acyclic graph with 5 levels is the hierarchical construction of the dynamic fault tree of the power system;

[0062] Step S12: Utilize the power system dynamic fault tree constructed above to characterize the system's functional structure, with directed edges representing the state parameter transfer methods at each level: A l The stress field simulation model is transmitted to B. m B m The data is transmitted to C through a system-level dynamic simulation model. p C p By weight coefficient W i,p Passed to D i Finally, D i Through weight coefficient W sys,i Passed to R sys .

[0063] Step S2: Establish a multi-stress field simulation model for the sub-circuit, converting external excitation into component stress, and determine key parameters through sensitivity analysis; combine component stress to establish a failure physics model for the key parameters, and substitute the degradation curves and failure time samples obtained by random sampling into the multi-stress field simulation model of the sub-circuit to obtain the output parameters of the sub-circuit containing performance degradation and functional failure characteristics, completing the transfer from the component layer to the sub-circuit layer; the specific process includes the following steps:

[0064] Step S21, Establish X sub Each sub-circuit has its own multi-stress field simulation model, which transforms external excitation into component stress. The multi-stress field simulation model includes the component electrical simulation model E. l (t)=G E (t,S,T) and thermal simulation model T l (t)=G T (t,E,M), and the electrical and thermal simulation models can transfer parameters to each other.

[0065] The electrical simulation model incorporates external electrical stress S and temperature stress T (where T = {T1, ..., T2}). a ,…},T a The temperature of the a-th component is converted into the component's electrical stress E based on time t. l (t);

[0066] The thermal simulation model incorporates the ambient temperature stress M and the electrical stress E of all components in the sub-circuit (where E = {V1, I1, P1, ... V1}). b ,I b ,P b ,},V b I b P b The voltage, current, and power of the b-th component are converted into the thermal stress T on the surface of each component l. l (t);

[0067] Step S22: Based on the multi-stress field simulation models corresponding to each of the constructed sub-circuits, the component state parameters A are analyzed for each sub-circuit. l Sensitivity analysis was performed to determine the key parameter A. k The steps are as follows:

[0068] During the life cycle, A was respectively... l First and second order sensitivity analyses were performed using Sobol to examine the impact of a single parameter variation of ±20% on the sub-circuit output (i.e., the simulation results of the multi-stress field simulation model) and the coupling effect of any two parameters variation of ±20% on the sub-circuit output. Parameters with first and second order sensitivity values ​​both higher than ±1% were identified as critical parameter A. k Key parameter A k Including the intrinsic parameters R of the components k Key node parameter V k , can be represented as A k ={R k V k}

[0069] Step S23: Establish the physical failure model (PoF) for key parameters, including the characterization of key parameter A. k Degradation model A over time k (t)=G P (t,Τ l (t),E l (t) and characterizing key parameter A k Failure model F for the time to occurrence of hard faults (open circuit, short circuit) k (t)=G F (t,Τ l (t),E l (t)), G P (·) and G F (·) are functions used to represent the degradation model and the failure model, respectively;

[0070] By using Monte Carlo random sampling, degradation curves (data corresponding to the degradation curves) and failure time samples corresponding to each key parameter are generated. The data corresponding to the degradation curves and the open-circuit and short-circuit data after reaching the failure time are then substituted into the multi-stress field simulation model to obtain the sub-circuit output parameters. These sub-circuit output parameters can represent performance degradation characteristics and functional failure characteristics. The multi-stress field simulation model in this embodiment is: [V n (t),I n [(t)]=G M (t,S,A k (t),F k (t)), where G M (·) is a function used to represent the obtained multi-stress field simulation model, V n (t), I n (t) represents the voltage and current parameters of the sub-circuit output node n as it degenerates with time t, V n (t), I n (t) can represent performance degradation characteristics and functional failure characteristics.

[0071] Step S3: Determine the parameters that significantly affect system failure, i.e., the system failure precursor parameters, through Pearson correlation coefficient analysis. Establish a system-level dynamic simulation model based on the power system's dynamic fault tree. Substitute the failure precursor parameters into the system-level dynamic simulation model to obtain the functional unit output parameters under equal stress, completing the transfer from the sub-circuit layer to the functional unit layer. The specific process includes the following steps:

[0072] Step S31: Through Pearson correlation coefficient analysis, select parameters that have a significant impact on power system failure from the key parameters identified in S22, i.e., system failure precursor parameters:

[0073] Calculate the Pearson correlation coefficient r for the input and output of S23, where:

[0074]

[0075] Where, x m y m These are the input values ​​of key parameters of the sub-circuit and the output values ​​(simulation values) of the sub-circuit output parameters. and It corresponds to the mean, s x and s y This represents the corresponding standard deviation, and N is the number of data points.

[0076] The parameter whose correlation coefficient r with the unit output curve is greater than a given threshold Z is determined as the system fault precursor parameter A. S System fault precursor parameter A S Including the intrinsic parameters R of sensitive components SSensitive key node parameter V S , can be represented as A S ={R S V S};

[0077] Step S32: Based on the topology of the power system dynamic fault tree in step S12, a dynamic logic gate component is constructed from simulation components. The simulation components include: sequential switches, programmable switches, registers, basic logic gates, etc. A system-level dynamic simulation model is established based on the dynamic logic gate component to obtain the interface between the sub-circuit level and the functional unit level, thereby realizing the conversion of the qualitative power system dynamic fault tree model into a quantitative simulation model.

[0078] Step S33: First, obtain the system fault precursor parameter A based on historical data. S This includes: R-values ​​fitted from historical data under equal stress. S V corresponding to historical test data under equal stress S ; The non-uniform stress historical test data V corresponding to the sensitive key node parameters test As a supplement, multiply it by the correction factor μ (0 < μ < 1) to obtain V′. S =(1-μ)V S +μV test ;

[0079] Will be by V S Corrected to V′ S The corresponding system fault precursor parameter A S Substituting into the system-level dynamic simulation model, the calculation aims to minimize the error between the simulated value and the actual value of the system-level dynamic simulation model. The value of this factor is changed through an optimization algorithm to correct the correction factor μ. The output parameters of the functional unit under constant stress are obtained through simulation. The output parameters of the functional unit can be represented as a degradation curve with a distribution, i.e., the degradation curve of the output parameters of the functional unit.

[0080] The above process allows V test A more accurate approximation of the data table under equal stress.

[0081] It should also be noted that the output parameters of a functional unit are a simulation result, and the certain conditions of the functional unit reflected by this result are called the functional unit state.

[0082] Step S4: Determine the functional unit degradation model based on the functional unit output parameter degradation curve under equal stress. Determine the functional unit reliability based on the functional unit degradation model and failure threshold, and use the weighting coefficient W. i,p The process of transferring functional units to the system task profile layer includes:

[0083] Step S41: Determine the functional unit degradation model based on the functional unit output parameter degradation curve under equal stress, and analyze the reliability of the functional unit based on the functional unit degradation model and failure threshold.

[0084] The p-th (p = 1, 2, ..., X) func The functional unit degradation model is represented as follows:

[0085] D p (t)=d(t;Φ)+ξ (2)

[0086] Among them, D p (t) represents the cumulative degradation amount, d(t; Φ) represents the degradation trajectory function of the degradation curve, Φ is the coefficient in the function (electrical, thermal stress, etc.); ξ is an uncertainty parameter that follows a normal distribution (related to individual tolerance differences and environmental noise).

[0087] Parameter D p (t) is a continuously differentiable monotonically increasing function, when D p (t)>D f Failure occurs at time t. Then, the failure probability function F of the p-th functional unit at time t is... p (t) is:

[0088]

[0089] Among them G f Let be the standard normal cumulative distribution function, with a mean of μ and a standard deviation of σ.

[0090] The reliability R of the output parameter of the p-th functional unit is calculated based on the failure probability function. out,p :

[0091] R out,p (t)=1-F p (t) (4)

[0092] Step S42: Solve for C according to the working conditions. p To D i weight coefficients W i,p The weighting coefficients of the output parameters of the p-th functional unit in the i-th system task profile; t 第p个功能单元时长 t represents the working time of the p-th functional unit in the i-th system task profile; 第i个系统任务剖面总时长 This represents the total duration corresponding to the i-th system task profile.

[0093] The reliability of the i-th system task profile is determined by the reliability of the output parameters of p functional units based on W. i,p Determined by weighting and normalization.

[0094] S5, System Task Profile Status Parameters Di Through weight coefficient W sys,i The data is then passed to the overall system task layer to ultimately determine reliability characteristics such as the overall system task reliability; this includes the following steps:

[0095] Determine D i To R sys weight coefficients Among them, W sys,i Let t be the weighting coefficient of the total system tasks over the entire lifecycle for the i-th system task profile; 第i个任务剖面出现次数对应的时长 t represents the total duration of the i-th system task profile calculated based on the number of occurrences. 系统总任务 Indicates the total duration of all system tasks;

[0096] The overall system task reliability throughout its entire lifecycle is determined by the reliability of each i-system task profile based on W. sys,i Determined by weighting and normalization.

[0097] Reliability R based on functional unit output parameters out，p The overall mission reliability R of the comprehensive life-cycle evaluation index system sys The reliability curve is obtained by calculating using equation (5).

[0098]

[0099] Based on reliability, other reliability characteristics such as unreliability, failure rate, and reliable lifetime can be obtained.

[0100] Example

[0101] This embodiment focuses on an aircraft power system. Its primary function is to perform voltage conversions of varying timings and magnitudes according to different preset task flows, serving as a fundamental guarantee for the normal operation of the aircraft's electrical equipment. The aircraft power system is a multi-level system comprising six interconnected sub-circuit units used for power supply, control, communication, and other functions. Utilizing system layering and dynamic fault tree construction, and considering the multi-level and multi-tasking characteristics of complex electronic systems, the reliability of the aircraft power system is predicted, taking into account the functional failures and performance degradation of key internal components, as well as the multi-level impacts.

[0102] First, based on the aircraft power system schematic and according to the signal hierarchy, the system is divided into five levels: component layer, sub-circuit layer, functional unit layer, system mission profile layer, and overall system mission layer (output reliability characteristics). The hierarchical system parameter transfer relationships are as follows: Figure 2 In reality, the reliability of information that can be obtained often decreases with each level. By treating each level as a network node using a directed acyclic graph construction method, the system includes X. comp +X sub +Xfunc +X sys +1 node: X comp individual components, X sub Individual circuits, X func Functional units, X sys One system task profile, one overall system task. The system network node state set is Ω = {A} l B m C p D i ,R sys}, where A l (l=1,2,...,X comp ) indicates component status parameters, B m (m=1,2,...,X sub ) represents the sub-circuit state parameters, C p (p = 1, 2, ..., X) func ) represents the functional unit state parameter, D i (i = 1, 2, ..., X) sys ) represents the system task profile state parameters, R sys It represents the overall reliability characteristics of the system (reliability, failure rate, etc.).

[0103] A multi-level power supply system (including multiple sub-circuit modules that perform various functions such as power supply, control, and communication and have mutual influences) is represented using the dynamic fault tree method. The functional structure of the power supply is characterized as follows: Figure 3 As shown. And the state parameter transfer method at each level is represented by directed edges, by A. l The stress field simulation model is passed to B. m B m The data is transmitted to C through a system-level dynamic simulation model. p C p By weight coefficient W i,p Passed to D i Finally, D i Through weight coefficient W sys,i Passed to R sys .

[0104] For this system, the overall system task reliability characteristic R sys The evaluation index is the weighted result of the output parameters of the three functional units: C1 monitoring, C2 driving, and C3 power supply. C1 includes independent sub-circuit units such as B1 control sub-circuit and B2 communication sub-circuit; C2 includes B3 control sub-circuit and B4 power supply sub-circuit; and C4 includes independent sub-circuit units such as B5 battery module sub-circuit and B6 control sub-circuit.

[0105] Secondly, taking the B4 power supply circuit as an example, sub-circuit-level modeling is performed to realize the transfer from the component layer to the sub-circuit. A multi-stress field simulation model of the B4 power supply circuit is established using EDA software, as follows: Figure 4 This process transforms external excitation into component stress. First, a simulation model of the thermistor is constructed. Based on test data of the component's electrical parameters at different ambient temperatures, a model is established to show how the component's parameters change with temperature. For example, the power MOSFET is modeled based on its main electrical parameters at 25℃ and 175℃. The component's electrical simulation model is E. l (t)=G E (t,S,T), thermal simulation model is T l (t)=G T (t,E,M). The electrical simulation model incorporates external electrical stress S and temperature stress T (where T={T1,…,T…) a ,…},T a The temperature of the a-th component is converted into the component's electrical stress E based on time t. l (t); The thermal simulation model incorporates the ambient temperature stress M and the electrical stress E of all components in the sub-circuit (where E = {V1, I1, P1, ... V1}). b ,I b ,P b ...}, where Vb, Ib, and Pb are the voltage, current, and power of the b-th component, respectively, and are converted into the thermal stress T on the surface of each component l. l (t);

[0106] First- and second-order sensitivity analyses of Sobol were performed to determine key parameters. Based on the B4 simulation model, parameter A was... l Input variables A1 to A9 are labeled. During their lifespan, the effects of a single parameter on the sub-circuit output signal V4, and the coupling effect of two parameters on the sub-circuit output signal V4, are analyzed when these parameters change by ±20%. The analysis results are as follows: Figure 5 As shown, if both the first-order and second-order sensitivities are higher than ±1%, it indicates that the impact of this parameter on the output will not be overwhelmed by noise during its lifespan, and it can be considered a critical parameter A. k ={R k V k}, including the intrinsic parameters R of the components k Key node parameter V k Since the first-order and second-order sensitivities of A7, A8, and A3 are all higher than ±1%, they are identified as key parameters.

[0107] Establish PoF models for key parameters, including: a degradation model A characterizing the key parameters. k (t)=G P (t,Τ l (t),E l(t)). Specifically, conventional devices such as resistors, capacitors, and transistors adopt the Arrhenius physical model, such as the resistance degradation model of resistor A7 as shown in equation (1), which satisfies the normal distribution; for DC / DC converters and other devices that cannot be covered by the physical model, the empirical degradation model is used as a supplement, for example: the degradation curve of circuit node signal A3 can be constructed by extending the Eileen model as shown in equation (2):

[0108]

[0109] Where R0 is the initial value of the intrinsic parameters of the component (resistance value in A7), R k (t) represents the intrinsic parameters of the component that degrade over time. V0 represents the initial value of the critical node parameter (output voltage in A3), V k (t) Degradation values ​​of critical node parameters over time. V l For set E l The voltage stress in (t), T l For set T l Thermal stress (in Kelvin) in (t), where t is time and k is the Boltzmann constant 8.617 × 10⁻⁶. -5 eV / K, ε1 and ε2 are random quantities. Activation energy E a Other parameters m, p, a, and b are undetermined coefficients, obtained through fitting experimental data. Specifically for A7:

[0110] R7(t)=R0~N(3400,33.12 2 )

[0111]

[0112] Failure model F characterizing hard fault (open circuit, short circuit) states of key parameters k (t)=G F (t,Τ l (t),E l (t)). The specific form is as follows: For example, the failure model of the A8 power chip is as shown in equation (3), which satisfies a normal distribution.

[0113]

[0114] Where F k (t) represents the component's failure lifetime. l For set E l The current stress in (t) is ε3, which is a random quantity.

[0115] In fact, A k (t)=G P (t,Τ l (t),E l(t) is a general expression of formulas (1) and (2), F k (t)=G F (t,Τ l (t),E l (t) is a general expression of formula (3).

[0116] Based on parameter optimization software, such as iSIGHT, a sub-circuit degradation simulation calculation platform is established, such as... Figure 6 Degradation curves and failure (lifetime end) time samples were generated using Monte Carlo random sampling. The degradation curve data and open-circuit and short-circuit data after reaching the failure time were then substituted into the simulation model, such as... Figure 7 In the B4 sub-circuit, the power MOSFETs operate in PWM mode, regulating the power of each output load according to the drive signal. The final sub-circuit output parameters are: [V n (t),I n [(t)]=G M (t,S,A k (t),F k (t)), where V n (t), I n (t) represents the voltage and current parameters of the output node n of the sub-circuit as it degrades over time t. These parameters can represent performance degradation characteristics and functional failure characteristics. For example, Vt represents the voltage and current parameters of the output node n of the sub-circuit as it degrades over a certain time t. n I n It is a degenerative characteristic, and when t is long, V n I n This will become a failure characteristic. Similarly, EDA software can be used to build models of each sub-circuit included in the power supply system and perform simulation analysis.

[0117] Then, further analysis using the Pearson correlation coefficient was conducted to determine the key parameter A. k ={R k V k The parameters that significantly affect the failure of the power supply system are determined in the sub-circuits A7, A8, and A3, i.e., the system failure precursor parameters. The Pearson correlation coefficient r is given by equation (4):

[0118]

[0119] x m and y m These are the input values ​​and simulation output values ​​of the key parameters of the sub-circuit, respectively. and It corresponds to the mean, s x and s yHere, Z represents the standard deviation, and N is the number of data points. Analysis shows that the Pearson correlation coefficients (r) of curves A7, A8, and A3 relative to V4 are 0.891, 0.853, and 0.716, respectively. A7 and A8 are greater than Z (in this analysis scenario, a value greater than Z = 0.85 is considered to meet the correlation requirement), making them the weakest parameters for B4, and thus identified as precursor parameters for system faults. Similarly, precursor parameters for system faults in other sub-circuits are determined.

[0120] A dynamic system model is established. Based on the location and type of fault precursor parameters, dynamic logic gate components are constructed in EDA software using simulation components such as sequential switches, programmable switches, registers, and basic logic gates. This enables the conversion from a qualitative dynamic fault tree model to a quantitative simulation model. For example, a simulated sub-circuit centered on the power supply is integrated into the model. When the voltage of multiple power supplies falls below a certain threshold, it will lead to the failure of subsequent sub-circuits. This simulation is achieved using a voltage-controlled switch component. The various sub-circuits are connected according to the system connection method, resulting in the final dynamic system model as follows: Figure 8 .

[0121] Substituting the system fault precursor parameters of subcircuits B4 (A7 and A8) along with those of other subcircuits into the dynamic simulation model, including: R values ​​fitted based on historical data under constant stress. S Historical test data V S Additionally, the non-uniform stress historical test data V of sensitive key node parameters will be used. test As a supplement, before substitution, multiply by the correction factor μ (0 < μ < 1), which is defaulted to 0.5, i.e., V′ S =0.5V S +0.5V test This allows for adjustment of the weight of this parameter in subsequent reliability prediction corrections, thereby improving the accuracy of reliability predictions. During the calculation process, an optimization algorithm is used to change the value of this factor, aiming to minimize the error between the simulated and actual values. Simulations yielded the functional unit output parameters under constant stress (including output degradation curves with distributions), where the degradation curve of the functional unit C2 output parameters is shown below. Figure 9 ;

[0122] Finally, the reliability is analyzed using the obtained degradation curves of the functional unit output parameters under equal stress, combined with the functional unit degradation model and failure threshold. Output voltage is a key output parameter determining the lifetime of C2, and it is related to other output parameters in series, directly determining whether the C2 control module can function. The degradation model of functional unit C2 is expressed as:

[0123] D2(t)=d(t;Φ)+ξ (5)

[0124] Where D2(t) is the cumulative degradation, d(t; Φ) is the degradation trajectory function representing the degradation curve, Φ is the coefficient in the function, and ξ is an uncertainty parameter that follows a normal distribution (related to individual tolerance differences and environmental noise). The parameter D2(t) is a continuously differentiable monotonically increasing function; when D2(t) > D... f Failure occurs at time t. Therefore, the failure probability function F2(t) at time t is:

[0125]

[0126] Among them G f Let be a standard normal cumulative distribution function with a mean of μ and a standard deviation of σ. The reliability R of the output parameters of functional unit C2 is calculated based on the lifetime distribution function. out,2 For example:

[0127] R out,2 (t)=1-F2(t) (7)

[0128] Based on this calculation process, the reliability curve of functional unit C2 under equal stress is obtained as follows: Figure 10 Similarly, the reliability curves of other functional units were obtained. Based on the existing experimental data, the lifespan of the functional units all met the reliability prediction range at a 95% confidence level, and the model accuracy met the requirements.

[0129] Subsequently, C was solved according to the working conditions. p To D i weight coefficients For example, the weighting coefficient W of the output parameter of the second functional unit C2 in the first system task profile D1. 1,2 =0.6. Solve for D. i To R sys weight coefficients For example, the weight coefficient W of the total system task in the entire life cycle of the first system task profile D1. sys,1 =0.2. The final system comprehensive evaluation coefficient calculation model is shown below:

[0130]

[0131] With 0.2(0.1R) out,1 +0.6R out,2 +0.3R out,3 For example, 0.2 indicates that the execution frequency of task profile D1 in this system accounts for 20%, R out,1 R out,2 R out,3The reliability of the output parameters corresponding to functional units C1, C2, and C3 are shown in the figures. 0.1, 0.6, and 0.3 represent the normalized values ​​of the weighting coefficients for the operating time of C1, C2, and C3 as a percentage of the total time of the first system task profile. The final comprehensive evaluation parameters of the power system and the overall system task reliability curve are shown below. Figure 11 The power system lifespan is approximately 1790 days at a reliability level of 0.9. Furthermore, other reliability characteristics can be obtained using this weighted method. By coupling the parameters of the multi-stress field simulation model and adaptively adjusting the correction factors for supplementary system failure precursor parameters, the model's predicted results can be corrected. Compared with historical test data of similar equipment, the lifespan prediction accuracy can reach over 90%, which is quite accurate.

[0132] The above examples of the present invention are merely illustrative of the computational model and process of the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is impossible to exhaustively list all possible implementations here. Any obvious variations or modifications derived from the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A multi-level reliability prediction method for power systems based on fault precursor parameters, characterized in that, Includes the following steps: S1. Characterizing the functional structure of the system based on the dynamic fault tree of the power system; The power system dynamic fault tree includes a component layer, a sub-circuit layer, a functional unit layer, a system task profile layer, and a system overall task layer. Directed edges represent the state parameter transmission method at each level, and component state parameters... The sub-circuit state parameters are transmitted to the sub-circuit layer through a multi-stress field simulation model. The functional unit state parameters are passed to the functional unit layer through the system-level dynamic simulation model. By weight coefficients The system task profile state parameters are passed to the system task profile layer. Through weight coefficients Passed to the system's overall task layer; S2. For each sub-circuit, establish a multi-stress field simulation model that transforms external excitation into component stress, and analyze the component state parameters. Sensitivity analysis to determine key parameters A failure physics model for key parameters is established by combining component stress. The degradation curve and failure time sample obtained by sampling are substituted into the multi-stress field simulation model of the sub-circuit to obtain the sub-circuit output parameters that can represent the performance degradation characteristics and functional failure characteristics; thus realizing the transfer from the component layer to the sub-circuit layer. S3. Based on the key parameters and output parameters of the sub-circuit, the parameters affecting the failure of the power supply system, i.e., the system fault precursor parameters, are determined through correlation coefficient analysis. A system-level dynamic simulation model is established based on the dynamic fault tree of the power supply system. The fault precursor parameters are substituted into the system-level dynamic simulation model to obtain the functional unit output parameters under equal stress, realizing the transfer from the sub-circuit layer to the functional unit layer. The functional unit output parameters can be represented as a degradation curve with distribution, i.e., the functional unit output parameter degradation curve. S4. Determine the functional unit degradation model based on the functional unit output parameter degradation curve under equal stress, and determine the reliability of the functional unit output parameters based on the functional unit degradation model and failure threshold. And through weight coefficients Implement the transfer from functional units to the system task profile layer; the weighting coefficients for implementing the transfer from functional units to the system task profile layer are as follows: in, The weighting coefficients of the output parameters of the p-th functional unit in the i-th system task profile; This represents the working time of the p-th functional unit in the i-th system task profile; This represents the total duration corresponding to the i-th system task profile; S5, System Task Profile Status Parameters Through weight coefficients The data is then passed to the overall system task layer to ultimately determine reliability characteristics such as the overall system task reliability. The specific process includes the following steps: Determine the weight coefficients ,in, Let be the weight coefficient of the total system tasks for the i-th system task profile throughout its entire lifecycle; This represents the total duration of the i-th system task profile calculated based on the number of occurrences. Indicates the total duration of all system tasks; Reliability R based on functional unit output parameters out，p Determine the overall mission reliability throughout its entire lifecycle. : （5） This represents the number of system task profiles. Based on overall task reliability Make a reliability prediction.

2. The method for constructing a multi-level reliability prediction model for a power system based on fault precursor parameters according to claim 1, characterized in that, Step S2 includes the following steps: S21, Establish The simulation model of the multi-stress field corresponding to each sub-circuit transforms the external excitation into the component stress. S22. Based on the multi-stress field simulation models corresponding to each of the constructed sub-circuits, the component state parameters of each sub-circuit are analyzed separately. Sensitivity analysis was performed to determine key parameters. ; S23. Determine the failure physical model for key parameters, including the parameters characterizing the key parameters. Degradation model over time and characterizing key parameters Failure model based on the time to occurrence of hard fault states , and These are functions used to represent the degradation model and the failure model, respectively; By random sampling, degradation curves and failure time samples corresponding to each key parameter are generated. The data corresponding to the degradation curves and the data corresponding to open circuits and short circuits after reaching the failure time are substituted into the multi-stress field simulation model to obtain the sub-circuit output parameters. The sub-circuit output parameters can represent the performance degradation characteristics and functional failure characteristics.

3. The method for constructing a multi-level reliability prediction model for a power system based on fault precursor parameters according to claim 2, characterized in that, The multi-stress field simulation model includes an electrical simulation model of the components. and thermal simulation model Furthermore, the electrical and thermal simulation models can transfer parameters to each other; The electrical simulation model incorporates external electrical stress S and temperature stress. Transformed into component electrical stress based on time t ; The thermal simulation model converts the ambient temperature stress M and the electrical stress of all components in the sub-circuit into thermal stress on the surface of each component l. .

4. The method for constructing a multi-level reliability prediction model for a power system based on fault precursor parameters according to claim 3, characterized in that, Step S22 includes the following steps: During the life cycle, respectively First and second order Sobol sensitivity analyses were performed to analyze the impact of single-parameter changes on the sub-circuit output and the coupling effect of changes in any two parameters on the sub-circuit output; parameters whose first and second order sensitivity values ​​are both higher than the sensitivity threshold were identified as critical parameters. Key parameters Includes component intrinsic parameters Key node parameters .

5. The method for constructing a multi-level reliability prediction model for a power system based on fault precursor parameters according to claim 4, characterized in that, The aforementioned multi-stress field simulation model: ,in, It is a function used to represent the simulation model corresponding to the multi-stress field. , These are the voltage and current parameters of the output node n of the sub-circuit as it degenerates over time t. , It can represent performance degradation characteristics and functional failure characteristics.

6. A method for constructing a multi-level reliability prediction model for a power system based on fault precursor parameters, as described in any one of claims 1 to 5, characterized in that, Step S3 specifically includes the following steps: S31. Based on the key parameters and output parameters of the sub-circuit, Pearson correlation coefficient analysis is used to screen out parameters affecting power system failure from the key parameters determined in S22, i.e., system fault precursor parameters. Including the intrinsic parameters R of sensitive components S Sensitive key node parameter V S ; S32. Based on the topology of the dynamic fault tree of the power system, a dynamic logic gate component is composed of simulation components. The simulation components include: sequential switches, programmable switches, registers, basic logic gates, etc. A system-level dynamic simulation model is established based on the dynamic logic gate component to obtain the interface between the sub-circuit level and the functional unit level, thereby realizing the conversion of the qualitative power system dynamic fault tree model into a quantitative simulation model. S33. First, obtain the system fault precursor parameter A based on historical data. S This includes: R-values ​​fitted from historical data under equal stress. S V corresponding to historical test data under equal stress S ; The non-uniform stress historical test data V corresponding to the sensitive key node parameters test As a supplement, it is multiplied by the correction factor μ to obtain ; Will be Revised to The corresponding system fault precursor parameter A S Substituting into the system-level dynamic simulation model, the calculation aims to minimize the error between the simulated value and the actual value of the system-level dynamic simulation model. The value of this factor is changed through an optimization algorithm to correct the correction factor μ. The output parameters of the functional unit under constant stress are obtained through simulation. The output parameters of the functional unit can be represented as a degradation curve with a distribution, i.e., the degradation curve of the output parameters of the functional unit.

7. The method for constructing a multi-level reliability prediction model for a power system based on fault precursor parameters according to claim 6, characterized in that, Step S4 describes the process of determining the functional unit degradation model based on the functional unit output parameter degradation curve under equal stress, which includes: The degradation model of the p-th functional unit is represented as follows: （2） in, Let d(t;Φ) be the cumulative degradation amount, d(t;Φ) be the degradation trajectory function representing the degradation curve, and Φ be the coefficient in the function; The uncertainty parameter is that it follows a normal distribution.

8. The method for constructing a multi-level reliability prediction model for a power system based on fault precursor parameters according to claim 7, characterized in that, Step S4 describes the process of determining the reliability of a functional unit based on its degradation model and failure threshold, which includes: Failure threshold Failure occurs at time t, and the failure probability function F of the p-th functional unit at time t is given. p (t) is: （3） in The function is the standard normal cumulative distribution function with a mean of 1 / 2. The standard deviation is ; The reliability of the output parameters of the p-th functional unit is calculated based on the failure probability function. .

Images