New energy vehicle charging pile fault monitoring method and system

Abstract

The invention discloses a new energy automobile charging pile fault monitoring method and system, and the method comprises the following steps: S1, collecting charging pile thermal radiation field data through a micro-lens array, constructing multi-source input in combination with current, voltage and power time sequence data, and calculating a temperature gradient in real time; S2, dynamically generating a safety threshold based on a depth deterministic strategy gradient algorithm; s3, when the monitoring data exceed a dynamic safety threshold, triggering a fault positioning process; and S4, calculating a fault probability Pfail based on the historical failure rate of the root fault equipment, and inputting a fault value-at-risk model to output an early warning level and a time window. According to the invention, a thermal radiation detection error is compressed to + / -1.3 DEG C by reconstructing a temperature field through a light field, and by combining 30fps high-frequency sampling, early overheating (the temperature rise rate gt; 0.5 DEG C / s), the detection rate is increased to 98.2%; the dynamic threshold engine quantifies the life loss of the part in real time through the aging factor delta tau, and the false alarm rate is reduced to 4.3%.

Description

Technical Field

[0001] This invention relates to the field of charging pile monitoring technology, specifically to a method and system for monitoring faults in new energy vehicle charging piles. Background Technology

[0002] With the increasing popularity of electric vehicles, safety incidents caused by DC charging pile malfunctions are increasing by an average of 37.2% annually. Traditional monitoring solutions face three major technical bottlenecks:

[0003] Insufficient sensing accuracy: Infrared thermal imagers are affected by ambient light interference, resulting in a temperature detection error of ±3℃ (the allowable value of IEC 61851-23 standard), leading to an early overheating missed detection rate as high as 26.8%.

[0004] Rigid early warning mechanism: Fixed threshold strategies (such as alarms for >80℃) cannot adapt to the dynamic process of component aging. Before a fire broke out in a 7kW pile of a certain brand due to the progressive failure of the relay, the system did not trigger any alarms.

[0005] Lack of location capability: Existing impedance spectroscopy analysis requires shutdown for testing and cannot distinguish between concurrent faults such as contactor arcing (failure rate 43%) and capacitor bulging (failure rate 29%). Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for monitoring faults in new energy vehicle charging piles, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for monitoring faults in new energy vehicle charging piles, comprising the following steps:

[0008] S1. Collect thermal radiation field data of the charging pile through a microlens array, combine it with current, voltage, and power time-series data to construct a multi-source input, and calculate the temperature gradient in real time.

[0009] S2. A safety threshold is dynamically generated based on a deep deterministic strategy gradient algorithm, wherein the threshold integrates ambient temperature, humidity, power gradient and equipment aging factor;

[0010] S3. When the monitoring data exceeds the dynamic safety threshold, the fault location process is triggered:

[0011] Computing the vulnerability of power topology nodes using graph-diffused convolutional networks;

[0012] The antigen-antibody matching mechanism is used to identify fault characteristics and root cause equipment.

[0013] S4. Calculate the failure probability P based on the historical failure rate of the equipment with root cause failures. fail Input the fault risk value model and output the warning level and time window.

[0014] The state space of the deep deterministic policy gradient algorithm in step S2 is defined as follows:

[0015]

[0016] Where T is temperature, H is humidity, and P is power. Δτ is the power gradient, and Δτ is the aging factor calibrated through accelerated aging tests.

[0017] The formula for calculating the aging factor Δτ is:

[0018]

[0019] Where, σ thermal For thermal stress, σ c N is the critical thermal stress of the material. cycle N represents the number of charge-discharge cycles. f is the material fatigue life factor calibrated by accelerated aging test, and m is the shape parameter of the Weibull distribution.

[0020] The antigen-antibody matching mechanism in step S3 includes:

[0021] The current harmonic distortion rate f1 and the temperature gradient are used. The packet loss rate is encoded as an antigen vector Ag;

[0022] Matching antibody library Abs is calculated based on affinity:

[0023]

[0024] The formula for reconstructing the thermal radiation field in step S1 is:

[0025]

[0026] Where, Φ i denoted as lens transmittance, O as object light field, PSF as point spread function, β as ambient light compensation coefficient, and sampling rate ≤ 30%.

[0027] The fault risk value in step S4 is defined as follows:

[0028]

[0029] Where the loss function is L = ω1·P fail +ω2·t urgent P fail Based on the historical failure rate calculation of equipment with root cause failures, t urgent This represents the time urgency coefficient.

[0030] The tail of the fault probability distribution is fitted using extreme value theory, and the distribution function is:

[0031]

[0032] Where ξ is the shape parameter, β is the scale parameter, and μ is the position parameter, which is fitted using historical fault data.

[0033] The reward function of the dynamic threshold engine is:

[0034]

[0035] Where FPR is the false positive rate, FNR is the false negative rate, and ω1+ω2=1, is dynamically optimized through reinforcement learning.

[0036] The ambient light compensation coefficient β of the light field virtual sensing module was determined by dark field calibration test, with a calibration period of ≤24 hours.

[0037] A fault monitoring system for new energy vehicle charging piles includes:

[0038] The light field virtual sensing module includes a microlens array with an aperture ≤100μm and an optical transmission matrix processor, with an array density ≥200 lenses / square centimeter;

[0039] Dynamic threshold engine with built-in LSTM-GRU hybrid network and DDPG optimizer;

[0040] The core of fault localization integrates a graph diffusion convolutional network and a clone selection algorithm unit.

[0041] The cloud-based early warning platform deploys a fault risk value calculator and an extreme value theory analyzer.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] 1. The optical field reconstruction temperature field of this invention compresses the thermal radiation detection error to ±1.3℃. Combined with 30fps high-frequency sampling, it increases the detection rate of early overheating (temperature rise rate > 0.5℃ / s) to 98.2%. The dynamic threshold engine quantifies the component life loss in real time through the aging factor Δτ, reducing the false alarm rate to 4.3%. The immune matching mechanism realizes fault component-level localization with an accuracy of 95.7% and a response time shortened to 47ms.

[0044] 2. The gallium arsenide microlens array of this invention replaces 12 discrete sensors, reducing the hardware cost of single-pile monitoring by 62.7%; the edge-cloud collaborative architecture supports concurrent processing of 50 piles, reducing computing power consumption by 78%; the online antibody library update mechanism enables the system's misdiagnosis rate to continuously decrease over time (<2.1% after 2000 hours).

[0045] 3. The FoVaR early warning model of this invention predicts the risk of failure 4 hours in advance (the traditional method is only 0.5 hours), provides a quantitative handling time limit, and the five-level early warning system is integrated with the operation and maintenance procedures, which improves the timeliness rate of manual intervention for high-risk failures to 100%. The extreme value distribution model is adapted to the temperature and humidity differences between the north and south, and maintains an early warning accuracy rate of 92.4% in environments ranging from -25℃ to 50℃. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the process of the present invention;

[0047] Figure 2 This is a system block diagram of the present invention. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] Please see Figure 1-2 The present invention provides a technical solution: a method for monitoring faults in new energy vehicle charging piles, comprising the following steps:

[0050] S1. Collect thermal radiation field data of the charging pile through a microlens array, combine it with current, voltage, and power time-series data to construct a multi-source input, and calculate the temperature gradient in real time.

[0051] S2. A safety threshold is dynamically generated based on a deep deterministic strategy gradient algorithm, wherein the threshold integrates ambient temperature, humidity, power gradient and equipment aging factor;

[0052] S3. When the monitoring data exceeds the dynamic safety threshold, the fault location process is triggered:

[0053] Computing the vulnerability of power topology nodes using graph-diffused convolutional networks;

[0054] The antigen-antibody matching mechanism is used to identify fault characteristics and root cause equipment.

[0055] S4. Calculate the failure probability P based on the historical failure rate of the equipment with root cause failures. fail Input the fault risk value model and output the warning level and time window.

[0056] The state space of the deep deterministic policy gradient algorithm in step S2 is defined as follows:

[0057]

[0058] Where T is temperature, H is humidity, and P is power. Δτ is the power gradient, and Δτ is the aging factor calibrated through accelerated aging tests.

[0059] The formula for calculating the aging factor Δτ is:

[0060]

[0061] Where, σ thermal For thermal stress, σ c N is the critical thermal stress of the material. cycle N represents the number of charge-discharge cycles. f is the material fatigue life factor calibrated by accelerated aging test, and m is the shape parameter of the Weibull distribution.

[0062] The antigen-antibody matching mechanism in step S3 includes:

[0063] The current harmonic distortion rate f1 and the temperature gradient are used. The packet loss rate is encoded as an antigen vector Ag;

[0064] Matching antibody library Abs is calculated based on affinity:

[0065]

[0066] The formula for reconstructing the thermal radiation field in step S1 is:

[0067]

[0068] Where, Φ i denoted as lens transmittance, O as object light field, PSF as point spread function, β as ambient light compensation coefficient, and sampling rate ≤ 30%.

[0069] The fault risk value in step S4 is defined as follows:

[0070]

[0071] Where the loss function is L = ω1·P fail +ω2·t urgent P fail Based on the historical failure rate calculation of equipment with root cause failures, t urgent This represents the time urgency coefficient.

[0072] The tail of the fault probability distribution is fitted using extreme value theory, and the distribution function is:

[0073]

[0074] Where ξ is the shape parameter, β is the scale parameter, and μ is the position parameter, which is fitted using historical fault data.

[0075] The reward function of the dynamic threshold engine is:

[0076]

[0077] Where FPR is the false positive rate, FNR is the false negative rate, and ω1+ω2=1, is dynamically optimized through reinforcement learning.

[0078] The ambient light compensation coefficient β of the light field virtual sensing module was determined by dark field calibration test, with a calibration period of ≤24 hours.

[0079] A fault monitoring system for new energy vehicle charging piles includes:

[0080] The virtual light field sensing module employs a gallium arsenide microlens array (lens aperture 80μm, density 220 lenses / cm²). 2 The optical transmission matrix is ​​processed by a Xilinx Zynq-7000 FPGA, with an ambient light compensation coefficient β = 0.18 (obtained through dark field calibration).

[0081] Dynamic threshold engine: Deployed on NVIDIA Jetson TX2 edge computing unit (1.3 TFLOPS computing power), running LSTM-GRU hybrid network (128 hidden layers).

[0082] The core of fault localization is a cloud-based dual Tesla V100 GPU running a graph diffusion convolutional network (5 layers) and a clone selection algorithm (antibody library size 1024).

[0083] Early warning platform: The FoVaR calculator is running on an Alibaba Cloud ECS server (8 cores, 32GB).

[0084] Step S1: The microlens array acquires the thermal radiation field at 30fps, and the temperature field is calculated using the reconstruction formula;

[0085] Output: Temperature gradient (Calculate the temperature difference between three adjacent frames; trigger an alarm if the threshold is >0.5℃ / s).

[0086] Example 1:

[0087] Step S1:

[0088] Collect the base radiation value under shaded conditions;

[0089] The ambient light compensation coefficient β is calculated to be 0.18;

[0090] Output the temperature field matrix T(x,y) according to the reconstruction formula;

[0091] Calculate the temperature gradient

[0092] Step S2:

[0093] Obtain the state vector S t =[35,60,120,0.8,0.63] T ;

[0094] The DDPG network output security threshold δsafe = 0.57;

[0095] determination Trigger location;

[0096] Step S3:

[0097] Construct antigen Ag = [0.17, 1.2, 0.05] T ;

[0098] Antibody library affinity was calculated to be 0.91.

[0099] Output matching result: main relay contact;

[0100] Step S4:

[0101] Historical query failure rate = 0.88;

[0102] The calculated loss is L = 0.6 × 0.88 + 0.4 × 0.9 = 0.888;

[0103] Fitting the extreme value distribution;

[0104] Output: FoVaR 0.95 =86 (Orange alert, time window 4 hours)

[0105] Warning level:

[0106] Green (Safe): If 0 ≤ FoVaR < 40;

[0107] Blue (attention): If 40 ≤ FoVaR < 60;

[0108] Yellow (Warning): If 60 ≤ FoVaR < 80;

[0109] Orange (High Risk): If 80 ≤ FoVaR < 90;

[0110] Red (Emergency): If FoVaR ≥ 90.

[0111] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

Claims

1. A method for monitoring faults in new energy vehicle charging piles, characterized in that, Includes the following steps: S1. Collect thermal radiation field data of the charging pile through a microlens array, combine it with current, voltage, and power time-series data to construct a multi-source input, and calculate the temperature gradient in real time. S2. A safety threshold is dynamically generated based on a deep deterministic strategy gradient algorithm, wherein the threshold integrates ambient temperature, humidity, power gradient and equipment aging factor; S3. When the monitoring data exceeds the dynamic safety threshold, the fault location process is triggered: Computing the vulnerability of power topology nodes using graph-diffused convolutional networks; The antigen-antibody matching mechanism is used to identify fault characteristics and root cause equipment. S4. Calculate the failure probability P based on the historical failure rate of the equipment with root cause failures. fail Input the fault risk value model and output the warning level and time window.

2. The method for monitoring faults in new energy vehicle charging piles according to claim 1, characterized in that: The state space of the deep deterministic policy gradient algorithm in step S2 is defined as follows: Where T is temperature, H is humidity, and P is power. Δτ is the power gradient, and Δτ is the aging factor calibrated through accelerated aging tests.

3. The method for monitoring faults in a new energy vehicle charging pile according to claim 2, characterized in that: The formula for calculating the aging factor Δτ is: Where, σ thermal For thermal stress, σ c N is the critical thermal stress of the material. cycle N represents the number of charge-discharge cycles. f is the material fatigue life factor calibrated by accelerated aging test, and m is the shape parameter of the Weibull distribution.

4. The method for monitoring faults in new energy vehicle charging piles according to claim 1, characterized in that: The antigen-antibody matching mechanism in step S3 includes: The current harmonic distortion rate f1 and the temperature gradient are used. The packet loss rate is encoded as an antigen vector Ag; Matching antibody library Abs is calculated based on affinity:

5. The method for monitoring faults in new energy vehicle charging piles according to claim 1, characterized in that: The formula for reconstructing the thermal radiation field in step S1 is: Where, Φ i denoted as lens transmittance, O as object light field, PSF as point spread function, β as ambient light compensation coefficient, and sampling rate ≤ 30%.

6. The method for monitoring faults in new energy vehicle charging piles according to claim 1, characterized in that: The fault risk value in step S4 is defined as follows: Where the loss function is L = ω1·P fail +ω2·t urgent P fail Based on the historical failure rate calculation of equipment with root cause failures, t urgent This represents the time urgency coefficient.

7. A method for monitoring faults in new energy vehicle charging piles according to claim 6, characterized in that: The tail of the fault probability distribution is fitted using extreme value theory, and the distribution function is: Where ξ is the shape parameter, β is the scale parameter, and μ is the position parameter, which is fitted using historical fault data.

8. The method for monitoring faults in new energy vehicle charging piles according to claim 1, characterized in that: The reward function of the dynamic threshold engine is: Where FPR is the false positive rate, FNR is the false negative rate, and ω1+ω2=1, is dynamically optimized through reinforcement learning.

9. A method for monitoring faults in new energy vehicle charging piles according to claim 1, characterized in that: The ambient light compensation coefficient β of the light field virtual sensing module was determined by dark field calibration test, with a calibration period of ≤24 hours.

10. A fault monitoring system for new energy vehicle charging piles, used to implement the detection and adjustment method according to any one of claims 1-9, characterized in that, include: The light field virtual sensing module includes a microlens array with an aperture ≤100μm and an optical transmission matrix processor, with an array density ≥200 lenses / square centimeter; Dynamic threshold engine with built-in LSTM-GRU hybrid network and DDPG optimizer; The core of fault localization integrates a graph diffusion convolutional network and a clone selection algorithm unit. The cloud-based early warning platform deploys a fault risk value calculator and an extreme value theory analyzer.

Images