Hydrogen production power supply fault detection method, system and medium

Abstract

The invention relates to the technical field of power supply fault detection, in particular to a hydrogen production power supply fault detection method and system and a medium. The method comprises the following steps: collecting the current operation power and environment temperature of a hydrogen production power supply in real time; if the current operation power is lower than a preset first power threshold value and the environment temperature is lower than a preset temperature threshold value, starting a preset high-frequency sampling mode, and collecting a switching tube driving voltage waveform of a hydrogen production power supply and a direct-current output current waveform of an electrolytic reactor; calculating a driving voltage oscillation period according to the switching tube driving voltage waveform, and calculating a current fluctuation period according to the DC output current waveform; and when the driving voltage oscillation period is in a preset period range and the ratio of the current fluctuation period to the driving voltage oscillation period is in a preset ratio range, determining that the hydrogen production power supply has a critical oscillation fault. According to the method, the critical oscillation transmitted across the system can be identified in time by associating the cooperative waveform characteristics of the power supply and the electrolytic reactor, and potential damage to the membrane electrode is avoided.

Description

Technical Field

[0001] This invention relates to the field of power supply fault detection technology, and in particular to a method, system and medium for detecting faults in hydrogen production power supplies. Background Technology

[0002] Proton exchange membrane (PEM) electrolyzers are widely used in wind power / solar photovoltaic (PV) hydrogen production microgrid systems due to their fast dynamic response and high hydrogen purity. In such scenarios, the electrolyzer power supply needs to frequently track fluctuating renewable energy output and rapidly adjust within a range of 10%–120% of rated power. In actual operation, it has been found that when the electrolyzer power supply is under light load (e.g., below 30% of rated power) and the ambient temperature drops sharply, a critical oscillation phenomenon occurs in the power supply's internal control circuit: the switching transistor drive voltage experiences periodic critical conduction, and its oscillation frequency (approximately 200Hz–2kHz) is exactly synchronized with the fluctuation frequency of the output current on the DC side of the electrolyzer. Although this oscillation does not trigger conventional overvoltage or overcurrent protection, it creates periodic mechanical stress inside the electrolyzer. After long-term operation, this will lead to fatigue cracks in the proton exchange membrane, ultimately causing premature failure of the electrolyzer.

[0003] However, existing detection methods only focus on independent faults in the power source or the electrolytic reactor itself, lacking effective means to monitor the dynamic interaction characteristics of the power source-electrolytic reactor coupled system under light load and temperature alternating conditions, and thus cannot identify and warn of the aforementioned critical oscillation faults in the early stages. Summary of the Invention

[0004] Therefore, it is necessary for the present invention to provide a method, system and medium for detecting hydrogen production power source failures in order to solve at least one of the above-mentioned technical problems.

[0005] To achieve the above objectives, a method for detecting faults in a hydrogen production power source includes the following steps: Step S1: Real-time acquisition of the current operating power and ambient temperature of the hydrogen production power source; Step S2: If the current operating power is lower than the preset first power threshold and the ambient temperature is lower than the preset temperature threshold, then start the preset high-frequency sampling mode to collect the switching transistor drive voltage waveform of the hydrogen production power supply and the DC output current waveform of the electrolytic reactor. Step S3: Calculate the driving voltage oscillation period based on the switching transistor driving voltage waveform, and calculate the current fluctuation period based on the DC output current waveform; Step S4: When the driving voltage oscillation period is within the preset period range and the ratio of the current fluctuation period to the driving voltage oscillation period is within the preset ratio range, it is determined that the hydrogen production power supply has a critical oscillation fault. Step S5: When a critical oscillation fault is detected, output a fault alarm signal.

[0006] The beneficial effects of this invention are as follows: On the one hand, high-frequency sampling is triggered only when the operating power is less than or equal to the preset first power threshold and the ambient temperature is less than or equal to the temperature threshold. This concentrates computing power on the operating window where critical oscillations are most likely to be triggered, allowing weak oscillation components in the 200Hz–2kHz frequency band to be extracted cycle by cycle. This avoids data redundancy and DSP resource waste caused by blind sampling throughout the entire time period, and enables latent oscillations to be effectively revealed from background noise.

[0007] On the other hand, by using the driving voltage oscillation period as a benchmark, the current fluctuation period is locked to it by ratio. By using the physical nature of forced synchronization between the two to set a dual threshold of period and ratio, seemingly normal DC ripple can be distinguished from true critical oscillation. The false alarm rate is reduced from more than 15% in the traditional threshold method to less than 2%, and the detection delay is shortened to the millisecond level.

[0008] On the other hand, when the alarm is issued, the electrolytic reactor is only subjected to periodic stress with a micron-level amplitude and has not yet formed fatigue cracks. Maintenance personnel can eliminate the oscillation source by increasing the power or adjusting the damping of the temperature control loop within the planned shutdown window. The membrane electrode life is extended from the original 3000 hours of premature failure to the same life as the reactor design, saving about 20% of the replacement cost per reactor.

[0009] Preferably, the present invention provides a hydrogen production power supply fault detection system for performing the hydrogen production power supply fault detection method described above, the hydrogen production power supply fault detection system comprising: The data acquisition module is used to collect the current operating power and ambient temperature of the hydrogen production power source in real time. The waveform sampling module is used to start a preset high-frequency sampling mode to collect the switching transistor drive voltage waveform of the hydrogen production power supply and the DC output current waveform of the electrolytic reactor if the current operating power is lower than a preset first power threshold and the ambient temperature is lower than a preset temperature threshold. The period calculation module is used to calculate the driving voltage oscillation period based on the switching transistor driving voltage waveform and the current fluctuation period based on the DC output current waveform. The fault determination module is used to determine that the hydrogen production power supply has a critical oscillation fault when the driving voltage oscillation period is within a preset period range and the ratio of the current fluctuation period to the driving voltage oscillation period is within a preset ratio range. The alarm module is used to output a fault alarm signal when a critical oscillation fault is detected.

[0010] Preferably, the present invention also provides a computer medium storing a program that can be loaded by a processor and executed as described above for detecting hydrogen production power source failures. Attached Figure Description

[0011] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description with reference to the accompanying drawings: Figure 1 A flowchart illustrating the steps of a hydrogen production power source fault detection method according to an embodiment is shown.

[0012] Figure 2 A schematic diagram of a multi-batch fault confirmation mechanism according to an embodiment is shown.

[0013] Figure 3 A schematic diagram of a hydrogen production power source fault detection system architecture is shown in one embodiment. Detailed Implementation

[0014] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0015] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0016] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0017] To achieve the above objectives, please refer to Figures 1 to 3 This invention provides a method for detecting faults in hydrogen production power sources, comprising the following steps: Step S1: Real-time acquisition of the current operating power and ambient temperature of the hydrogen production power source; Step S2: If the current operating power is lower than the preset first power threshold and the ambient temperature is lower than the preset temperature threshold, then start the preset high-frequency sampling mode to collect the switching transistor drive voltage waveform of the hydrogen production power supply and the DC output current waveform of the electrolytic reactor. Step S3: Calculate the driving voltage oscillation period based on the switching transistor driving voltage waveform, and calculate the current fluctuation period based on the DC output current waveform; Step S4: When the driving voltage oscillation period is within the preset period range and the ratio of the current fluctuation period to the driving voltage oscillation period is within the preset ratio range, it is determined that the hydrogen production power supply has a critical oscillation fault. Step S5: When a critical oscillation fault is detected, output a fault alarm signal.

[0018] Preferably, the conditions for activating the preset high-frequency sampling mode in step S2 further include: If the current operating power is lower than the preset second power threshold, and the rate of decrease of the ambient temperature per unit time exceeds the preset temperature change rate threshold, the preset high-frequency sampling mode will be forcibly activated. In some embodiments, the preset second power threshold refers to the power lower limit set in advance by the system, which is used to determine whether the hydrogen production power supply has entered a low power operation state; the rate of decrease of ambient temperature per unit time refers to the continuous acquisition of ambient temperature by temperature sensor and the calculation of temperature change per unit time; when both conditions are met simultaneously, the system immediately forces the start of the preset high-frequency sampling mode.

[0019] In some embodiments, the preset high-frequency sampling mode is a phased, high-speed data acquisition strategy. Its activation condition is independent of the conventional sampling logic and it only takes effect when a dual trigger signal of low power and rapid cooling is detected. Once activated, this mode will prioritize locking the switching transistor drive voltage waveform at a high sampling rate. Furthermore, it will reduce the sampling rate to acquire the DC output current waveform of the electrolytic cell, thereby taking into account both high-frequency details and long-term trends, and ensuring that the dynamic response of the electrolytic cell can be fully recorded even under extreme cooling conditions.

[0020] Wherein, the preset first power threshold is greater than the preset second power threshold; The preset high-frequency sampling mode includes a phased sampling strategy. In the first phase, the switching transistor drive voltage waveform is acquired at a preset first sampling frequency. In the second phase, the DC output current waveform of the electrolytic reactor is acquired at a preset second sampling frequency. The preset first sampling frequency is higher than the preset second sampling frequency.

[0021] In some embodiments, the staged sampling strategy is a high-speed diagnostic mechanism triggered when the hydrogen production power supply experiences a low-power + rapid cooling abnormal condition: the first stage involves sampling the pulse voltage at the IGBT half-bridge drive port at 1MHz (preset first sampling frequency) for 40 seconds. The first stage is a snapshot to capture drive delays or spikes caused by cold starts of the electrolytic cell; the second stage immediately slows down to 200kHz (preset second sampling frequency) and collects the DC bus current of the electrolytic cell within a 5ms window to identify current oscillations or interruption risks caused by sudden temperature drops; the data from the two stages are spliced ​​together by timestamp through the same high-speed cache.

[0022] In some embodiments, the differentiated design of a preset first sampling frequency higher than a preset second sampling frequency is intended to prioritize ensuring that high-frequency fault characteristics are not missed under the limited DSP computing power and storage resources of the hydrogen production power source: the first stage of 1MHz sampling rate only captures the rising and falling edges of the driving voltage with nanosecond-level precision to detect the initial distortion of Hall zero drift during the switching transient of the power device; the second stage of 200kHz sampling rate covers the complete envelope of the electrolytic reactor output current to capture the slow drift of insulation impedance disturbance at low temperature; the two stages are seamlessly switched to avoid data explosion caused by high sampling throughout the process and to ensure that critical fault transients are fully recorded.

[0023] Preferably, step S3 includes: The switching transistor drive voltage waveform is processed by a sliding window based on a preset window width and a preset sliding step size to obtain waveform segment sequences corresponding to multiple time windows. In some embodiments, sliding window processing is used to divide the continuous switching transistor drive voltage waveform into short, overlapping segments of fixed length so that the periodic characteristics of the drive pulse can be extracted in each short segment. The preset window width refers to the length of time captured each time, and the preset sliding step size refers to the time increment of each forward sliding of the window. The window width ensures that at least two complete switching transients are included, while the step size is smaller than the window width, so that adjacent windows overlap, thereby avoiding the failure to detect instantaneous frequency drift caused by a sudden drop in temperature.

[0024] In some embodiments, the switching transistor drive voltage waveform is continuously acquired by the DSP at a sampling rate of 1MHz and cyclically stored in RAM; the system sets the window width to 50µs (corresponding to 50 sampling points) and the sliding step size to 10µs (10 sampling points). After each sliding, a new waveform segment is captured and the waveform segment sequence is generated sequentially; this parameter combination can capture two complete drive pulses within 50µs and achieve a 5 / 4 overlap rate through a tight step size of 10µs, so that the subsequent cycle calculation has a time resolution of 0.5µs.

[0025] Within each window, the rising edge trigger point of the switching transistor drive voltage waveform is extracted, and the time interval between adjacent trigger points is calculated as half an oscillation period. In this embodiment of the invention, the DSP first performs 3-point moving average denoising on the current 50µs waveform segment, and then uses a fixed level of 0.3V as the threshold for rising edge crossing detection; whenever the waveform jumps from below 0.3V to above 0.3V with a slope greater than 0.1V / µs, the index of that sampling point is recorded as a valid trigger point, and the index difference between two consecutive trigger points is recorded. n multiplied by the 1µs sampling interval gives the half-cycle. = n×1µs; if adjacent If the difference between n and the sampling point exceeds ±5, it is determined to be noise-triggered and discarded, ensuring that the half-cycle calculation is derived only from the actual switching action.

[0026] In one implementation of this invention, after denoising, the window waveform detected a total of 4 crossovers. The second crossover, with a slope of only 0.08V / µs, was determined to be a false trigger and was removed. The three valid rising edge indices that were retained were 12, 32, and 52, corresponding to half-cycles of 20µs and 20µs, respectively. This result was directly used for the subsequent 16-point weighted average, successfully identifying the 20ns-level micro-drift caused by low temperature.

[0027] Step S3-3: Take the weighted average of K consecutive half-cycles, and multiply the weighted average by 2 to obtain the driving voltage oscillation period of the current window, where K is an integer greater than or equal to 10 and less than or equal to 100; In this embodiment of the invention, the DSP selects the most recent 16 (K=16) effective half-cycles. The samples are assigned linear weights based on their time of origin, with the newest sample having a weight of 1 / 16 and the oldest sample having a weight of 1 / 256. The weighted average is obtained by summing these weights. Multiplying this by 2 gives the current window drive voltage oscillation period. If a certain half of the cycle is If the deviation is greater than 20%, it is considered an abnormal transient and is removed. New samples are automatically added to ensure... It reflects minute drifts rather than occasional spikes.

[0028] Of particular importance, after determining the driving voltage oscillation period, the following is also included: Calculate the peak variation coefficient of the voltage waveform within each window; When the coefficient of variation exceeds the preset coefficient threshold, the calculation result of the oscillation period of the corresponding window is marked as low confidence. In some embodiments, the peak variation coefficient is used to quantify the instability of the driving voltage waveform within the window, preventing period calculation distortion due to glitches or noise; the system obtains the current window after completing step S3-3. Then, immediately extract the maximum value within the 50µs waveform segment. and minimum value Calculate peak value = – and the window mean The coefficient of variation (CV) is calculated by comparing the values. / When CV > 15% (preset coefficient threshold), significant amplitude fluctuations are detected within the judgment window, and the current value is then... Marked as low confidence level, it will not participate in the subsequent weighted average.

[0029] In some embodiments, the DSP employs a confidence bit mechanism: after each window ends, it first executes... , The test is performed, and then the CV is calculated using a single-cycle division instruction; if CV = 18% exceeds the limit, the bit is set. The highest bit of the register is 1, indicating low confidence, and the data point is automatically removed in the next weighting cycle; for example, in a vehicle bump test, if the third window introduces a ±0.4V spike due to mechanical vibration, with a CV of 20%, it is marked as low confidence and excluded. An outlier of 19.2µs was found; the subsequent 16-point average used only high-confidence data, and the final output... =20.0µs, error <0.5µs.

[0030] The DC output current waveform is bandpass filtered to obtain the filtered waveform. In this embodiment of the invention, a second-order Butterworth IIR bandpass filter is deployed on the DC output current waveform sampled at 200kHz. The passband is set to 500Hz–5kHz, which covers the current oscillations commonly seen in electrolytic reactors during low-temperature cooling. The filter uses 16-bit fixed-point arithmetic, with a cutoff frequency error of <1%, which suppresses high-frequency switching noise while retaining the slow fluctuations caused by insulation impedance disturbances.

[0031] Calculate the autocorrelation function of the filtered waveform; In some embodiments, the autocorrelation function is used to extract hidden periodic components from the bandpass-filtered DC output current waveform; the 200kHz, 5ms long filter sequence output in steps S3-4 is then processed. (Total 1000 points) Perform unbiased autocorrelation calculation.

[0032] Most importantly, the specific calculation formula for unbiased autocorrelation is as follows: , Among the delays The maximum number of points is 200 (corresponding to 1ms). The multiplication and accumulation process is completed using a single-cycle MAC instruction from the DSP, taking approximately 30µs. The calculation result... exist The value at 0 is normalized to 1, and the rest are normalized to 1. The amplitude reflects the similarity between the signal and its own delayed version, thus highlighting the fundamental frequency fluctuations of the current hidden in the noise, without interrupting the bus current throughout the process.

[0033] in, For the autocorrelation function in delay The value at that location, The delay index has values ​​of 0, 1, 2, ... , This represents the total length of the sequence (number of sampling points). The value of the filtered current waveform at the i-th sampling point is... For the waveform in delay The instantaneous value after that, This represents the number of effective product terms, used for unbiased normalization.

[0034] In some embodiments, the DSP employs a segmented-accumulation strategy to reduce memory usage: first, the 1000-point filtered data is divided into 4 segments of 256 points each (with zero padding at the end); local autocorrelation is calculated for each segment; and then the 4 sets of intermediate results are accumulated to obtain the final result. For example, in a low-temperature startup test, the filtered current exhibited a weak oscillation of approximately 460µs, with a local amplitude of only ±0.3A, and the autocorrelation function was... A significant peak value of 0.65 is observed at point 92 (corresponding to 460µs), which is much higher than the noise floor value of 0.15. The system can accurately extract the current fluctuation period based on this peak value.

[0035] Identify the first significant peak in the autocorrelation function and use the time interval corresponding to the first significant peak as the current fluctuation period.

[0036] In some embodiments, a significant peak is defined as a point above the noise floor and a local maximum, used to derive from the autocorrelation function. Extract the true fluctuation period of the current; firstly, for The sequence was smoothed using a 3-point moving average, and then from... =1 to start scanning point by point, if If the following conditions are met simultaneously: ① amplitude > 0.3 (significance threshold), ② the point is greater than the two adjacent points (local maxima), then it is recorded as the first significant peak, and its corresponding delay index is... Multiply by the sampling interval t=5µs, thus obtaining the current ripple period. = × If no peak value meets the condition within 1ms, then the current is determined to have no periodic fluctuation and is set to t; =0.

[0037] In some embodiments, the DSP employs a threshold-slope dual-judgment mechanism: first, the noise floor = mean(R[10:50]) is calculated, and then the significance threshold is dynamically set to the floor + 0.2; for example, in a low-temperature start-up test, after smoothing R[ A peak value of 0.65 appears at τ=92, which is higher than the threshold of 0.35 and is a local maximum. The system immediately locks. =92, converted to =460µs; This result is consistent with the drive voltage cycle. =20µs ratio =1.1 falls within the range of 0.8-1.2, thus confirming that the current oscillation is synchronized with the power stage switching.

[0038] Preferably, step S3 further includes generating a feature matching signal, specifically including: The rate of change of the duration of the high-level drive voltage over N consecutive cycles is calculated based on the waveform of the drive voltage of the switching transistor to determine the duty cycle variation characteristics, where N is an integer greater than or equal to 5 and less than or equal to 20. In some embodiments, the duty cycle variation characteristics are used to capture the drive pulse width drift caused by a sudden temperature drop or Hall zero drift accumulation during the on-load aging process of the hydrogen production power supply; firstly, the drive voltage oscillation period obtained in step S3-3 is... Perform period alignment, and then count the duration of the high level within the same period. After caching continuously for 5 ≤ N ≤ 20 cycles, calculate The first linear slope of the sequence is the rate of change. D=( – ) / (N× The rate of change is output in the form of ‰, when | When D|>1‰, it is determined to be an abnormal duty cycle drift.

[0039] In some embodiments, the DSP takes N=10, and the latched 10 cycles... Perform a sliding linear fit on the data; the least squares slope within the window is the current value. D; if D being greater than +1‰ twice consecutively indicates that the drive high level widens cycle by cycle, possibly due to a decrease in the gate threshold voltage at low temperatures; if If D is less than -1‰, it indicates a shrinking duty cycle or an increased oscillation of the electrolytic reactor current; this positive or negative change characteristic is written to the status register in real time.

[0040] Calculate the ratio of the effective values ​​of the second to fifth harmonics to the effective value of the fundamental wave in the DC output current waveform to determine the harmonic distortion. In some embodiments, harmonic distortion is used to characterize the current distortion of the hydrogen production power supply during load aging due to duty cycle drift or Hall zero drift accumulation; first, the current fluctuation period is... By performing integer-cycle truncation, a 128-point synchronous sampling sequence is obtained. Further, an embedded FFT kernel is invoked to extract the RMS values ​​of the 1st to 5th harmonics. According to THD= Calculate the harmonic distortion (THD); this ratio is output as a percentage, and when THD > 3%, it is considered abnormal distortion.

[0041] In some embodiments, the DSP takes 5ms of current waveform data (1000 points), performs a 256-point segmented average to reduce random noise, and then performs a radix-2 FFT; the computation time is <40µs, and the duty cycle changes abruptly. Decline and , The THD rose sharply from 1.2% to 4.1% during the synchronous rise, triggering the harmonic over-limit flag; this flag, along with the duty cycle change characteristics and insulation response bit value, was written into the status register.

[0042] When the duty cycle change characteristic exceeds the preset change characteristic threshold and the harmonic distortion exceeds the preset distortion threshold, a feature matching signal is generated.

[0043] In some embodiments, the characteristic matching signal is the ultimate alarm for both time-domain drift and frequency-domain distortion anomalies during the aging process of the hydrogen production power source under load; the duty cycle change rate D and harmonic distortion (THD) are simultaneously fed into a dual-window comparator: when | When D|>1‰ (preset change characteristic threshold) and THD>3% (preset distortion threshold), it is determined to be a duty cycle-harmonic joint anomaly, and the GPIO output is immediately pulled high as a feature matching signal. This signal is a single clock cycle pulse, directly fed into the abnormal frame tracing queue, used to trigger instantaneous shutdown before the next switching cycle, to ensure that the electrolytic reactor and power devices are not subjected to further aging impact.

[0044] In some embodiments, the DSP employs an event-triggered mechanism: the duty cycle rate of change is updated every 10ms, and the THD is updated every 5ms; when two consecutive comparison results both satisfy | If D>1‰ and THD>3%, then the feature matching flag is locked and the current feature is frozen. The D and THD values ​​are written to the read-only register to prevent subsequent noise refresh; the lock triggers a 1µs high-level pulse to drive the optocoupler-isolated shutdown link, so that the boost transistor drive is soft-drated to zero within 10µs.

[0045] Preferably, in step S4, determining that the hydrogen production power source has experienced a critical oscillation fault also requires meeting a time duration condition, which includes: When the drive voltage oscillation period is detected to be within the preset period range for the first time and the ratio of the current fluctuation period to the drive voltage oscillation period is within the preset ratio range, the timer is started. In some embodiments, a preset cycle range and a preset ratio range together constitute the entry conditions for the hydrogen production power supply to enter steady-state aging assessment; the driving voltage oscillation cycle is used as the threshold for aging assessment. With current fluctuation period Feed into a two-condition comparator: when Falling within 15µs–25µs (preset period range) and the ratio ρ= / When the value falls within the range of 0.8–1.2 (preset ratio range), it is determined that the fuel cell-power cascade loop is in the controllable resonance region, and a 32-bit free-running timer is immediately started; the timer accuracy is 1µs.

[0046] In some embodiments, the DSP employs an edge-triggered mechanism: and Updated every 5ms, when two consecutive comparison results both satisfy 15µs ≤ If the value is ≤25µs and 0.8≤ρ≤1.2, the steady-state access flag is locked, and a single pulse enable is sent to the timer module; the timer starts accumulating from zero and pauses freezing when any subsequent update exceeds the range, and the accumulated value is written to a read-only register as the effective aging time; for example, in a high-altitude test, =20µs, =22µs, ρ=1.1 falls exactly within the dual window. After the timer starts, it continuously accumulates for 8.6s until the temperature drops sharply, causing ρ to rise to 1.3 and the timer is paused. During this period, all duty cycle and harmonic data are marked as the same aging period.

[0047] If the condition that the driving voltage oscillation period is within the preset period range and the ratio of the current fluctuation period to the driving voltage oscillation period is within the preset ratio range is met for a continuous preset observation period, the time continuity condition is satisfied.

[0048] In some embodiments, the preset observation time and the preset time ratio threshold together constitute a dual requirement for the duration and proportion of the hydrogen production power source's aging steady state; based on the activated 1µs precision timer, checks are performed every 5ms within a continuous 2s (preset observation time). and If the time falls within the range of 15µs–25µs and 0.8–1.2, the effective flag counter is incremented. When the observation time ends, the effective counter is divided by the total number of checks (400 times). If the percentage is >85% (preset time ratio threshold), the time continuity condition is determined to be met, and a steady-state aging admission signal is generated.

[0049] In some embodiments, the DSP internally uses a state machine: upon power-up, the valid counter and the total counter are cleared; after entering the 2-second observation window, a check is performed every 5ms during an interrupt period. and If the condition is met, the valid counter is incremented by 1, and the total number of attempts is fixed at 1. The percentage is calculated immediately upon the end of the window. If the percentage = 344 / 400 = 86%, exceeding the 85% threshold, the persistence qualification flag is set, and the current window is locked. D and THD baseline values; for example, in a low-temperature start-up test, only 82% of the time was spent in the first 1.8 seconds due to environmental disturbances. If an occasional boundary violation occurs, continue to observe and do not exit prematurely until the percentage rises to 86% after a full 2 ​​seconds to confirm that the time continuity condition is met.

[0050] Preferably, in step S4, determining that the hydrogen production power source has a critical oscillation fault also requires meeting characteristic matching conditions, which include: Obtain the feature matching signal. When the feature matching signal is in a valid state, the feature matching condition is met.

[0051] In some embodiments, the feature matching signal is the ultimate attack alarm for hydrogen production power supply aging diagnosis, and its effective state is equivalent to the trigger token of the abnormal frame; the duty cycle change rate D and harmonic distortion (THD) are fed into a dual-window comparator: when | When D|>1‰ and THD>3%, the hardware comparator outputs a high-level single pulse, which is the feature matching signal. The signal width is fixed at 1µs and is directly latched into the status register bit [Match]. Once [Match]=1, it is determined that the feature matching condition is met.

[0052] In some embodiments, the DSP employs an event-triggered mechanism: refreshing every 5ms. If both D and THD exceed the limit in two consecutive comparisons, then set [Match] and freeze the current value. D, THD, and timestamp are used to prevent noise-driven refresh; the feature matching signal synchronously drives the optocoupler isolation link, enabling the boost diode drive to be softly dated to zero within 10µs; for example, in a high-altitude low-temperature test... If D=1.2‰ and THD=4.1% both exceed the limit, and [Match] is set to 1, then ( (D, THD, periodic number) are encapsulated as an abnormal frame and pushed into the traceback queue.

[0053] Preferably, step S2 further includes synchronously acquiring at least one operating parameter of the electrolytic reactor cooling system; in step S4, determining that the hydrogen production power source has a critical oscillation fault also requires meeting auxiliary parameter conditions, which include: After determining that the driving voltage oscillation period is within the preset period range and the ratio of the current fluctuation period to the driving voltage oscillation period is within the preset ratio range, the deviation state of the operating parameters is determined according to the preset normal operating range. In some embodiments, the deviation state is used to quantify the degree of drift of the actual operating parameters of the hydrogen production power source relative to the reference value within the steady-state aging window; after confirming that the time duration condition is qualified, the current real-time bus current Ireal-time, bus voltage Ureal-time, and returned status code Sreal-time are subtracted item by item from the preset historical triplet (Ihistory, Uhistory, Shistory) to obtain the current offset. I. Voltage Deviation U and status code offset S; If any offset exceeds the allowable band—for example, | I|>0.5A、| U|>0.3V or If S≠0, then the parameter is determined to be in a deviated state.

[0054] In some embodiments, the DSP employs a three-channel parallel comparison strategy: after locking the steady-state window, it first performs a Q15 format fixed-point subtraction between the real-time I and the historical I, and then shifts the result 3 bits to the right to obtain a resolution of ±0.1A. I; Calculate similarly U、 S; If all three offsets are within the allowable band, then the zero offset flag is set, and subsequent drift amplitudes are directly returned to zero; if If I = +0.7A exceeds the limit, the forward current drift is recorded and immediately marked as a candidate for the amplitude command of the subsequent reverse current pulse.

[0055] When the operating parameters deviate, the auxiliary parameter conditions are met; The operating parameters include cooling water flow rate or cooling water temperature.

[0056] Preferably, step S4 is followed by enhanced conditions for fault confirmation, specifically including: After the first determination of a critical oscillation fault, maintain the preset high-frequency sampling mode and repeat steps S3 and S4 in the subsequent M consecutive preset sampling batches. In some embodiments, a critical oscillation fault refers to a situation where, during the operation of a hydrogen power supply under load, the driving voltage cycle or current fluctuation cycle suddenly jumps out of the steady-state window and lasts for two check cycles. This is considered a critical state that may induce drift-zero drift-insulation joint inspection failure. Once triggered, the DSP immediately locks the current 1MHz / 200kHz dual-frequency sampling clock, maintains the high-frequency sampling mode, and repeats steps S3 (duty cycle-harmonic extraction) and S4 (insulation disturbance application-response latching) in a loop within subsequent M=8 preset sampling batches (5ms per batch) to ensure high-density snapshots of the fault evolution process.

[0057] In some embodiments, the DSP employs a batch-sequence dual-counter structure: after the first determination of critical oscillation, the CRIT_OSC flag is immediately set to mask any frequency reduction requests; at the end of each sampling batch, the current counter is... D, THD, insulation response bits, and batch number (0→7) are bundled and written to the circular buffer; if the fault is not resolved after the 8th batch, the last frame of data is frozen and an abnormal aging frame is forcibly generated, driving a momentary shutdown; for example, in a high-altitude low-temperature test, The critical oscillation was triggered by a sudden increase to 28µs. The system maintained high-frequency sampling and captured the complete evolution of THD from 2% to 5% and the insulation response bit from 0x00 to 0x02 in 8 consecutive batches. The proportion of the abnormal frame exceeded the threshold when the batch number was 3, and the abnormal frame was encapsulated in advance. The hydrogen production power supply continued to output 1.6kW until the end of batch 8 before power was cut off, ensuring that the critical oscillation was recorded with high fidelity throughout.

[0058] If more than half of the M preset sampling batches show a critical oscillation fault, then the critical oscillation fault is confirmed. In some embodiments, a majority-determination mechanism is used to filter out sporadic disturbances, ensuring that only persistent critical oscillations are identified as system-level faults. After completing M=8 high-frequency sampling batches, the system immediately counts the number of batches judged as critical oscillations. If this number is ≥5 (i.e., 8×0.5+1), the critical oscillation fault establishment flag is set, and the last frame is locked. D, THD, and the insulation response bit serve as the main body of subsequent abnormal frames; this majority voting strategy requires no additional hardware and relies entirely on the DSP's internal counter, giving the final decision at the end of the last batch.

[0059] In some embodiments, the DSP employs a 1+8 accumulator structure: the first batch determined to be critical sets the accumulator to 1, and each subsequent critical batch increments by 1; if a non-critical batch appears midway, the accumulator remains unchanged to prevent bounce; for example, in a high-altitude low-temperature test, batches 0–2… If the boundary is continuously exceeded and the accumulator is set to 3, batch 3 is skipped due to instantaneous recovery. Batches 4-7 are critical again, and the accumulator eventually reaches 7. If more than half of the batches are 5, the critical oscillation fault is confirmed, and a high-level lockout pulse is generated.

[0060] Otherwise, the initial determination result is determined to be a transient interference, and the fault flag of the hydrogen production power source is cleared; In some embodiments, transient interference is used to eliminate the misleading effect of occasional glitches on the determination of critical oscillation; after 8 high-frequency sampling batches are completed, if the number of batches judged as critical oscillation is <5 (not more than half), the first limit violation is considered to be a random disturbance rather than a real fault, and the CRIT_OSC flag is pulled low and the written abnormal data cache is cleared; at the same time, the DSP sends a fault flag clearing pulse to the driver board to restore the standard sampling frequency.

[0061] In some embodiments, the DSP employs a veto logic: after completing the statistics for the 8th batch, if the accumulator value is ≤4, a transient interference branch is triggered. This involves first blocking the writing of the abnormal frame, then pulling the fault flag level output by the GPIO low, and resetting the duty cycle and harmonic baseline. For example, in a vehicle acceleration / deceleration test, batch 0 fails due to a loose power cable. If the jump is detected as critical, and all 7 batches return to normal, and the accumulator = 1, then it is determined to be a transient interference. The fault flag is cleared and the sampling is reduced to normal.

[0062] Where M is a positive integer greater than or equal to 2.

[0063] It should be noted that in this case, Figure 2 This diagram illustrates a multi-batch fault confirmation mechanism. The upper part of the diagram shows the time-series distribution of M=8 consecutive sampling batches, with the batch order represented by a horizontal time axis, arranged sequentially from batch 0 to batch 7. Each batch is labeled with a rectangle. Batches judged to be in a "critical" state (batches 0, 1, 2, 4, 5, 6, 7) are displayed with white text filled in black, while batch 3 (batch 3) is displayed with black text in a hollow box. The judgment result for each batch is indicated below the rectangle. The middle part of the diagram shows the state evolution curve of the critical oscillation accumulator. The horizontal axis represents the batch number, and the vertical axis represents the accumulated value. The curve starts with an accumulated value of 1 for batch 0, accumulates through batches 1 and 2 to 3, and batch 3, being judged to be normal, retains an accumulated value of 3. Further, batches 4, 5, 6, and 7 continue to accumulate, finally reaching an accumulated value of 7 in batch 7. The diagram uses a horizontal dashed line to mark the half threshold = 5. The accumulator curve exceeds this threshold after batch 5 and remains above it, demonstrating the judgment condition of exceeding half.

[0064] Preferably, the present invention provides a hydrogen production power supply fault detection system for performing the hydrogen production power supply fault detection method described above, the hydrogen production power supply fault detection system comprising: The data acquisition module 101 is used to collect the current operating power and ambient temperature of the hydrogen production power source in real time. The waveform sampling module 102 is used to start a preset high-frequency sampling mode and collect the switching tube drive voltage waveform of the hydrogen production power supply and the DC output current waveform of the electrolytic reactor if the current operating power is lower than a preset first power threshold and the ambient temperature is lower than a preset temperature threshold. The period calculation module 103 is used to calculate the driving voltage oscillation period based on the switching transistor driving voltage waveform and to calculate the current fluctuation period based on the DC output current waveform. The fault determination module 104 is used to determine that the hydrogen production power supply has a critical oscillation fault when the driving voltage oscillation period is within a preset period range and the ratio of the current fluctuation period to the driving voltage oscillation period is within a preset ratio range. The alarm module 105 is used to output a fault alarm signal when a critical oscillation fault is detected.

[0065] Preferably, the present invention also provides a computer medium storing a program that can be loaded by a processor and executed as described above for detecting hydrogen production power source failures.

[0066] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0067] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for detecting a failure of a hydrogen production power source, characterized by, The method comprises the following steps: Step S1: Real-time acquisition of the current operating power of the hydrogen production power supply and the ambient temperature; Step S2: If the current operating power is lower than a preset first power threshold and the ambient temperature is lower than a preset temperature threshold, a preset high-frequency sampling mode is started, and the switch tube driving voltage waveform of the hydrogen production power supply and the direct current output current waveform of the electrolysis stack are collected; Step S3: The driving voltage oscillation period is calculated according to the switch tube driving voltage waveform, and the current fluctuation period is calculated according to the direct current output current waveform; Step S4: When the driving voltage oscillation period is within a preset period range and the ratio of the current fluctuation period to the driving voltage oscillation period is within a preset ratio range, it is determined that the hydrogen production power supply has a critical oscillation fault; Step S5: When it is determined that the critical oscillation fault occurs, a fault alarm signal is output.

2. The hydrogen production power supply fault detection method of claim 1, wherein, The condition for starting the preset high-frequency sampling mode in step S2 further includes: If the current operating power is lower than a preset second power threshold and the descending rate of the ambient temperature within a unit time exceeds a preset temperature change rate threshold, the preset high-frequency sampling mode is forcibly started; Wherein, the preset first power threshold is greater than the preset second power threshold; The preset high-frequency sampling mode includes a phased sampling strategy, the switch tube driving voltage waveform is collected at a preset first sampling frequency in the first phase, and the direct current output current waveform of the electrolysis stack is collected at a preset second sampling frequency in the second phase, wherein the preset first sampling frequency is higher than the preset second sampling frequency.

3. The method of claim 1, wherein, Step S3 includes: According to the preset window width and the preset sliding step, the switch tube driving voltage waveform is processed by sliding window to obtain a waveform segment sequence corresponding to multiple time windows; In each window, the rising edge trigger point of the switch tube driving voltage waveform is extracted, and the time interval of adjacent trigger points is calculated as half an oscillation period; The weighted average value of the continuous K half periods is taken, and the weighted average value is multiplied by 2 to obtain the driving voltage oscillation period of the current window, wherein K is an integer greater than or equal to 10 and less than or equal to 100; The direct current output current waveform is bandpass filtered to obtain a filtered waveform; The autocorrelation function of the filtered waveform is calculated; The first significant peak value in the autocorrelation function is identified, and the time interval corresponding to the first significant peak value is taken as the current fluctuation period.

4. The method of claim 1, wherein, Step S3 further includes generating a feature matching signal, specifically including: Based on the switch tube driving voltage waveform, the change rate of the driving voltage high level duration in the continuous N periods is calculated to determine the duty cycle change feature, wherein N is an integer greater than or equal to 5 and less than or equal to 20; The ratio of the effective value of the second to fifth harmonics to the effective value of the fundamental wave in the direct current output current waveform is calculated to determine the harmonic distortion degree; When the duty cycle change feature exceeds the preset change feature threshold and the harmonic distortion degree exceeds the preset distortion degree threshold, the feature matching signal is generated.

5. The method of claim 1, wherein, In step S4, the determination that the hydrogen production power supply has a critical oscillation fault also needs to meet a time duration condition, and the time duration condition includes: When the driving voltage oscillation period is within the preset period range and the ratio of the current fluctuation period to the driving voltage oscillation period is within the preset ratio range for the first time, a timer is started; If the time ratio of the condition that the driving voltage oscillation period is within the preset period range and the ratio of the current fluctuation period to the driving voltage oscillation period is within the preset ratio range is greater than the preset time proportion threshold, the time persistence condition is met.

6. The method of claim 1, wherein, In step S4, the determination of the critical oscillation fault of the hydrogen production power supply also needs to meet a feature matching condition, and the feature matching condition includes: Obtaining a feature matching signal, and when the feature matching signal is in an effective state, the feature matching condition is met.

7. The method of claim 1, wherein, Step S2 further includes synchronously collecting at least one operating parameter of the electrolysis stack cooling system; in step S4, the determination of the critical oscillation fault of the hydrogen production power supply also needs to meet an auxiliary parameter condition, and the auxiliary parameter condition includes: After the condition that the driving voltage oscillation period is within the preset period range and the ratio of the current fluctuation period to the driving voltage oscillation period is within the preset ratio range is met, the deviation state of the operating parameter is determined according to the preset normal operation range; When the operating parameter deviates, the auxiliary parameter condition is met; wherein the operating parameter includes the cooling water flow or the cooling water temperature.

8. The method of claim 1, wherein, After step S4, an enhanced condition for fault confirmation is further included, specifically including: After the first determination of the critical oscillation fault, the preset high-frequency sampling mode is maintained, and steps S3 and S4 are repeatedly executed in the subsequent M preset sampling batches; If the determination results of more than half of the batches in the M preset sampling batches are all critical oscillation faults, it is confirmed that the critical oscillation fault is established; Otherwise, the first determination result is determined to be transient interference, and the fault flag of the hydrogen production power supply is cleared; Wherein M is a positive integer greater than or equal to 2.

9. A hydrogen-producing power supply fault detection system characterized by comprising: A hydrogen production power supply fault detection system for executing the hydrogen production power supply fault detection method of claim 1, comprising: A data acquisition module for collecting the current operating power and the ambient temperature of the hydrogen production power supply in real time; A waveform sampling module for starting a preset high-frequency sampling mode if the current operating power is lower than a preset first power threshold and the ambient temperature is lower than a preset temperature threshold, and collecting the switch tube driving voltage waveform of the hydrogen production power supply and the direct current output current waveform of the electrolysis stack; A period calculation module for calculating the driving voltage oscillation period according to the switch tube driving voltage waveform, and calculating the current fluctuation period according to the direct current output current waveform; A fault determination module for determining that the hydrogen production power supply has a critical oscillation fault when the driving voltage oscillation period is within the preset period range and the ratio of the current fluctuation period to the driving voltage oscillation period is within the preset ratio range; An alarm module for outputting a fault alarm signal when it is determined that a critical oscillation fault occurs.

10. A computer storage medium, characterized in that, A program is stored, which can be loaded and executed by a processor to perform the hydrogen production power supply fault detection method of any one of claims 1 to 8.

Images