Control Method of Integrated Training System for Wind-Solar Hybrid Hydrogen Production and Fuel Cell Power Generation

Abstract

This invention provides a control method for an integrated training system of wind-solar hybrid hydrogen production and fuel cell power generation, comprising: synchronously collecting operating data of a proton exchange membrane electrolyzer, a buffer hydrogen storage cylinder, a hydrogen pipeline, a fuel cell, a DC bus, and an energy storage battery through a central control unit; performing time alignment, validity processing, and rolling diagnostic window division through an edge computing module; generating three types of conserved residual sequences based on the deviations between the theoretical water consumption, theoretical hydrogen production, and theoretical oxygen production of the proton exchange membrane electrolyzer and the corresponding actual material quantities; determining an active detection window when the residuals continuously exceed limits, generating an optimal active detection parameter set through a bat algorithm, and controlling the fuel cell hydrogen discharge valve to perform short-pulse opening; extracting hydrogen discharge echo feature sets and cross-matching them with the three types of conserved residual sequences to obtain the hydrogen energy link anomaly location results; and performing hierarchical control and recovery verification based on the hydrogen energy link anomaly location results.

Description

Technical Field

[0001] This invention relates to the field of integrated energy control technology for hydrogen energy systems and electric vehicles, and in particular to a control method for an integrated training system for wind-solar hybrid hydrogen production and fuel cell power generation for integrated energy supply of electric vehicles. Background Technology

[0002] With the development of wind power, photovoltaic power generation, and the electric vehicle (pure electric vehicle and hydrogen fuel cell vehicle) industries, utilizing renewable energy to drive water electrolysis to produce hydrogen, and then storing hydrogen energy and supplying power to electric vehicles through hydrogen storage devices and fuel cells, has become a core development direction for the comprehensive utilization of new energy and the interaction between vehicles, grids, and power stations. To meet the practical training and microgrid control strategy verification needs related to the new energy vehicle industry, integrated training systems that combine wind power generation devices, photovoltaic power generation devices, water electrolysis hydrogen production devices, hydrogen storage devices, fuel cells, and electrical loads have emerged in existing technologies.

[0003] Existing training systems typically use threshold judgments based on hydrogen concentration, hydrogen storage pressure, fuel cell hydrogen inlet pressure, stack voltage, and operating temperature. When these parameters reach the set thresholds, control actions such as alarms and stopping hydrogen production are executed. While this type of control can handle anomalies with obvious characteristics, its ability to identify anomalies in their early stages is limited. Anomalies such as minor leaks in the hydrogen pipeline, partial blockages in the hydrogen supply pipeline, sluggish operation of the hydrogen inlet solenoid valve, and insufficient water replenishment in the proton exchange membrane electrolyzer can already affect the system's operational stability before the hydrogen concentration reaches the alarm threshold.

[0004] Because the proton exchange membrane electrolyzer, buffer hydrogen storage tank, hydrogen pipeline, and fuel cell are connected sequentially, abnormalities in different components may exhibit similar operating characteristics. For example, a decrease in actual hydrogen production could be caused by an abnormality in the proton exchange membrane electrolyzer, or by a minor leak in the hydrogen pipeline or flow sensor drift; a slow recovery of the hydrogen inlet pressure in the fuel cell could be caused by insufficient hydrogen supply capacity, or by a sluggish operation of the hydrogen inlet solenoid valve or partial blockage in the hydrogen pipeline. Judging solely based on a single parameter or fixed threshold is insufficient to accurately pinpoint the source of the abnormality, easily leading to false alarms, missed alarms, or delayed responses.

[0005] Therefore, this invention proposes a control method for an integrated training system of wind-solar hybrid hydrogen production and fuel cell power generation. The information disclosed in the background section is only for enhancing understanding of the background of this disclosure and may therefore contain prior art information that is not common knowledge to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing a control method for an integrated training system for wind-solar hybrid hydrogen production and fuel cell power generation, thereby solving the technical problems mentioned in the background section.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The control method for the integrated training system of wind-solar hybrid hydrogen production and fuel cell power generation includes the following steps: S1. The central control unit synchronously collects hydrogen energy link operation data, and the edge computing module completes time alignment, validity processing and rolling diagnostic window division to generate hydrogen energy link diagnostic data package. S2. Calculate the theoretical water consumption, theoretical hydrogen production and theoretical oxygen production based on the operating current sequence of the proton exchange membrane electrolyzer, and compare them with the actual water consumption, actual hydrogen production and actual oxygen production to generate three types of conservation residual sequences. When at least one type of conservation residual continuously exceeds the allowable range of the corresponding residual, an active detection request is generated. S3. Determine the active detection window based on the active detection request and the safe operation parameter table. Within the active detection safety constraint group, determine the optimal active detection parameter group using the bat algorithm. Control the fuel cell hydrogen exhaust valve to perform short pulse opening and generate a pulse response data packet. S4. Extract hydrogen emission echo feature groups from the impulse response data packets, and perform cross-matching of the hydrogen emission echo feature groups with three types of conserved residual sequences to obtain the hydrogen energy link anomaly localization results. S5. Generate a hierarchical control instruction set based on the anomaly location results of the hydrogen energy link, execute the corresponding control actions, determine the anomaly handling results based on the post-control verification data packet, and update the normal operation benchmark library and the active detection reference library.

[0008] S1 specifically includes: collecting sensor data and controller feedback data from the proton exchange membrane electrolyzer, buffer hydrogen storage cylinder, fuel cell, DC bus, energy storage battery, and electrical load at a unified sampling time through the central control unit, adding timestamps and valid sampling identifiers to form the raw data stream of the hydrogen energy link; performing time alignment, abnormal data removal, and gas flow standard state conversion on the raw data stream of the hydrogen energy link through the edge computing module to form the valid data sequence of the hydrogen energy link; calculating the actual water consumption, actual hydrogen production, actual oxygen production, and single-chip voltage dispersion according to the rolling diagnostic window, and combining the operating data, status data, and valid data channel identifiers to form the hydrogen energy link diagnostic data package.

[0009] S2 specifically includes: reading the operating current sequence, operating temperature sequence, and inlet deionized water conductivity sequence from the hydrogen energy link diagnostic data package; integrating the operating current sequence; and calculating the theoretical water consumption, theoretical hydrogen production, and theoretical oxygen production based on the Faraday efficiency correction coefficient in the normal operation benchmark library to form a theoretical material quantity set; correcting the actual water consumption based on the non-electrolysis water consumption benchmark; and comparing the corrected actual water consumption, actual hydrogen production, and actual oxygen production with the theoretical material quantity set to form three types of conserved residual sequences. When at least one type of conserved residual exceeds the allowable range of the corresponding residual for three consecutive valid residual windows, an active detection request containing residual triggering information and valid data channel identifier is generated; the valid residual window is a rolling diagnostic window where all input data channels are valid and the proton exchange membrane electrolyzer is in a stable operating state.

[0010] S3 specifically includes: reading active detection requests and hydrogen energy link diagnostic data packets; determining the active detection window and active detection safety constraint group based on hydrogen concentration, hydrogen storage pressure, fuel cell hydrogen inlet pressure, DC bus voltage, energy storage battery state of charge, and active detection power margin; calling the active detection reference library within the active detection safety constraint group; optimizing the single opening duration, short pulse interval duration, number of openings, and response acquisition duration of the fuel cell hydrogen exhaust valve using the bat algorithm to obtain the optimal active detection parameter group; controlling the fuel cell hydrogen exhaust valve to perform short pulse opening according to the optimal active detection parameter group, synchronously acquiring response data, and forming a pulse response data packet.

[0011] S4 specifically includes: reading the pulse response data packet, segmenting the data according to the actual opening and closing times of the fuel cell hydrogen exhaust valve, extracting the hydrogen inlet pressure drop amplitude, hydrogen inlet pressure recovery time, hydrogen storage pressure disturbance amplitude, output current deviation, and improvement ratio of single-chip voltage dispersion to form a hydrogen exhaust echo feature set; cross-matching the hydrogen exhaust echo feature set with three types of conserved residual sequences, calling the anomaly feature template set to calculate the anomaly localization confidence of candidate anomaly types, and forming a candidate anomaly localization result set; The location confidence difference between the highest and second-highest anomaly location confidence is compared. If the confidence difference reaches the preset value, the anomaly location result of the hydrogen energy link is determined. If it does not reach the preset value, a retest request is generated. After the retest request is generated, it is fed back to step S3 to re-determine whether there is an active detection window. If the active detection window is confirmed to exist, the active detection parameter group for retest is generated by the bat algorithm, and the hydrogen exhaust valve of the fuel cell is controlled to re-execute the short pulse opening.

[0012] S5 specifically includes: reading the anomaly location results of the hydrogen energy link, the three types of conserved residual sequences, and the hydrogen energy link diagnostic data package; calculating the control risk value according to the hierarchical control rule table; generating a hierarchical control instruction set; entering the safety control takeover state; issuing the hierarchical control instruction set according to the execution order; forming a control instruction confirmation record based on the actual equipment status feedback; generating a post-control verification data package; recalculating the three types of conserved residuals based on the post-control verification data package; recalculating the hydrogen emission echo characteristic group when the verification impulse response data is included; determining the verification deviation; restoring operation, raising the control level, or executing level 4 risk disposal according to the anomaly handling results; and updating the normal operation benchmark library and active detection reference library when the recovery conditions are met.

[0013] The beneficial effects of this invention are as follows: This invention synchronously collects and effectively processes operational data from the proton exchange membrane electrolyzer, buffer hydrogen storage tank, hydrogen pipeline, and fuel cell. Based on the deviations between theoretical water consumption, theoretical hydrogen production, and theoretical oxygen production and their corresponding actual material quantities, it generates three types of conserved residual sequences. This allows for early identification of abnormal signs before the hydrogen concentration reaches the alarm threshold, improving the timeliness of anomaly detection in the hydrogen energy chain. When active detection safety conditions are met, the fuel cell hydrogen discharge valve is controlled to open with a short pulse, and hydrogen discharge echo feature groups formed by changes in hydrogen inlet pressure, hydrogen storage pressure disturbances, and single-chip voltage dispersion are extracted. By cross-matching the hydrogen discharge echo feature groups with the three types of conserved residual sequences, it is possible to distinguish between proton exchange membrane electrolyzer anomalies, minor leaks in the hydrogen pipeline, delayed action of the hydrogen inlet solenoid valve, partial blockage in the hydrogen pipeline, and abnormal gas-water state of the fuel cell, reducing the risk of misjudgment caused by relying solely on a single parameter.

[0014] This invention utilizes the Bat Algorithm to optimize the single-opening duration, short-pulse interval duration, number of short-pulse openings, and response acquisition duration of the fuel cell hydrogen exhaust valve within an active detection safety constraint group. While ensuring that the hydrogen storage pressure, DC bus voltage, and battery state of charge remain within allowable ranges, it improves the distinguishability of abnormal responses and reduces the impact of active detection on normal power supply. An edge computing module locally performs data alignment, residual calculation, hydrogen exhaust echo feature extraction, and anomaly localization. Based on the anomaly localization results of the hydrogen energy link, a hierarchical control command set is generated. Parameter correction, operational restrictions, anomaly branch isolation, or emergency handling are executed according to the risk level, improving the targeting and efficiency of anomaly handling.

[0015] This invention recalculates three types of conservation residuals and hydrogen emission echo characteristic groups after anomaly handling, and determines whether the anomaly has been eliminated based on the verification deviation. Equipment operation is restored step-by-step only when recovery conditions are met, and verified data is updated to the normal operation benchmark library and active detection reference library, thus forming a closed-loop control process of anomaly identification, active detection, anomaly location, hierarchical control, and recovery verification. Applicable to integrated training systems for wind-solar hybrid hydrogen production and fuel cell power generation, this invention can transform the formation, identification, location, and handling of hydrogen energy link anomalies into reproducible training content. It facilitates simulation training for the safe operation and fault handling of hydrogen energy systems, as well as the operation and maintenance of integrated energy supply stations for electric vehicles (integrated photovoltaic-storage-hydrogen charging stations), providing strong technical support for the new energy vehicle and hydrogen energy industries. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the control method of the integrated power generation training system according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the integrated power generation training system framework according to an embodiment of the present invention. Detailed Implementation

[0017] 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.

[0018] Example 1: As Figure 1 As shown, this embodiment provides a control method for an integrated training system for wind-solar hybrid hydrogen production and fuel cell power generation, including the following steps: S1. The central control unit synchronously collects hydrogen energy link operation data, and the edge computing module completes time alignment, validity processing and rolling diagnostic window division to generate hydrogen energy link diagnostic data package. S2. Calculate the theoretical water consumption, theoretical hydrogen production and theoretical oxygen production based on the operating current sequence of the proton exchange membrane electrolyzer, and compare them with the actual water consumption, actual hydrogen production and actual oxygen production to generate three types of conservation residual sequences. When at least one type of conservation residual continuously exceeds the allowable range of the corresponding residual, an active detection request is generated. S3. Determine the active detection window based on the active detection request and the safe operation parameter table. Within the active detection safety constraint group, determine the optimal active detection parameter group using the bat algorithm. Control the fuel cell hydrogen exhaust valve to perform short pulse opening and generate a pulse response data packet. S4. Extract hydrogen emission echo feature groups from the impulse response data packets, and perform cross-matching of the hydrogen emission echo feature groups with three types of conserved residual sequences to obtain the hydrogen energy link anomaly localization results. S5. Generate a hierarchical control instruction set based on the anomaly location results of the hydrogen energy link, execute the corresponding control actions, determine the anomaly handling results based on the post-control verification data packet, and update the normal operation benchmark library and the active detection reference library.

[0019] S1 specifically includes the following sub-steps: S110. Establish a unified terminology and synchronously collect raw data from the hydrogen energy link: The proton exchange membrane electrolyzer, hydrogen purification and drying unit, buffer hydrogen storage cylinder, hydrogen pipeline, proportional valve, hydrogen inlet solenoid valve, fuel cell, and fuel cell exhaust valve, connected sequentially, are collectively defined as the hydrogen energy link. The local computing program deployed in the central control unit for performing data alignment, data validity processing, parameter calculation, feature extraction, and anomaly localization operations is collectively defined as the edge computing module. The data interval that the edge computing module extracts according to a preset window duration and moves forward according to a preset update step size is collectively defined as the rolling diagnostic window. The central control unit, acting as a unified clock source, collects raw data from the hydrogen energy link according to a basic sampling period of 100ms.

[0020] The operating current, electrolysis chamber voltage, operating temperature, and operating status of the proton exchange membrane electrolyzer are uploaded by the hydrogen production controller; the makeup water flow rate is collected by a makeup water flow sensor installed on the external deionized water makeup pipeline; the deionized water tank level is collected by a level sensor installed inside the deionized water tank; the inlet deionized water conductivity is collected by a conductivity sensor installed on the proton exchange membrane electrolyzer inlet pipeline; the hydrogen outlet flow rate is collected by a hydrogen outlet flow sensor installed downstream of the hydrogen purification and drying unit; the oxygen outlet flow rate is collected by an oxygen outlet flow sensor installed downstream of the oxygen-side gas-liquid separation unit; the corresponding absolute gas pressure and absolute gas temperature are collected by pressure sensors and temperature sensors installed near each flow sensor.

[0021] Hydrogen storage pressure and temperature are collected by pressure and temperature sensors located in the buffer hydrogen storage tank; hydrogen concentration is collected by hydrogen concentration sensors located near the proton exchange membrane electrolyzer, buffer hydrogen storage tank, and fuel cell; valve status is fed back by valve limit switches or valve actuators; fuel cell hydrogen inlet pressure is collected by a pressure sensor located downstream of the hydrogen inlet solenoid valve and upstream of the fuel cell hydrogen inlet; fuel cell operating status, output current, and single-cell voltage sequence are uploaded by the fuel cell controller; DC bus voltage is collected by a DC bus voltage sensor; energy storage battery state of charge and available discharge power are uploaded by the energy storage battery management system; load power is calculated based on the values ​​collected by voltage and current sensors of the electrical load unit.

[0022] When the energy storage battery management system does not directly output the available discharge power of the energy storage battery, the edge computing module calculates it based on the energy storage battery terminal voltage, allowable discharge current, and bidirectional DC-DC converter efficiency. The central control unit adds timestamps, device numbers, sensor numbers, and sampling validity identifiers to each data point, forming the raw data stream of the hydrogen energy link. For example, if the central control unit has issued a closing command to the hydrogen inlet solenoid valve, but the valve limit switch still indicates an open state, the valve state should be recorded as open, and the target command state should not replace the actual feedback state.

[0023] S120. Perform time alignment and validity processing on the raw data stream of the hydrogen energy link: The edge computing module performs time alignment on the raw data stream of the hydrogen energy link formed in step S110, based on a unified sampling time. For continuous measurements such as operating current, voltage, pressure, temperature, flow rate, conductivity, and load power, linear interpolation is used between two adjacent valid sample values; for discrete state quantities such as valve status and operating status, the most recent valid feedback state before the unified sampling time is used.

[0024] The edge computing module discards data lacking timestamps, data exceeding the sensor's measurement range, data with communication latency exceeding 200ms, and data whose variation exceeds a preset upper limit. The sensor's measurement range is determined according to the corresponding sensor's product manual, and the preset upper limit is determined based on normal operation calibration records. If the same data channel fails to obtain valid data for three consecutive basic sampling cycles, that data channel is marked as unavailable. Data channels marked as unavailable cannot be used for subsequent residual calculations or to trigger active detection.

[0025] The outlet flow rates of hydrogen and oxygen are converted to standard state gas flow rates using the following formula: In the formula, The standard state gas flow rate at sampling time t is expressed in L / min. The actual state gas flow rate at sampling time t, in L / min; This represents the absolute pressure of the gas, expressed in kPa. This is the absolute temperature of the gas, expressed in Kelvin (K). The absolute pressure under standard conditions is taken as 101.325 kPa; The absolute temperature is 273.15 K; g indicates the type of gas, either hydrogen or oxygen.

[0026] If the flow sensor directly outputs the standard state gas flow rate, the conversion will not be repeated. For example, if the hydrogen outlet flow rate is uploaded every 500ms and the fuel cell hydrogen inlet pressure is uploaded every 100ms, time alignment should be completed before proceeding to subsequent calculations; data from different sampling times should not be directly compared. After processing, a valid data sequence for the hydrogen energy link is formed.

[0027] S130. Generate hydrogen energy link diagnostic data packets according to the rolling diagnostic window: The edge computing module continuously extracts rolling diagnostic windows from the valid data sequence of the hydrogen energy link formed in step S120, with a window duration of 10 seconds and an update step size of 1 second. Each rolling diagnostic window has a unique window number, a window start time, and a window end time.

[0028] Actual water consumption is calculated using the following formula: In the formula, This represents the actual water consumption within the current scrolling diagnostic window, in units of mol. This is the density of deionized water, expressed in g / L. The total volume of replenished water obtained within the current scrolling diagnostic window, in liters (L). and These are the window start times. and window end time The corresponding deionized water tank storage volume, in liters (L); Let be the molar mass of water, taken as 18.015 g / mol.

[0029] The water volume of the deionized water tank is obtained based on the liquid level data and the pre-calibrated correspondence between liquid level and water volume. For example, if the replenishment water volume is 0.05L, when the water volume of the deionized water tank decreases from 2.00L to 1.80L, the actual water consumption volume is 0.25L, not 0.05L.

[0030] Actual hydrogen production and actual oxygen production are calculated using the following formula: In the formula, This represents the actual gas production, expressed in mol. The standard molar volume of the gas is taken as 22.414 L / mol. The flow rate is the standard state gas flow rate, in L / min; the sampling time is uniformly converted to min during integration; g is used to obtain the actual hydrogen production when hydrogen is sampled, and g is used to obtain the actual oxygen production when oxygen is sampled. and These are the start and end times of the aforementioned scrolling diagnostic window, respectively.

[0031] The voltage dispersion of a single chip is calculated using the following formula: In the formula, The voltage dispersion of a single cell is expressed in V; N is the number of single cells in the fuel cell. Let V be the voltage of the j-th individual cell, in volts (V). is the average voltage of a single cell, in V; j is the cell number.

[0032] The operating current sequence, electrolysis chamber voltage sequence, operating temperature sequence, inlet deionized water conductivity sequence, actual water consumption, actual hydrogen production, actual oxygen production, hydrogen storage pressure sequence, hydrogen storage temperature sequence, hydrogen concentration sequence, valve status sequence, fuel cell hydrogen inlet pressure sequence, fuel cell output current sequence, single-cell voltage sequence, single-cell voltage dispersion sequence, DC bus voltage sequence, energy storage battery state of charge sequence, energy storage battery available discharge power sequence, load power sequence, and valid identifiers of each data channel are combined to form a hydrogen energy link diagnostic data package. This hydrogen energy link diagnostic data package serves as the common input for steps S210, S220, and S310.

[0033] S2 specifically includes the following sub-steps: S210, Generate theoretical material quantity set: The theoretical water consumption, theoretical hydrogen production, and theoretical oxygen production calculated based on the proton exchange membrane electrolyzer operating current sequence within the current rolling diagnostic window are uniformly defined as the theoretical material quantity group; the water loss caused by circulating water evaporation, liquid discharge, and gas-liquid entrainment is uniformly defined as the non-electrolysis water consumption benchmark quantity; and the data set formed during the normal operation calibration stage and updated only by the data verified in step S530 is uniformly defined as the normal operation benchmark library.

[0034] The edge computing module reads the operating current sequence, operating temperature sequence, inlet deionized water conductivity sequence, and actual water consumption from the hydrogen energy link diagnostic data package generated in step S130, and integrates the operating current sequence: In the formula, Q represents the charge amount corresponding to the current scrolling diagnostic window, in C. The current of the proton exchange membrane electrolyzer at sampling time t is expressed in amperes (A). and These represent the start and end times of the scrolling diagnostic window, respectively.

[0035] A normal operating baseline library was established before the system was put into practical training. Under the conditions of no leakage, no water shortage, no abnormal oxygen side discharge, and normal sensor operation, the proton exchange membrane electrolyzer was controlled to operate within the preset operating current range, operating temperature range, and inlet deionized water conductivity range, and no less than 30 rolling diagnostic windows were collected under each calibration condition; the Faraday efficiency correction coefficient was obtained based on the corresponding charge and actual hydrogen production, and the normal fluctuation range of non-electrolysis water consumption baseline and three types of conservation residuals was recorded.

[0036] The edge computing module reads the Faraday efficiency correction coefficients from the normal operating benchmark library based on the current operating current range, operating temperature range, and inlet deionized water conductivity range. When the current operating conditions are between two adjacent calibration ranges, linear interpolation is used to determine the corresponding coefficients; when the current operating conditions exceed the coverage of the normal operating benchmark library, the current rolling diagnostic window is marked as an ineffective residual window.

[0037] The theoretical material quantity set is calculated according to the following formula: In the formula, The theoretical hydrogen production is expressed in mol. Theoretical oxygen production, in mol. This represents the theoretical water consumption, expressed in mol. is the Faraday efficiency correction factor; F is the Faraday constant, taken as 96485 C / mol.

[0038] For example, within a 10-second rolling diagnostic window, if the operating current is 20A for the first 5 seconds and 30A for the next 5 seconds, the entire operating current sequence should be integrated; the 30A value at the end of the window should not be used for calculation. The theoretical material quantity set should be used as the input for step S220.

[0039] S220. Generate three types of conserved residual sequences: The rolling diagnostic window where all input data channels are valid and the proton exchange membrane electrolyzer is in a stable operating state is uniformly defined as the effective residual window; the residual fluctuation range obtained under the same operating conditions during the normal operation calibration phase is uniformly defined as the residual allowable range; and the data sequence formed by arranging the water-side conserved residual, hydrogen-side conserved residual, and oxygen-side conserved residual in the order of the rolling diagnostic window is uniformly defined as the three types of conserved residual sequences.

[0040] The edge computing module reads the non-electrolysis water consumption baseline from the normal operation baseline library and corrects the actual water consumption obtained in step S130: In the formula, To correct for actual water consumption, the unit is mol; The actual water consumption obtained in step S130 is expressed in mol. This is the baseline amount of water consumed without electrolysis, in mol. For example, if the actual water consumption is 0.060 mol and the baseline amount of water consumed without electrolysis is 0.010 mol, the corrected actual water consumption is 0.050 mol.

[0041] The edge computing module will correct the actual water consumption, actual hydrogen production, and actual oxygen production by comparing them with the theoretical material quantities: In the formula, The conservation residual is the material type x; x is water, hydrogen, or oxygen. This corresponds to the theoretical material quantity, in mol. The actual amount of material used for comparison is expressed in mol. A lower limit for positive numbers is set to prevent the denominator from being zero.

[0042] The water side uses the corrected actual water consumption, the hydrogen side uses the actual hydrogen production, and the oxygen side uses the actual oxygen production. The water-side conservation residual is denoted as... The hydrogen-side conservation residual is denoted as The oxygen-side conservation residual is denoted as The edge computing module saves the window number, three types of conserved residuals, the allowable range of residuals, and the valid identifier of the data channel according to the rolling diagnostic window sequence, forming a sequence of three types of conserved residuals. Rolling diagnostic windows that include start-up, stop, fault reset times, or unavailable data channels are not considered valid residual windows. For example, when the gas flow rate is not yet stable after the proton exchange membrane electrolyzer starts up, the hydrogen-side conserved residual may be temporarily low; this window should not be used for continuous over-limit judgment. The sequence of three types of conserved residuals is used as input for step S230.

[0043] S230. Generate an active detection request based on the continuous exceedance situation: The data formed by combining the residual type, deviation direction, maximum deviation magnitude, number of consecutive out-of-limit windows, and duration of consecutive out-of-limit events is uniformly defined as residual trigger information; the data formed by combining residual trigger information, request number, generation time, and valid data channel identifier is uniformly defined as active detection request.

[0044] The edge computing module reads the three types of conserved residual sequences formed in step S220 in chronological order. When any type of conserved residual exceeds the allowable range of the corresponding residual within three consecutive valid residual windows, the conserved residual of that type is deemed to meet the continuous exceedance condition. If there are invalid residual windows between two adjacent valid residual windows, the number of consecutive exceedance windows is recounted.

[0045] If a single valid residual window exceeds its limit, and subsequent valid residual windows recover to the corresponding allowable residual range, only the corresponding residual record is retained, and no active detection request is generated. An active detection request must include at least the request number, generation time, type of residual continuously exceeding the limit, deviation direction, maximum deviation magnitude, number of consecutive exceeding windows, duration of consecutive exceeding, corresponding valid residual window number, valid identifiers for each data channel, proton exchange membrane electrolyzer operating status, hydrogen storage pressure, hydrogen concentration, fuel cell operating status, and request execution status.

[0046] The request execution status includes pending judgment, prohibited active detection, and permitted entry into safety condition judgment. When the hydrogen concentration reaches the emergency interlock threshold, there is an unavailable data channel, or the central control unit has executed an emergency shutdown command, the request execution status is set to prohibited active detection; otherwise, the request execution status is set to permitted entry into safety condition judgment, and the active detection request is sent to step S310.

[0047] For example, if the hydrogen-side conservation residual has been below the corresponding allowable residual range for three consecutive valid residual windows, but the hydrogen concentration has reached the emergency interlock threshold, the short-pulse opening of the fuel cell hydrogen discharge valve must not continue; the emergency interlock process should proceed directly. The active detection request serves as the input for step S310 to determine the active detection safety conditions and is not directly used as the anomaly location result.

[0048] S3 specifically includes the following sub-steps: S310. Determine the active detection window: The set of parameters used to limit the short-pulse opening range of the fuel cell hydrogen exhaust valve is uniformly defined as the active detection safety constraint group; the time interval for the active detection request to meet safety conditions is uniformly defined as the active detection window; and the hydrogen concentration warning threshold, the allowable range of hydrogen storage pressure, the minimum value of fuel cell hydrogen inlet pressure, the allowable range of DC bus voltage, the minimum value of energy storage battery state of charge, the allowable short-term power drop of fuel cell, and the reserved power stored in the central control unit are uniformly defined as the safe operation parameter table.

[0049] The edge computing module reads the active detection request generated in step S230, and reads the hydrogen concentration, hydrogen storage pressure, fuel cell hydrogen inlet pressure, DC bus voltage, energy storage battery state of charge, energy storage battery available discharge power, fuel cell output current, single-cell voltage dispersion and valid identifier of each data channel from the hydrogen energy link diagnostic data packet generated in step S130.

[0050] The active detection power margin is calculated using the following formula: In the formula, For active detection of power margin, the unit is W; This refers to the usable discharge power of the energy storage battery, measured in W. This refers to the allowable short-term power drop of the fuel cell, expressed in watts (W). Reserved power, unit is W.

[0051] An active detection window is established when all data channels are valid, the hydrogen concentration is below the warning threshold, the hydrogen storage pressure and DC bus voltage are within the allowable range, the fuel cell hydrogen inlet pressure and energy storage battery state of charge are not lower than the corresponding minimum values, the active detection power margin is not less than 0, and the fuel cell has not undergone any start-up, shutdown, fault reset, or routine hydrogen discharge actions within the first 5 seconds. The active detection safety constraint group includes at least the single allowable activation duration range, the short pulse interval range, the upper limit of the number of short pulse activations, the response acquisition duration range, and the allowable duration of active detection.

[0052] For example, if the available discharge power of the energy storage battery is 180W, the allowable short-term power drop of the fuel cell is 120W, and the reserved power is 30W, the active detection power margin is 30W; when the available discharge power of the energy storage battery drops to 130W, the active detection power margin is -20W, and active detection should not be performed. The active detection window serves as the input for step S320.

[0053] S320. Determine the optimal active detection parameter set using the bat algorithm: The calibration data set generated before the system was put into operation under normal and safe and controllable fault simulation conditions is uniformly defined as the active detection reference library; the data consisting of the single opening duration of the fuel cell hydrogen exhaust valve, the short pulse interval duration, the number of short pulse openings, and the response acquisition duration is uniformly defined as the candidate active detection parameter set; and the candidate active detection parameter set obtained by the bat algorithm is uniformly defined as the optimal active detection parameter set.

[0054] The active detection reference library stores calibration records according to hydrogen storage pressure range, fuel cell hydrogen inlet pressure range, fuel cell output current range, and single-cell voltage dispersion range. Each calibration record includes the corresponding parameter set, hydrogen inlet pressure response, hydrogen storage pressure disturbance, DC bus voltage disturbance, single-cell voltage dispersion response, and additional hydrogen consumption.

[0055] The candidate active detection parameter set corresponding to the i-th virtual bat is represented as follows: In the formula, the parameter vector For the i-th virtual bat, there is a candidate active detection parameter set; Duration of a single activation, in milliseconds; The duration of the short pulse interval is expressed in milliseconds (ms). This refers to the number of times the short pulse is activated. The response time is measured in milliseconds (ms).

[0056] The edge computing module generates multiple candidate active detection parameter groups within the active detection security constraint group, and calls the calibration record matching the current state from the active detection reference library to calculate the fitness according to the following formula: In the formula, The fitness of the i-th candidate active detection parameter set; For pressure response discrimination; The distinguishability of the voltage dispersion response of a single chip; This represents the hydrogen storage pressure disturbance. This represents the DC bus voltage disturbance. This is for additional hydrogen consumption; to Each value has a positive weight and a sum of 1. All evaluation indicators are normalized to between 0 and 1.

[0057] The bat algorithm updates the candidate active detection parameter set according to the following formula: In the formula, Let be the search frequency corresponding to the i-th virtual bat; and These are the lower limit and upper limit of search frequency, respectively; A random number between 0 and 1; velocity vector The search speed at the k-th iteration; parameter vector This represents the candidate active probe parameter set for the k-th iteration; parameter vector. This is the candidate active detection parameter set with the highest fitness in the k-th iteration.

[0058] Boundary correction is performed after each update, and the number of short pulse activations is rounded down. When the maximum number of iterations is reached or the fitness is not improved after a preset number of iterations, the optimal active detection parameter set is output. If the active detection reference library lacks matching records, a preset safe initial parameter set is used to complete low-disturbance detection; long-duration, multiple short pulses must not be used directly. The optimal active detection parameter set is used as input for step S330.

[0059] In a specific practical training scenario, the parameters and weights of the Bat Algorithm's fitness function are set as follows: The positive weights of the fitness evaluation index are set as follows: =0.35, =0.25, =0.15, =0.15, =0.10 (sum of 1); Lower bound of search frequency for the bat algorithm Set to 0, maximum search frequency. Set to version 2.0, and the maximum number of iterations is set to 50.

[0060] S330, Execute short pulse activation and generate pulse response data packet: The synchronously collected data before, during, and after the fuel cell hydrogen exhaust valve opens, along with the active detection request number, optimal active detection parameter set, actual action feedback, and early termination flag, are uniformly defined as a pulse response data packet. The central control unit sends a short pulse opening command to the fuel cell controller according to the optimal active detection parameter set formed in step S320. The fuel cell controller drives the fuel cell hydrogen exhaust valve to operate and feeds back the actual opening and closing times to the central control unit.

[0061] Before the first short pulse is activated, the edge computing module sets a 1-second baseline acquisition period and continuously acquires the fuel cell hydrogen inlet pressure, hydrogen storage pressure, fuel cell output current, single-cell voltage sequence, single-cell voltage dispersion, DC bus voltage, energy storage battery state of charge, hydrogen concentration, and fuel cell hydrogen exhaust valve status according to a 20ms sampling cycle.

[0062] The change in response is calculated using the following formula: In the formula, This represents the change in response at sampling time t. This refers to the response data corresponding to sampling time t; This represents the average response data during the baseline acquisition period before the short pulse of the hydrogen discharge valve is activated.

[0063] During active detection, the energy storage battery provides short-term power compensation. When the hydrogen concentration reaches the warning threshold, the fuel cell hydrogen inlet pressure or hydrogen storage pressure is lower than the minimum value, the DC bus voltage is lower than the minimum value, the energy storage battery state of charge is lower than the minimum value, the active detection power margin is less than 0, communication is interrupted, or the feedback status of the fuel cell hydrogen discharge valve is inconsistent with the command, the central control unit immediately terminates the remaining short pulses and closes the fuel cell hydrogen discharge valve.

[0064] For example, if the central control unit has sent a shutdown command, but the valve limit switch still indicates an open state, the abnormal operation of the fuel cell hydrogen exhaust valve should be recorded, and further detection should be prohibited. The pulse response data packet should include at least the planned time, the actual time, the original sequence of each response data, the sequence of response changes, the valid data channel identifier, the early termination identifier, and the reason for early termination. The pulse response data packet serves as the input for step S410.

[0065] S4 specifically includes the following sub-steps: S410, Extraction of hydrogen emission echo characteristic group: The average values ​​of fuel cell hydrogen inlet pressure, hydrogen storage pressure, fuel cell output current, single-cell voltage dispersion, and DC bus voltage during the baseline acquisition period before the short pulse opening of the fuel cell hydrogen exhaust valve are uniformly defined as the hydrogen exhaust echo baseline. The data set formed by quantifying the pressure response, current response, and single-cell voltage dispersion response caused by the short pulse opening of the fuel cell hydrogen exhaust valve is uniformly defined as the hydrogen exhaust echo feature group. The edge computing module reads the pulse response data packet generated in step S330 and segments the data according to the actual opening and closing times of the fuel cell hydrogen exhaust valve, without substituting the planned time for the actual action time.

[0066] The pressure drop at the hydrogen inlet is calculated using the following formula: In the formula, This represents the decrease in hydrogen inlet pressure, expressed in kPa. The average hydrogen inlet pressure in the hydrogen exhaust echo baseline is expressed in kPa. The hydrogen inlet pressure of the fuel cell at sampling time t is expressed in kPa. This represents the actual opening range of the hydrogen exhaust valve in the fuel cell.

[0067] The hydrogen inlet pressure recovery time is calculated using the following formula: In the formula, The time for hydrogen inlet pressure to recover is expressed in milliseconds (ms). This is the actual closing time of the fuel cell hydrogen exhaust valve; The allowable deviation for pressure recovery read from the active probe reference library, in kPa. and The definition is the same as above. If the data is not recovered by the end of the recovery acquisition period, an incomplete recovery marker is recorded.

[0068] The improvement ratio of single-chip voltage dispersion is calculated according to the following formula: In the formula, To improve the voltage dispersion of a single chip by a certain percentage; The average value of the single-chip voltage dispersion in the hydrogen emission echo baseline, in V; The voltage dispersion of a single chip at sampling time t, expressed in V; This is the recovery data acquisition interval after the fuel cell hydrogen exhaust valve is closed; The lower limit of positive numbers defined in step S220 is used.

[0069] The edge computing module synchronously calculates the hydrogen storage pressure disturbance amplitude, fuel cell output current deviation, DC bus voltage disturbance amplitude, consistency of adjacent short pulse responses, invalid short pulse identifiers, and reasons for premature termination. For example, if the central control unit plans to open the fuel cell hydrogen discharge valve at 1000ms, but the valve limit switch only responds with an open status at 1080ms, the pressure response should be extracted starting from 1080ms to avoid misjudging the valve delay as a hydrogen pipeline malfunction. All features are combined to form a hydrogen discharge echo feature set, which serves as the input for step S420.

[0070] S420. Generate a candidate anomaly localization result set: The template set consisting of three types of conserved residual features and hydrogen emission echo features labeled according to anomaly type in the active detection reference library is uniformly defined as the anomaly feature template set; the data set consisting of candidate anomaly types, candidate anomaly links, anomaly location confidence, and supporting evidence is uniformly defined as the candidate anomaly location result set. The edge computing module reads the three types of conserved residual sequences formed in step S220, the residual triggering information formed in step S230, the hydrogen emission echo feature group formed in step S410, and the active detection reference library established in step S320.

[0071] The edge computing module first screens candidate abnormal links based on three types of conservation residuals: when the water-side conservation residual, hydrogen-side conservation residual, and oxygen-side conservation residual all exceed the corresponding residual allowable range, the proton exchange membrane electrolyzer is listed as a candidate abnormal link; when the hydrogen-side conservation residual is consistently low, while the water-side and oxygen-side conservation residuals are within the corresponding residual allowable range, the hydrogen storage buffer and safety management unit or hydrogen pipeline is listed as a candidate abnormal link; when all three types of conservation residuals are within the corresponding residual allowable range, but the hydrogen emission echo characteristics are abnormal, the fuel cell is listed as a candidate abnormal link.

[0072] After completing the candidate anomaly screening, the corresponding template is retrieved from the anomaly feature template set, and the anomaly localization reliability corresponding to the candidate anomaly type is calculated according to the following formula: In the formula, represents the anomaly localization confidence level corresponding to candidate anomaly type a; L represents the number of features involved in the matching. For the current number Observational characteristics; For the candidate exception type a, the first Item template center value; For the candidate exception type a, the first The feature allows for fluctuation scale; For the first Each feature has a weight, and all weights are positive numbers with a sum of 1. The lower limit of positive numbers defined in step S220 is used.

[0073] The abnormal feature template set is established by normal calibration and fault simulation within the safety boundary before the system is put into operation. It includes at least templates for insufficient water replenishment, poor oxygen side discharge, minor leakage in hydrogen pipeline, sluggish action of hydrogen inlet solenoid valve, partial blockage in hydrogen pipeline, tendency of fuel cell anode to accumulate water, and sensor drift.

[0074] Taking the "minor leak in hydrogen pipeline" anomaly feature template as an example, its parameter structure and reference center value in the template set are shown below: Template center value of the first matching feature (deviation of hydrogen-side conserved residual). Set to -5%, feature allowable fluctuation scale s to 2%, weight Set to 0.4; the template center value of the second matching feature (prolongation of hydrogen inlet pressure recovery time). Set to +200ms, feature allowable fluctuation scale s is set to 50ms, weight Set it to 0.6.

[0075] For example, if the hydrogen-side conservation residual is consistently low, while the water-side and oxygen-side conservation residuals are both within their respective allowable ranges, and the hydrogen inlet pressure recovery time is prolonged while valve feedback is normal, the matching degree of the hydrogen pipeline micro-leakage template should be improved; leakage should not be directly determined solely based on the hydrogen-side conservation residual. In cases where the single-chip voltage dispersion does not improve after the fuel cell hydrogen exhaust valve short-pulse opening, only an abnormal fuel cell gas-water state should be output, without directly determining membrane dryness. The candidate anomaly types and their anomaly location confidence scores are combined to form a candidate anomaly location result set, which serves as the input for step S430.

[0076] S430. Determine the anomaly location result of the hydrogen energy link: The difference between the highest anomaly location reliability and the second highest anomaly location reliability is uniformly defined as the location reliability error value; the data fed back to step S320 when the location result is unclear is uniformly defined as a retest request. The edge computing module compares the location reliability error value with the preset reliability error value pre-stored in the active detection reference library.

[0077] When the location confidence level reaches the preset confidence level, the candidate anomaly type corresponding to the highest anomaly location confidence level is determined as the hydrogen energy link anomaly location result; when the location confidence level does not reach the preset confidence level, a retest request is generated. The retest request includes at least the previous optimal active detection parameter set, the candidate anomaly type corresponding to the highest anomaly location confidence level, the candidate anomaly type corresponding to the second highest anomaly location confidence level, indistinguishable features, the number of retests, and the parameter adjustment direction.

[0078] The edge computing module sends the retest request back to step S310 to re-determine whether an active detection window exists. If an active detection window is confirmed, step S320 generates a retest active detection parameter set different from the previous optimal active detection parameter set using the bat algorithm. The number of retests does not exceed two. If the maximum number of retests is reached and the anomaly still cannot be distinguished, the anomaly type is marked as requiring manual confirmation, and a protective localization result is output based on the candidate anomaly type with the highest risk level in the candidate anomaly localization result set.

[0079] For example, minor leaks in the hydrogen pipeline and delayed operation of the hydrogen inlet solenoid valve can both prolong the hydrogen inlet pressure recovery time. In such cases, the recovery acquisition time can be extended, and feedback from the valve limit switch can be used for further differentiation. The anomaly location results of the hydrogen energy link should include at least the anomaly link, anomaly type, risk level, anomaly location reliability, supporting evidence, number of retests, a flag indicating pending manual confirmation, a flag prohibiting further active detection, and a suggested control action type, and serve as input for step S510.

[0080] S5 specifically includes the following sub-steps: S510. Generate a hierarchical control command set based on the anomaly location results of the hydrogen energy link: The data table used to record abnormal links, abnormal types, risk levels, control actions, control action execution order, control target values, control stabilization waiting time, and recovery conditions is uniformly defined as a hierarchical control rule table; the state in which the central control unit suspends the execution of user-defined energy management strategy instructions that conflict with abnormal handling, and the corresponding equipment is prioritized for control by safety control instructions is uniformly defined as a safety control takeover state; and all control instructions generated for the same abnormal location result and their execution order are uniformly defined as a hierarchical control instruction set.

[0081] The edge computing module reads the hydrogen energy link anomaly location results formed in step S430, reads the water-side conservation residual, hydrogen-side conservation residual and oxygen-side conservation residual from the three types of conservation residual sequences formed in step S220, and reads the hydrogen concentration, hydrogen storage pressure, fuel cell hydrogen inlet pressure, DC bus voltage and energy storage battery state of charge from the hydrogen energy link diagnostic data package formed in step S130.

[0082] For the h-th safe operating data item, the degree of safety deviation is calculated according to the following formula: In the formula, This represents the degree of safety deviation corresponding to the h-th safe operation data item; This is the current safe operating data; For safety reference values; This is the safety boundary value; This is used as a hazard indicator; a value of 1 is used when an increase in data leads to an increase in risk, and a value of -1 is used when a decrease in data leads to an increase in risk. The lower limit of positive numbers defined in step S220 is used.

[0083] The edge computing module calculates the control risk value according to the following formula: In the formula, To control risk values; H represents the inherent risk coefficient corresponding to anomaly type a; H is the number of safe operation data involved in risk assessment. The weight corresponding to the degree of safety deviation of the h-th item; To determine the weights for uncertainty; The anomaly localization confidence level corresponding to the anomaly type a finally determined in step S430; The definition of safety deviation is the same as above.

[0084] In a specific training control embodiment, the preset values ​​of various risk parameters are as follows: Taking the "minor leak in the hydrogen pipeline" anomaly as an example, its corresponding inherent risk coefficient is... Set to 0.4; weights corresponding to each degree of safety deviation. The uniform average is set to 0.1; the positioning uncertainty weight is... Set to 0.2.

[0085] The graded control rule table classifies risk levels into four levels—Level 1, Level 2, Level 3, and Level 4—based on control risk values. Level 1 risk corresponds to parameter correction and enhanced monitoring; Level 2 risk corresponds to operational restrictions; Level 3 risk corresponds to stopping abnormal equipment and isolating abnormal branches; and Level 4 risk corresponds to stopping hydrogen-related equipment, closing the hydrogen inlet solenoid valve, starting the exhaust system, and performing a safe venting. When the hydrogen concentration reaches the emergency interlock threshold, the hydrogen storage pressure reaches the allowable upper limit, the fuel cell hydrogen exhaust valve cannot be closed, or critical data channel communication is interrupted, there is no need to calculate the control risk value; the risk is directly determined as Level 4.

[0086] For situations requiring power reduction, the target control value is determined using the following formula: In the formula, The target operating current of the proton exchange membrane electrolyzer is expressed in amperes (A). This is the current operating current of the proton exchange membrane electrolyzer, in amperes (A). The percentage reduction in hydrogen production current corresponding to risk level r; The target output power of the fuel cell, expressed in watts (W). This represents the current output power of the fuel cell, in watts (W). The percentage reduction in fuel cell output power corresponding to risk level r.

[0087] Under Level 4 risk, the target operating current of the proton exchange membrane electrolyzer and the target output power of the fuel cell are both set to 0. For example, if a minor leak in the hydrogen pipeline has been located, but the user-defined energy management strategy still requires increasing the hydrogen production current, the central control unit enters a safety control takeover state, blocking the command to increase the hydrogen production current and prioritizing the execution of stopping hydrogen production and isolating abnormal branches. The control actions, target equipment, control target values, execution sequence, and maximum execution duration are combined to form a hierarchical control instruction set, which serves as the input for step S520.

[0088] S520: Execute the hierarchical control instruction set and generate a control post-verification data packet: The data formed by combining the control commands issued by the central control unit, the feedback received by the equipment controller, and the actual status feedback of the equipment is uniformly defined as the control command confirmation record; the rolling diagnostic window continuously captured after the hierarchical control commands have been executed and after the control stabilization waiting period is uniformly defined as the post-control verification window; the data formed by combining the control command confirmation record, the running data corresponding to the post-control verification window, and the verification impulse response data obtained when execution is allowed is uniformly defined as the post-control verification data packet.

[0089] After the central control unit enters the safety control takeover state, it issues control commands in the following order: stopping or reducing power, closing valves, isolating abnormal branches, starting the exhaust system, and performing a safety venting. The actual execution result of the proton exchange membrane electrolyzer's operating current is fed back by the hydrogen production controller; the valve closing result is fed back by the valve limit switch or valve actuator; the fuel cell output power is fed back by the fuel cell controller; the exhaust system's operating status is confirmed by the exhaust system's operating feedback signal; and the safety venting execution result is jointly confirmed by the safety venting valve limit feedback and the change in hydrogen storage pressure.

[0090] Each control command confirmation record must include at least the control command number, the corresponding anomaly location result number, the target device, the control action, the control target value, the command issuance time, the maximum execution duration, the actual status feedback, the execution success flag, and the execution failure reason. If the actual device status feedback is inconsistent with the control target, or if the control target is not achieved even after the maximum execution duration, the execution success flag will be set to failure, and the risk level will be increased.

[0091] For example, if the central control unit has issued a command to close the hydrogen inlet solenoid valve, but the valve limit switch still indicates that it is open, it should not be recorded as the valve being closed; instead, a Level 4 risk response should be executed directly. For Level 1 and Level 2 risks, after the control action is completed, a post-control verification window should be established according to the control stability waiting time in the hierarchical control rule table. For Level 3 risks, a low-disturbance verification short pulse is permitted only when the abnormal branch has been isolated, the hydrogen concentration is below the warning threshold, and the active detection conditions specified in step S310 are met again. For Level 4 risks, verification short pulses are not permitted; only safe operation data should be continuously collected, and manual reset should be awaited.

[0092] The post-control verification data package includes at least the hierarchical control instruction set, control instruction confirmation record, operating current sequence, actual water consumption, actual hydrogen production, actual oxygen production, hydrogen storage pressure sequence, hydrogen storage temperature sequence, hydrogen concentration sequence, valve status sequence, fuel cell hydrogen inlet pressure sequence, fuel cell output current sequence, single-cell voltage dispersion sequence, DC bus voltage sequence, energy storage battery state of charge sequence, data channel valid identifier, and verification impulse response data, and serves as the input for step S530.

[0093] S530. Verify the anomaly handling results and update the normal operation benchmark library and the active detection reference library: The degree to which the three types of conservation residuals and hydrogen emission echo characteristic groups deviate from the normal range after control is uniformly defined as the verification deviation degree; the state in which the abnormality handling is completed, the verification deviation degree meets the recovery conditions, and the safe operation data is within the allowable range is uniformly defined as the recovery allowable state.

[0094] The edge computing module reads the control post-verification data packet formed in step S520, and re-executes steps S210 and S220 to obtain the water-side conservation residual, hydrogen-side conservation residual, and oxygen-side conservation residual after control. When the control post-verification data packet contains verification impulse response data, step S410 is re-executed to obtain the hydrogen emission echo characteristic group after control.

[0095] The deviation is calculated using the following formula: In the formula, To verify the degree of deviation; , and These are the water-side conservation residuals, hydrogen-side conservation residuals, and oxygen-side conservation residuals after control. , and These are the absolute values ​​of the corresponding residual allowable range boundaries in the normal operating benchmark library; The normalized distance between the controlled hydrogen emission echo characteristic set and the normal state template; To actively detect the upper limit of distance corresponding to the normal state in the reference library.

[0096] If the verification deviation of the verification window is no greater than 1 after three consecutive control cycles, and the hydrogen concentration, hydrogen storage pressure, DC bus voltage, and energy storage battery state of charge are all within the allowable range, the anomaly is determined to be eliminated and the system enters the recovery allowable state; if the verification deviation decreases but is still greater than 1, the system returns to step S510 to increase the control level or extend the execution time of the control action; if the verification deviation does not decrease or continues to increase, level 4 risk handling is performed.

[0097] For example, if the pre-control verification deviation is 2.0, it decreases to 1.3 after control is implemented. Although the anomaly has been improved, operation must still not be resumed. After entering the recovery permission state, the central control unit maintains the isolation of the abnormal branch and allows the proton exchange membrane electrolyzer to start at low power. After continuous verification is passed, the operating current of the proton exchange membrane electrolyzer and the output power of the fuel cell are restored step by step, and finally the safety control takeover state is released.

[0098] Only data from areas lacking fault simulation injection, unavailable data channels, pending manual confirmation, and manually enforced control are written to the normal operation baseline library; verified hydrogen emission echo characteristic groups are written to the normal state partition of the active detection reference library; fault simulation data are written only to the corresponding fault simulation state partition of the active detection reference library. The updated normal operation baseline library and active detection reference library serve as inputs for the next control cycle.

[0099] Example 2: Figure 2As shown, this embodiment provides an integrated training system for wind-solar hybrid hydrogen production and fuel cell power generation, including: The system includes a wind-solar hybrid power generation unit, a DC bus, an energy storage battery unit, a proton exchange membrane electrolyzer, a hydrogen storage buffer and safety management unit, a fuel cell power generation unit, an electrical load unit, and a central control unit; the wind-solar hybrid power generation unit, the energy storage battery unit, the proton exchange membrane electrolyzer, the fuel cell power generation unit, and the electrical load unit are all connected to the DC bus. Specifically, the power load unit may include an electric vehicle charging pile simulation module and an electric vehicle battery simulation load connected thereto, which is used to realize the comprehensive power supply training of the wind-solar-hydrogen storage system to electric vehicles in the off-grid state of the microgrid.

[0100] The hydrogen storage buffer and safety management unit includes a hydrogen purification and drying device, a buffer hydrogen storage cylinder, hydrogen pipelines, a proportional valve, a hydrogen inlet solenoid valve, a pressure sensor, a temperature sensor, a hydrogen concentration sensor, branch valves, and a safety vent valve; the proton exchange membrane electrolyzer, hydrogen purification and drying device, buffer hydrogen storage cylinder, hydrogen pipelines, proportional valve, hydrogen inlet solenoid valve, fuel cell, and fuel cell vent valve are connected in sequence to form a hydrogen energy link; The central control unit includes an edge computing module, a data acquisition and processing module, a conserved residual generation module, an active detection and control module, an anomaly localization module, and a hierarchical control verification module; The data acquisition and processing module is used to synchronously acquire the operation data of the hydrogen energy link, perform time alignment, validity processing and rolling diagnostic window division on the operation data, and generate hydrogen energy link diagnostic data packages. The conservation residual generation module is used to calculate the theoretical water consumption, theoretical hydrogen production and theoretical oxygen production based on the operating current sequence of the proton exchange membrane electrolyzer, and compare them with the actual water consumption, actual hydrogen production and actual oxygen production respectively to generate three types of conservation residual sequences. When at least one type of conservation residual continuously exceeds the corresponding residual allowable range, an active detection request is generated. The active detection control module is used to determine the active detection window based on the active detection request and the safe operation parameter table. Within the active detection safety constraint group, the bat algorithm is used to optimize the single opening duration, short pulse interval duration, number of short pulse openings and response acquisition duration of the fuel cell hydrogen exhaust valve to obtain the optimal active detection parameter group. The module then controls the fuel cell hydrogen exhaust valve to perform short pulse opening according to the optimal active detection parameter group, generating a pulse response data packet. The anomaly localization module is used to extract hydrogen emission echo feature groups based on impulse response data packets, and to perform cross-matching of the hydrogen emission echo feature groups with three types of conserved residual sequences to obtain the anomaly localization results of the hydrogen energy link. The hierarchical control verification module is used to generate a hierarchical control instruction set based on the anomaly location results of the hydrogen energy link, control the proton exchange membrane electrolyzer, hydrogen storage buffer and safety management unit and fuel cell power generation unit to perform corresponding control actions, and judge the anomaly handling results based on the post-control verification data package, and update the normal operation benchmark library and active detection reference library when the recovery conditions are met.

[0101] All the above formulas are performed using dimensionless numerical calculations; the relevant formulas are based on empirical models that approximate the real situation, obtained through extensive data collection and software simulation fitting. The preset parameters and thresholds involved in the formulas can be conventionally set and adjusted by those skilled in the art according to the physical constraints of the actual application scenario.

[0102] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0103] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0104] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A control method for an integrated training system of wind-solar hybrid hydrogen production and fuel cell power generation, characterized in that, Includes the following steps: S1. The central control unit synchronously collects hydrogen energy link operation data, and the edge computing module completes time alignment, validity processing and rolling diagnostic window division to generate hydrogen energy link diagnostic data package. S2. Calculate the theoretical water consumption, theoretical hydrogen production and theoretical oxygen production based on the operating current sequence of the proton exchange membrane electrolyzer, and compare them with the actual water consumption, actual hydrogen production and actual oxygen production to generate three types of conservation residual sequences. When at least one type of conservation residual continuously exceeds the allowable range of the corresponding residual, an active detection request is generated. S3. Determine the active detection window based on the active detection request and the safe operation parameter table. Within the active detection safety constraint group, determine the optimal active detection parameter group using the bat algorithm. Control the fuel cell hydrogen exhaust valve to perform short pulse opening and generate a pulse response data packet. S4. Extract hydrogen emission echo feature groups from the impulse response data packets, and perform cross-matching of the hydrogen emission echo feature groups with three types of conserved residual sequences to obtain the anomaly localization results of the hydrogen energy link.

2. The control method for the integrated training system of wind-solar hybrid hydrogen production and fuel cell power generation according to claim 1, characterized in that, Also includes: S5. Generate a hierarchical control instruction set based on the anomaly location results of the hydrogen energy link, execute the corresponding control actions, determine the anomaly handling results based on the post-control verification data packet, and update the normal operation benchmark library and the active detection reference library.

3. The control method for the integrated training system of wind-solar hybrid hydrogen production and fuel cell power generation according to claim 1, characterized in that, S1 specifically includes: The central control unit collects sensor data and controller feedback data from the proton exchange membrane electrolyzer, buffer hydrogen storage cylinder, fuel cell, DC bus, energy storage battery and electrical load at a unified sampling time, and adds timestamps and valid sampling identifiers to form the original data stream of the hydrogen energy link. The edge computing module performs time alignment, abnormal data removal, and gas flow standard state conversion on the raw data stream of the hydrogen energy link to form an effective data sequence of the hydrogen energy link. The actual water consumption, actual hydrogen production, actual oxygen production, and single-chip voltage dispersion are calculated according to the rolling diagnostic window, and the operating data, status data, and effective data channel identifiers are combined to form a hydrogen energy link diagnostic data package.

4. The control method for the integrated training system of wind-solar hybrid hydrogen production and fuel cell power generation according to claim 1, characterized in that, S2 specifically includes: Read the operating current sequence, operating temperature sequence, and inlet deionized water conductivity sequence from the hydrogen energy link diagnostic data package. Integrate the operating current sequence and calculate the theoretical water consumption, theoretical hydrogen production, and theoretical oxygen production based on the Faraday efficiency correction coefficient in the normal operation benchmark library to form a theoretical material quantity group. The actual water consumption is corrected based on the non-electrolysis water consumption benchmark. The corrected actual water consumption, actual hydrogen production, and actual oxygen production are compared with the theoretical material quantity set to form three types of conservation residual sequences.

5. The control method for the integrated training system of wind-solar hybrid hydrogen production and fuel cell power generation according to claim 4, characterized in that, Also includes: When at least one type of conserved residual exceeds the allowable range of the corresponding residual for three consecutive valid residual windows, an active detection request containing residual triggering information and a valid data channel identifier is generated; The effective residual window is a rolling diagnostic window that assumes all input data channels are valid and the proton exchange membrane electrolyzer is in a stable operating state.

6. The control method for the integrated training system of wind-solar hybrid hydrogen production and fuel cell power generation according to claim 1, characterized in that, S3 specifically includes: Read the active detection request and hydrogen energy link diagnostic data packet, and determine the active detection window and active detection safety constraint group based on hydrogen concentration, hydrogen storage pressure, fuel cell hydrogen inlet pressure, DC bus voltage, energy storage battery state of charge and active detection power margin; Within the active detection safety constraint group, the active detection reference library is called, and the Bat Algorithm is used to optimize the single opening duration, short pulse interval duration, number of openings, and response acquisition duration of the fuel cell hydrogen exhaust valve to obtain the optimal active detection parameter group. The fuel cell hydrogen exhaust valve is controlled to open with a short pulse according to the optimal active detection parameter set, and response data is collected simultaneously to form a pulse response data packet.

7. The control method for the integrated training system of wind-solar hybrid hydrogen production and fuel cell power generation according to claim 1, characterized in that, S4 specifically includes: Read the pulse response data packet, segment the data according to the actual opening and closing time of the fuel cell hydrogen exhaust valve, and extract the hydrogen inlet pressure drop amplitude, hydrogen inlet pressure recovery time, hydrogen storage pressure disturbance amplitude, output current deviation and single-chip voltage dispersion improvement ratio to form a hydrogen exhaust echo characteristic group. The hydrogen emission echo feature set is cross-matched with three types of conserved residual sequences, and the anomaly feature template set is called to calculate the anomaly localization confidence of candidate anomaly types, forming a candidate anomaly localization result set. The location confidence difference between the highest and second-highest anomaly location confidence is compared. If the preset confidence difference is reached, the anomaly location result of the hydrogen energy link is determined. If it is not reached, a retest request is generated.

8. The control method for the integrated training system of wind-solar hybrid hydrogen production and fuel cell power generation according to claim 7, characterized in that, Also includes: After generating the retest request, the retest request is fed back to step S3 to re-determine whether there is an active detection window. If it is confirmed that there is an active detection window, the active detection parameter group for retest is generated by the bat algorithm, and the hydrogen exhaust valve of the fuel cell is controlled to re-execute the short pulse opening.

9. The control method for the integrated training system of wind-solar hybrid hydrogen production and fuel cell power generation according to claim 2, characterized in that, S5 specifically includes: Read the anomaly location results of the hydrogen energy link, the three types of conserved residual sequences and the hydrogen energy link diagnostic data package, calculate the control risk value according to the hierarchical control rule table, and generate a hierarchical control instruction set. Upon entering the safety control takeover state, hierarchical control instruction sets are issued in the execution sequence. Control instruction confirmation records are formed based on the actual status feedback of the equipment, and a post-control verification data packet is generated. Based on the control post-verification data package, the three types of conserved residuals are recalculated, and the hydrogen emission echo characteristic group is recalculated when the verification impulse response data is included, to determine the verification deviation.

10. The control method for the integrated training system of wind-solar hybrid hydrogen production and fuel cell power generation according to claim 9, characterized in that, Also includes: Based on the results of the anomaly handling, operations can be restored, the control level can be raised, or a Level 4 risk response can be implemented. When the recovery conditions are met, the normal operation benchmark library and the active detection reference library can be updated.

Images