Hydrogen energy power supply integrated control system

Abstract

The application discloses a hydrogen energy power supply integrated control system and relates to the technical field of hydrogen energy power supply control. The system collects working condition characteristic signals in a cold start stage, identifies an exhaust action interval, a load climbing section and a gentle section in an initial green light window, and constructs an exhaust load timing evidence object sequence. A maximum risk set is screened based on a dictionary order threshold and a dominance relationship, and it is determined whether to enter an intervention mode. An execution token of a limited takeover mode or a side hanging suggestion is generated according to a risk level, and a rearrangement plan is generated based on available net capacity of the gentle section to control the exhaust valve to perform peak-shaving operation. Finally, the execution effect is quantitatively evaluated and the strategy is optimized through a closed-loop evaluation module. The application effectively solves the ice blocking problem caused by the exhaust-load overlap in the cold start process, and improves the start safety and the control intelligent level of the hydrogen energy power supply system.

Description

Technical Field

[0001] This invention relates to the field of hydrogen energy power control technology, and more specifically, to an integrated control system for hydrogen energy power. Background Technology

[0002] Hydrogen fuel cells are increasingly used in vehicles and distributed energy sources, but their cold start phase still presents significant challenges. During cold start-up, water generated by the electrochemical reaction freezes, hindering gas transport and damaging the catalyst and diffusion layers, leading to problems such as pore collapse and platinum particle agglomeration within the stack. To achieve rapid and safe cold start-up, control strategies must balance stack temperature rise rate with material protection, minimizing ice accumulation while meeting start-up time requirements. Existing research indicates that different start-up strategies significantly impact stack voltage, current distribution, and ice formation mechanisms; the choice of current density and load application method directly determine stack temperature distribution and ice growth. Furthermore, the goal of control strategies is not only to shorten start-up time but also to generate sufficient heat quickly while preventing excessive water accumulation.

[0003] During the cold start process of a hydrogen power system, the exhaust valve is used to remove nitrogen and inert gas from the reactor core, and the load is connected to heat the reactor core or output electrical energy. If the exhaust time overlaps with the load ramp-up phase, it may cause the output voltage to drop rapidly, the reactor core temperature to fluctuate, or even cause the system's self-protection shutdown. In research and development practice, it has been found that the risk of exhaust-load overlap is a key pain point in cold start control.

[0004] To address the above problems, this invention proposes a solution. Summary of the Invention

[0005] The purpose of this invention is to provide an integrated control system for hydrogen energy power supply, which solves the technical problem of equipment ice blockage caused by exhaust-load overlap in the prior art.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A hydrogen energy power integrated control system includes the following modules: a cold start timing monitoring and processing analysis module, used to collect cold start operating condition characteristic signals during the cold start phase, and extract the indicator light status and fan speed from the cold start operating condition characteristic signals to determine the green light initial window; and within the green light initial window, identify the exhaust action interval, load ramp-up segment and smooth segment, and construct a sequence of exhaust timing evidence objects, which includes a variety of evidence objects characterizing the exhaust and load timing relationship, voltage behavior and capacity characteristics;

[0008] The intervention assessment and management module is used to extract three risk features—negative overlap rate, voltage valley depth, and recovery constant—based on the negative time-series evidence object sequence, compare them in lexicographical order, and filter by dominance relationship to obtain the maximum risk set, which is the set of evidence objects that are not dominated by any other evidence object.

[0009] The peak-shifting and execution management module is used to determine whether to enter the intervention mode based on whether there are evidence objects with a negative overlap rate greater than zero in the maximum risk set. If it enters the intervention mode, it further determines whether to enter the side-by-side suggestion mode or the limited takeover mode based on the comparison of various risk characteristics with the corresponding preset thresholds, and generates the corresponding execution token. Based on the generated execution token, and based on the available net capacity of the smooth segment, a rearrangement plan is generated through a priority strategy to control the exhaust valve to perform the rearrangement operation within the initial green light window.

[0010] As a further aspect of the present invention, the cold start condition characteristic signals include stack-end voltage, output current, exhaust valve status, load connection mark, indicator light status, and fan speed, which are collected through the controller local area network and various sensors.

[0011] As a further aspect of the present invention: the indicator light status and fan speed in the cold start condition feature signal are extracted to determine the initial green light window. Specifically, the starting point of the initial green light window is defined as the rising edge time when the indicator light changes from red to green; the ending point is the time when the fan speed reaches a fixed upper limit. This initial green light window represents the time region in which negative overlap is likely to occur in the early stage of cold start.

[0012] As a further aspect of this invention: Within the initial green light window, the exhaust action interval, load ramp-up segment, and smooth segment are identified to construct a sequence of exhaust timing evidence objects. Specifically, the exhaust action interval is determined by monitoring the exhaust valve status signal and using an edge-triggered algorithm to identify its rising and falling edges. When multiple consecutive sampling points of the signal maintain a sudden level change, it is determined to be a valid transition. After debouncing processing, the exhaust action interval is obtained. Where i represents the i-th exhaust action, used to distinguish multiple exhaust events that occur during the cold start process, such as... For the i-th exhaust section, Let i be the starting time of the i-th actual exhaust, where i is a positive integer and i = 1, 2, ..., n, and n is the total number of exhausts within the initial green light window W.

[0013] The load ramp-up phase is calculated by differentiating the output current and determining the current change rate. When the change rate exceeds the set load ramp-up threshold and the duration is not less than the minimum duration threshold, the start and end times are recorded to obtain the load ramp-up phase. Where j represents the j-th load ramp-up, For the j-th load ramp-up segment, Let j be the starting time of the j-th load ramp-up, where j is a positive integer and j = 1, 2, ..., m, and m is the total number of load ramp-ups within the initial green light window W.

[0014] The smooth segment is defined as the section within the initial green light window W where the absolute value of the current change rate does not exceed a set current slope threshold and the duration is not less than the minimum duration. The output current within the initial green light window W is iterated through, and when both conditions are met simultaneously in a continuous time period, the start and end times are recorded. and To obtain a smooth segment Where k represents the k-th smooth segment, The available net capacity of the kth smooth segment is given by the value of k, which is a positive integer and k = 1, 2, ..., p, where p is the total number of smooth segments within the initial green light window W. The available net capacity of each smooth segment is obtained by subtracting the valve response protection time slot from the duration of the segment.

[0015] For each exhaust action, the load ramp-up segment with the closest time distance is selected, and the corresponding reference window, exhaust overlap rate, voltage valley depth, recovery constant and available net capacity are combined to form an evidence object containing the exhaust and load timing relationship, voltage behavior and smooth capacity characteristics; all evidence objects constitute the exhaust timing evidence object sequence in chronological order.

[0016] As a further aspect of this invention: the comparison and dominance relationship screening using a lexicographical order to obtain the maximum risk set specifically involves: characterizing each evidence object with a triplet consisting of three risk characteristics: negative overlap rate, voltage valley depth, and recovery constant; if any evidence object is not inferior to another evidence object in all three risk characteristics, and at least one risk characteristic is strictly greater and of higher risk, then it is determined that the evidence object dominates the other evidence object; all evidence objects are traversed and dominated objects are removed, and the remaining undominated evidence objects constitute the maximum risk set, and their evidence objects are marked as the main risk evidence objects; wherein, the comparison of the three risk characteristics is performed in a lexicographical order.

[0017] As a further aspect of the present invention: determining whether to enter the intervention mode specifically includes: the intervention mode specifically includes the side-by-side suggestion mode and the limited takeover mode; the preset thresholds corresponding to each risk characteristic are: the severe critical threshold and the moderate critical threshold of the negative overlap rate; the fuel cell undervoltage associated risk threshold corresponding to the fuel cell voltage valley depth and the fan heat dissipation associated risk threshold corresponding to the recovery constant;

[0018] Based on the obtained maximum risk set, if the maximum risk set is empty, or if the maximum risk set is not empty but the negative overlap rate of all primary risk evidence objects within it is zero, then it is determined that no intervention will be initiated; if the maximum risk set is not empty and there are primary risk evidence objects with a negative overlap rate greater than zero, then it is determined that intervention mode will be entered and the mode subdivision process will begin; in the mode subdivision process, if the negative overlap rate of any primary risk evidence object is greater than the severe critical threshold for negative overlap rate, or if at least two primary risk evidence objects simultaneously satisfy the negative overlap rate being greater than the moderate critical threshold for negative overlap rate and there is an auxiliary risk exceeding the limit, then it is determined that a limited takeover mode will be implemented; if all primary risk evidence objects satisfy the negative overlap rate being greater than zero but not greater than the moderate critical threshold for negative overlap rate and there is no auxiliary risk exceeding the limit, then it is determined that a side-by-side recommendation mode will be implemented; wherein, the auxiliary risk exceeding the limit refers to the stack voltage valley depth being greater than the stack undervoltage associated risk threshold or the recovery constant being greater than the fan heat dissipation associated risk threshold.

[0019] As a further aspect of the present invention: the side-mounted suggestion mode refers to the mode of generating optimization suggestions for the main controller to refer to without direct control; its specific execution process includes: generating a side-mounted suggestion construction drawing based on the available net capacity of the smooth segment through a capacity-distance-translation matching algorithm;

[0020] The capacity-distance-translation matching algorithm filters out smooth segments with sufficient available net capacity to accommodate the exhaust action, selects the segment with the largest time distance from the load ramp-up segment to maximize spatial isolation, and determines the new exhaust section within the segment and adds defined valve response protection time slots at both ends of it; finally, the construction drawing containing the new exhaust section, expected overlap rate and expected voltage valley depth is written into the execution token and marked as executed by the main controller.

[0021] As a further aspect of the present invention: the limited takeover mode refers to the mode of directly controlling the exhaust valve to perform rearrangement operations. Its specific execution process includes: generating a rearrangement demand boundary, which includes a risk control boundary extracted based on the main risk evidence object in the maximum risk set and execution constraints defined based on the mechanical characteristics of the exhaust valve and the safety requirements of the fuel cell stack; encapsulating the rearrangement demand boundary into an execution token and marking it as a direct execution path; wherein, the risk control boundary includes safety thresholds for the negative overlap rate, voltage valley depth, and recovery constant, and the execution constraints include the minimum and maximum duration of a single exhaust action, the time slot interval of the protection time slot, and the adaptation relationship between the total exhaust duration and the total available net capacity of the smooth segment.

[0022] As a further aspect of the present invention: Based on the generated execution token and the available net capacity of the smooth segment, a reordering plan is generated through a priority strategy, including: the priority strategy includes an avoidance strategy, a merging strategy, and a fragmentation strategy. The selection logic is to prioritize the avoidance strategy. If there exists a single smooth segment with an available net capacity greater than or equal to the total duration of the current exhaust action, then the smooth segment with the largest distance from the associated load ramp-up period is selected as the target segment, and a new exhaust interval is set in the center of it. If the avoidance strategy condition is not met, then the merging strategy is adopted. That is, if the time interval between the current exhaust action and the adjacent exhaust action is less than the set action interval threshold, and there exists a smooth segment that can accommodate the total exhaust duration after merging, and the total duration satisfies the constraint of the maximum duration of a single exhaust in the reordering requirement boundary, then the two exhaust actions are merged into one continuous action for execution. If none of the above priority strategies are satisfied, then the fragmentation strategy is adopted, that is, the current exhaust action is divided into multiple segments, based on the reordering requirement boundary... and maximum duration Map the maximum available net capacity to Where k represents the k-th smooth segment; The available net capacity of the k-th smooth segment;

[0023] Give the lower bound of the partition. ;in, The total duration of the i-th exhaust action is the minimum number of wafers determined by the maximum capacity of a single wafer.

[0024] Upper Bound of Fragments The maximum number of plates is determined by the minimum exhaust time;

[0025] The sharding search range is obtained by combining the lower and upper bounds of the sharding. From the lower bound of the fragment Start by successively probing the first minimum feasible number of segments f and output the position of each segment; the setting of the new exhaust interval and the segmented execution of the exhaust action must meet the valve response protection time slot constraint.

[0026] As a further aspect of the present invention, after the peak reduction and execution management module, a closed-loop evaluation module is also included. The steps are as follows: collect execution result data, calculate the overlap clearing rate, voltage valley depth improvement rate and recovery acceleration rate, mark the execution as valid, partially valid or invalid, and generate an adoption template or non-adoption list for reuse or avoidance in subsequent cold start processes.

[0027] The beneficial effects of this invention are:

[0028] (1) This invention, through the cold start timing monitoring and analysis module, accurately identifies the exhaust action interval, load ramp-up segment and smooth segment within the initial green light window, constructs a sequence of exhaust timing evidence objects, and realizes early identification and quantitative analysis of exhaust-load overlap risk during cold start; through the intervention assessment management module, the maximum risk set is selected based on the dictionary order threshold and dominance relationship, realizing rapid location and hierarchical assessment of high-risk evidence objects, and improving the system's response capability to complex risk scenarios; through the peak-shifting resolution and execution management module, the side-hanging suggestion mode or limited takeover mode is automatically selected according to the risk level, and a rearrangement plan is generated based on the available net capacity of the smooth segment, realizing intelligent scheduling and peak-shifting control of the exhaust valve, effectively reducing the risk of ice blockage; through the closed-loop assessment module, the execution results are quantitatively assessed and marked, forming an adoption template and a non-adoption list, realizing self-learning and continuous optimization of the control strategy, improving the system's adaptability and reliability under changing working conditions, and solving the technical problem of equipment ice blockage caused by exhaust load overlap in the prior art. Attached Figure Description

[0029] The invention will now be further described with reference to the accompanying drawings.

[0030] Figure 1 This is a schematic diagram of the module framework of an integrated control system for hydrogen energy power supply according to the present invention;

[0031] Figure 2 This is a schematic diagram of the overall implementation logic of the integrated control of hydrogen energy power supply according to the present invention. Detailed Implementation

[0032] 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.

[0033] Example 1

[0034] Please see Figure 1 As shown, the present invention discloses a hydrogen energy power integrated control system, which is applied to a hydrogen energy individual soldier power supply outdoor emergency portable power generation device. This device generates electricity through a hydrogen fuel cell to provide power to various electronic devices; and furthermore, it precisely manages the time relationship between the exhaust valve and the load interface during the cold start of the device, thereby reducing the risk of ice blockage caused by the overlap of exhaust and load.

[0035] Please see Figure 2 As shown, the present invention is an integrated control system for hydrogen energy power supply, comprising the following modules:

[0036] The cold start timing monitoring and analysis module is used to collect various cold start condition characteristic signals during the equipment cold start implementation phase. It uses the indicator light changing from red to green as the starting point and determines the endpoint based on fan speed stability and power-on time, defining the initial green light window. Within this window, it synchronously extracts the exhaust action interval and load ramp-up segment, identifies smooth segments, constructs a comparison window centered on a single exhaust, calculates various spatiotemporal correlation parameters, and derives the available net capacity by deducting the valve response protection time slot from the smooth segments. Finally, it encapsulates the correlation between a single exhaust and the most recent load ramp-up as a negative timing evidence object, outputting the evidence object sequence and the smooth segment set.

[0037] After each cold start trigger, the system collects characteristic signals of cold start operation, such as stack-end voltage U(t), output current I(t), exhaust valve status E(t), load connection marker L(t), indicator light status G(t), and fan speed F(t), through the controller area network (CAN bus) and various sensors. To improve data reliability, multi-modal synchronous acquisition is adopted, and a first-order hysteresis filter is used at the acquisition end to eliminate high-frequency noise. The sampling frequency is determined according to the stack dynamics and control requirements. In this embodiment, 1kHz is selected, which can cover the voltage and current changes at the moment of opening and closing of the exhaust valve. For voltage and current data, noise is removed by Kalman filtering or exponential smoothing.

[0038] The initial green light window is determined based on the indicator light status G(t) and fan speed F(t) of the device. Specifically, the starting point is the rising edge of the indicator light from red to green; the ending point is the time Tcap when the fan speed reaches the fixed upper limit, which is 20 seconds in this embodiment. The initial green light window W represents the time region in which negative overlap is likely to occur in the early stage of cold start. All subsequent interventions are limited to the initial green light window W to avoid interference with the overall machine control logic.

[0039] For the cold start characteristic signals within the initial green light window W, a hierarchical event detection algorithm is used to extract three types of core events: exhaust action interval, load ramp-up segment, and smooth segment. This provides quantifiable spatiotemporal features for subsequent exhaust timing analysis. The specific implementation steps are as follows:

[0040] By monitoring the exhaust valve status signal E(t), an edge-triggered algorithm is used to identify the rising edge (closed=0→open=1) and falling edge (open=1→closed=0) of the exhaust valve status signal: when the exhaust valve status signal maintains a level change (0→1 or 1→0) for 3 consecutive sampling points, for example, 3ms, it is determined to be a valid transition, and the time of the valid transition is recorded. (rising edge) and (Falling edge), and perform hysteresis debouncing processing on the effective transition time to obtain the exhaust action range. Where i represents the i-th exhaust action, used to distinguish multiple exhaust events that occur during the cold start process. For the i-th exhaust section, Let i be the starting time of the i-th actual exhaust, where i is a positive integer and i = 1, 2, ..., n, and n is the total number of exhausts within the initial green light window W.

[0041] The rate of change of current is calculated by differentiating the output current I(t) and using the central difference method. ;in, Ensure derivative accuracy; set load ramp-up threshold. It is calibrated through a cold start test, that is, the rate of change of current when the load suddenly increases. Typical values ​​are 0.3~0.8A / ms, with the middle value used to cover resistive and inductive loads;

[0042] when And the duration is greater than or equal to To avoid misjudgment due to short-term electromagnetic interference; record the start time. That is, the moment when the threshold is first exceeded and the termination moment. The moment when the load falls back below the threshold is the load ramp-up phase. Where j represents the j-th load surge, used to distinguish multiple load access events. For the j-th load ramp-up segment, Let j be the starting time of the j-th load ramp-up, where j is a positive integer and j = 1, 2, ..., m, and m is the total number of load ramp-ups within the initial green light window W.

[0043] It should be noted that, among them The minimum duration threshold for load ramp-up is typically 50ms. It is used to distinguish between continuous current changes requiring energy charging / discharging / mechanical response from actual load connection, which have a long duration, and transient electromagnetic interference / sensor noise such as switch bounce and electromagnetic pulse, which have a short duration. It can be set by statistical analysis of cold start connection experiments of typical loads or by professional technicians based on specific needs.

[0044] Search within the initial green light window W | |≤ε and duration is greater than or equal to The segment; where ε is the current slope threshold of approximately 0.2 A / ms. The minimum duration is determined by the system's minimum exhaust time and the valve response protection time slot; in this embodiment, it is set to 100 ms. The output current I(t) within the initial green light window W is iterated; when both of the above conditions are met simultaneously in a continuous time period, it is marked as a smooth segment. Record start and end times and To obtain a smooth segment Where k represents the k-th smooth segment, used to distinguish multiple periods of stable current. Let $p$ be the available net capacity of the $k$-th smooth segment, where $k$ is a positive integer and $k = 1, 2, ..., p$, and $p$ is the total number of smooth segments within the initial green light window $W$. The available net capacity of each smooth segment is... ,in, The duration of the k-th smooth segment is given by , and g is the valve response protection time slot of approximately 10 ms.

[0045] Define a reference window for each exhaust action. Where Δ is the control time width, which is 150 ms in this example; calculate the baseline voltage. Calculate the valley voltage based on the median voltage within 200 ms before exhaust. This represents the minimum voltage during exhaust.

[0046] For each exhaust action and its most recent load ramp-up phase, the system calculates the following metric: exhaust-load overlap rate. If there is no climbing section, then Voltage valley depth , recovery constant This is the time required for the voltage to recover from its lowest point to 95% of its baseline.

[0047] Furthermore, for each exhaust action, the load ramp-up segment with the closest time interval is selected. A sequence of negative temporal evidence objects is formed by combining these elements. Among them, the negative chronological sequence of evidence objects includes various types of evidence objects. This includes the timing relationship between exhaust and load, voltage behavior, and smooth capacity characteristics. These evidence objects, arranged in chronological order, constitute a sequence of exhaust timing evidence objects, which is the sole input for subsequent analysis and decision-making. Through the above unified data structure, this module ensures that each indicator is directly related to the risk of overlap between cold start and exhaust.

[0048] The intervention assessment and management module receives the sequence of negative time-series evidence objects, extracts risk judgment features, compares the evidence objects in lexicographical order within the same window, and filters them by dominance relationship to obtain the largest risk set. If the set is empty or the overlap rate is zero, it is determined not to enter the intervention mode. If the set exists but the indicators have not exceeded the warning level, it enters the side-by-side suggestion mode. If there is a high overlap rate or multiple indicators deteriorate simultaneously, it enters the limited takeover mode. If the device permissions are insufficient, it automatically reverts from the limited takeover mode to the side-by-side suggestion mode. In the side-by-side suggestion mode, a side-by-side suggestion construction diagram is generated, which assigns target smooth segments and translation positions to the exhaust gas that need to be adjusted and provides the expected improvement. In the limited takeover mode, an executable rearrangement time sequence is formed, and the above results are naturally encapsulated as an execution token.

[0049] The intervention modes include the side-by-side suggestion mode and the limited takeover mode;

[0050] To minimize the computational burden of the secondary decision-making process, this module first constructs a maximum risk set for the evidence objects within the initial green light window W. Specifically, this means: [details about each piece of evidence]. Using three risk characteristics: negative overlap rate, voltage valley depth, and recovery constant. The representation uses a combination of lexicographical thresholds and dominance relationships for filtering; if Non-inferior in all three risk characteristics ,Right now And at least one of the risk characteristics is more strictly defined and of higher risk. Then it is called Dominate Remove the dominant entity from any evidence object, iterate through all evidence objects, and after removing the dominated object, the remaining objects constitute the maximum risk set in the current green light initial window. ; and mark its evidence object as the main risk evidence object; in engineering implementation, negative overlap rate can be arranged first. Divide into buckets in descending order, and then within each bucket, sort by voltage valley depth. , Local dominance discrimination is performed to reduce the number of comparisons and improve real-time performance; where subscripts a and b are positive integers, representing two different evidence objects within the initial green light window W. The sequence number is used to distinguish the evidence object instances corresponding to different exhaust events, and supports pairwise comparative analysis of their dominance relationships;

[0051] Based on obtaining the maximum risk set First, conduct an initial screening of risk-free and risky scenarios to quickly eliminate those that do not require intervention:

[0052] If any of the following conditions are met, it will be determined that no intervention is needed, and the exit and DEP (Data Evidence Package) archive command will be output directly:

[0053] Maximum risk set An empty value indicates that all evidence objects are under control and there are no major risk evidence objects that need to be focused on.

[0054] or maximum risk set It is not empty, but the negative overlap rate of all main risk evidence objects is not empty. This indicates that there is no time overlap between exhaust and load, and the natural timing can meet safety requirements without intervention;

[0055] If the maximum risk set The negative overlap rate of non-empty objects with primary risk evidence This indicates a risk of overlapping exhaust loads, which is deemed necessary for intervention, and the process will proceed to a detailed mode determination process.

[0056] For scenarios requiring intervention, based on risk level segmentation, various preset thresholds are calibrated using data-driven methods (1000+ cold-start labeled samples). A classification model is constructed using Support Vector Machine (SVM) and Random Forest, and optimized using ROC curves to ensure a false positive rate ≤5% and a false negative rate ≤8%.

[0057] Overlap rate threshold: Critical threshold for moderate negative overlap rate That is, the threshold between slight and moderate overlap, corresponding to an ice blockage risk probability of ≈30%; when the negative overlap rate is eliminated... > Moderate threshold for negative overlap rate At that time, the risk level was upgraded from mild to moderate;

[0058] Negative overlap rate critical threshold This refers to the threshold between moderate and severe overlap, corresponding to an ice blockage risk probability of approximately 85%; when the negative overlap rate is... > Negative overlap rate critical threshold At that time, the risk level was upgraded from moderate to severe;

[0059] Auxiliary risk threshold: The auxiliary risk exceeding the limit is defined as the depth of the stack voltage valley. > Undervoltage associated risk threshold or recovery constant of fuel cell stack > Fan cooling associated risk threshold; among which, the fuel cell stack undervoltage associated risk threshold has an initial value of 0.5V, which is used to characterize the superposition of undervoltage risks caused by transient voltage drops in the fuel cell stack;

[0060] The initial value of the fan heat dissipation associated risk threshold is 0.8s, which is used to characterize the risk of insufficient heat dissipation capacity caused by slow voltage recovery of the fan.

[0061] It should be noted that the voltage valley depth of the fuel cell stack... With recovery constant The reason for merging them into a single auxiliary risk threshold, rather than judging them separately, is that both are non-overlapping risk enhancement indicators. That is, risk escalation only occurs when the auxiliary risk and the negative overlap risk are combined; if only the auxiliary risk exceeds the threshold... =0 but Dᵢ=0.6V, mostly transient interference, no need to trigger high-level takeover;

[0062] If the maximum risk set If any of the following situations exist, it is determined to be a limited takeover mode, requiring direct control of the exhaust valve to avoid high risks:

[0063] When the overlap rate of any primary risk evidence object > Negative overlap rate critical threshold This indicates a serious risk of overlapping exhaust and load times, with exhaust and load times overlapping by more than 60%, resulting in severe obstruction of heat dissipation channels.

[0064] When at least two primary risk evidence objects simultaneously meet the exclusion overlap ratio > Moderate threshold for negative overlap rate Furthermore, the auxiliary risks exceed the limits, indicating multiple superimposed risks, insufficient power or heat dissipation capacity of the fuel cell stack, and the fuel cell stack approaching its limits at both boundaries;

[0065] Conversely, if all primary risk evidence objects meet the following conditions... If there is no auxiliary risk exceeding the limit, it indicates a slight overlap risk, only a low degree of time conflict, and redundancy in both fuel cell power and thermal management. In this case, it is determined to be the side-by-side suggestion mode, in which only optimization suggestions are provided and no direct control is given.

[0066] After determining that the system is in limited takeover mode, the exhaust valve control permission needs to be verified via the hardware permission interface. The interface protocol is Modbus-RTU, and the baud rate is 9600bps.

[0067] If permissions permit, the main controller returns a permission control command, retains the limited takeover mode, and generates a reordering plan for direct execution.

[0068] If permissions are denied, the main controller will return a permission lock command. If valve control permissions are manually taken over, the system will automatically downgrade to the side-by-side recommended mode to avoid permission conflicts that could cause system crashes and ensure that intervention plans can be implemented.

[0069] Furthermore, when the mode is determined to be a side-by-side suggestion mode, the available net capacity of the smooth segment generated by the cold start timing monitoring and analysis module is... The following steps are taken to generate suggested construction drawings for side-mounted structures using a three-level matching algorithm of capacity-distance-translation amount:

[0070] AS1: Reserve sufficient net capacity gentle passages Ensure that the capacity can fully accommodate the exhaust gas; among which The duration of the i-th exhaust action;

[0071] AS2: Calculate smooth segments With load ramp-up section Time distance Select time distance The largest segment maximizes the spatial isolation between load and load;

[0072] If multiple time segments are the same distance apart, choose the translation amount. The smallest segment reduces disturbances to the original timing of the main controller; among which The original planned start time of the i-th exhaust action is the target time for exhaust opening preset in the timing planning of the main controller. The subscript i is associated with the i-th exhaust action, and e represents a specific exhaust event. It is used to calculate the disturbance of the bypass scheme to the original control timing.

[0073] AS3: Define valve response protection time slot initial value It is the initial value of the protection time slot calibrated based on the valve's mechanical response time, used to avoid transient interference from valve operation; in the new exhaust section Add at both ends Protect the time slot, generate a side-by-side suggested construction diagram, and directly write the side-by-side suggested construction diagram into the execution token, with the execution path marked as the main control execution;

[0074] The generated side-mounted suggested construction drawings include new exhaust sections, corresponding smooth segment markings, expected overlap rate, and expected voltage valley depth, for the main controller to refer to;

[0075] When the limited takeover mode is determined, the generated rearrangement requirement boundary must include risk control boundary and execution constraint information. The specific implementation is as follows:

[0076] BS1: Extract the time series data of the primary risk objects in the maximum risk set M(W). Non-primary risk objects are excluded from the boundary because their risks have already been covered by the dominant risk set. For example, only the negative overlap rate is retained. > Moderate threshold for negative overlap rate Or exhaust section with excessive auxiliary risk The associated load ramp-up phase ;

[0077] If a limited takeover mode is triggered due to the negative overlap rate exceeding the severe threshold for any primary risk evidence object in the maximum risk set M(W), indicating a high overlap risk, then the overlap rate control boundary is that the negative overlap rate is less than or equal to the overlap rate safety threshold. Among them, the overlap rate safety threshold The baseline value was 0.2 (i.e., 20%), calibrated using 1000+ cold start samples, and can be dynamically adjusted according to ambient temperature.

[0078] If at least two primary risk evidence objects are present in the maximum risk set M(W), and the limited takeover mode is triggered due to the excessive auxiliary risk, then the voltage valley depth control boundary and the recovery constant control boundary are respectively: the stack voltage valley depth is less than or equal to the stack undervoltage protection threshold. The recovery constant is less than or equal to the fan cooling hysteresis threshold. Among them, the undervoltage protection threshold of the fuel cell stack The base value is 0.4V, corresponding to an undervoltage protection threshold of 0.6V for the fuel cell stack; fan cooling hysteresis threshold. The baseline value is 0.6s. The recovery constant was measured experimentally when the fan is running at full speed. This can cause the heat accumulation rate to exceed the heat dissipation rate. Therefore, this can be used as a boundary, and the temperature can be dynamically adjusted based on feedback from the fan speed sensor.

[0079] BS2: Based on the mechanical characteristics of the exhaust valve and the safe operation requirements of the fuel cell stack, define the execution constraints for the limited control mode, specifically the exhaust time constraint, which includes the minimum duration of a single exhaust action. and maximum duration Its corresponding initial minimum duration The maximum duration is 50ms. It takes 500ms;

[0080] Protection slot constraints include the slot interval of the protection slots. Its initial value is 10ms and the capacity adaptation constraint is that the total duration of all exhausts is less than or equal to the total available net capacity of the smooth segments;

[0081] BS3: Integrate the aforementioned risk control boundaries and execution constraints into rearranged requirement boundaries, encapsulate them into execution tokens, and mark the execution path as direct execution.

[0082] The peak-shifting and execution management module receives the rearrangement requirement boundary and execution token from the reserved intervention assessment management module. Based on the matching relationship between the available net capacity of the smooth segment and the exhaust duration, it generates a rearrangement plan for the main risk evidence object of the largest risk set according to the priority of avoidance → merging → segmentation. According to the generated rearrangement plan, the objects are queued in the time order within the initial window of the green light, and the execution result label is executed sequentially by the single valve-controlled state machine.

[0083] Receive the rearrangement requirement boundary and execution token output from the reserved intervention assessment and management module. Based on the matching relationship between the available net capacity of the smooth segment and the exhaust duration, generate a rearrangement plan using a priority strategy, specifically targeting the set of maximum risks. The following rules apply sequentially to each primary risk evidence object:

[0084] If there exists a single smooth segment Hk such that the available net capacity Ck is greater than or equal to the total exhaust time, then the avoidance strategy is selected; that is, among the smooth segments that satisfy the available net capacity, the smooth segment with the largest distance from the load ramp-up period is selected as the target smooth segment; if they are in parallel, the one with the smallest overall translation amount is selected as the target smooth segment, and there is no need to change the number of exhausts.

[0085] The time slot interval is centered within the target smooth segment, with protective time slot constraints added at both ends. Generate a new exhaust interval; during execution, continuously monitor whether each evidence object exceeds its corresponding rearrangement requirement boundary; if it does, make minor adjustments within that segment, not exceeding 10% of the segment length;

[0086] If the avoidance strategy conditions are not met, but the time interval between the current exhaust action and the adjacent exhaust action is less than the action interval threshold. If there exists a smooth segment that can accommodate the total exhaust duration of the merged exhaust actions, then the merging strategy is selected, that is, the two exhaust actions are merged into a continuous exhaust action, and it is verified whether the total exhaust duration corresponding to the continuous exhaust action is less than or equal to the maximum duration of a single exhaust action at the rearrangement requirement boundary. If the maximum duration of a single exhaust action is less than or equal to the rearrangement requirement boundary, If so, prioritize finding a single, gentle segment to avoid as a whole;

[0087] Among them, the action interval threshold The action interval threshold is calibrated based on the mechanical characteristics of the exhaust valve. When the action interval is less than 200ms, there is no additional wear from continuous action, and the number of voltage valleys can be reduced after merging.

[0088] If the conditions are not met, switch to a sharding strategy;

[0089] If neither the avoidance strategy condition nor the merging strategy condition is met, then a fragmentation strategy is chosen to solve for the minimum number of feasible fragments and the time interval for each fragment, specifically:

[0090] For the maximum risk set For each primary risk evidence object, the minimum duration of a single exhaust action in the rearranged demand boundary is considered. and maximum duration Map the maximum available net capacity to ;

[0091] Give the lower bound of the partition. The minimum number of chips is determined by the maximum capacity of a single chip, ensuring capacity compatibility;

[0092] Upper Bound of Fragments The maximum number of wafers is determined by the minimum exhaust time, avoiding the risk of excessive wafer splitting;

[0093] The search range for the number of fragments is: ,from Start by successively probing the first minimum feasible number of slices f and output the position of each slice. Prioritize selecting fewer slices to reduce the number of exhaust times and reduce the probability of voltage valley depth superposition.

[0094] The specific implementation steps are as follows: Press the smooth segment... Sort in descending order; divide the corresponding exhaust time into f segments, with each segment having an initial duration of... First, verify ,like Skip the first feasible segment number f directly and match the maximum exhaust time constraint in the rearrangement requirement boundary;

[0095] If all segments satisfy If the total exhaust duration is equal to the current total exhaust duration, then the first minimum feasible number of segments f and the final exhaust action interval are obtained; otherwise, try f+1 until a feasible number of segments is found.

[0096] Based on the final exhaust action range obtained, the time slot interval at both ends of each plate maintains the protection time slot constraint. Maintain the minimum spacing between any two pieces;

[0097] Based on the aforementioned priority strategy, the corresponding strategy is selected for execution to generate a rearrangement plan. After the plan is generated, it is queued according to the time order within the window and executed sequentially by a single valve-controlled state machine: within each exhaust interval, the "open valve - hold - close valve" process is completed, and the minimum safety interval is strictly checked between segments and before and after load ramp-up; throughout the execution, the overlap rate, voltage valley depth, and recovery time after execution are calculated in real time and compared with the baseline before the window to obtain the results such as overlap clearance rate, valley depth improvement rate, recovery acceleration rate, and changes in exhaust frequency; if any safety constraint is triggered, the token is immediately pressed to roll back and the anomaly is recorded; at the end of the window, this module outputs the execution result label (valid / partially valid / invalid) and comparison evidence, which, together with the execution token, rearrangement plan, and actual execution trajectory, are archived as the sole input for the module's four-loop evaluation and strategy solidification.

[0098] The closed-loop evaluation module is used to summarize the execution result labels within the window, the sequence of evidence objects before execution, the execution token, and the original negative time-series evidence objects to form a closed-loop dataset for a cold start; calculate core quantities such as overlap clearance rate, voltage valley depth improvement rate, recovery acceleration rate, and changes in the number of exhausts; divide the execution process into effective, partially effective, and ineffective; if the strategy is effective multiple times under a specific working condition, the system generates an adoption template containing information such as strategy type and number of segments, which is reused first in the next matching working condition to shorten the calculation time; and establishes a list of ineffective windows that should not be adopted.

[0099] After each cold start, the system records the execution result labels (including whether the overlap was effectively reduced, whether ice blockage occurred, the improvement rate of voltage valley depth, the improvement rate of recovery time, etc.) together with the evidence object sequence, execution token and rearrangement sequence before execution, forming a closed-loop dataset of cold start events;

[0100] Based on the overlap clearance rate, voltage valley depth improvement rate, recovery acceleration rate, and exhaust frequency change rate calculated before and after execution, the execution process is marked as effective, partially effective, or ineffective. The judgment criteria are corrected through a large number of experiments. For example, if the overlap clearance rate exceeds 80% and the voltage valley depth improves by more than 50%, it is considered effective; if only some indicators are improved, it is considered partially effective; if there is no improvement or new problems are introduced, it is considered ineffective.

[0101] When a strategy is verified to be effective multiple times under specific operating conditions, the system generates an adoption template, which records the strategy type (avoidance, fragmentation, merging), number of fragments, protection time slots, warning threshold intervals, capacity characteristics, and characteristics of the initial green light window (relative start and end times), etc. After the adoption template is matched with the cold start operating condition characteristics, it will be reused preferentially in the next cold start to shorten the calculation time.

[0102] For strategies and related conditions that fail to execute, the system adds them to the list of strategies not to be adopted and marks the key factors that led to the failure, such as insufficient smoothing capacity, excessive minimum exhaust time constraint, and too short window, in order to avoid repeated attempts at the same strategy in the future. The entries in the list will be used as constraints for the strategy selection of Module 2 and Module 3 to ensure the convergence of the search space.

[0103] For windows that frequently fail, time out, or roll back, record the key causes (insufficient tolerance, insufficient smoothing capacity, excessive inter-slice interlocking consumption, excessively short windows, etc.) and create a list of unacceptable combinations. The next time a token is generated or a template is invoked, these infeasible combinations will be excluded first to narrow the search space and reduce invalid calculations and unnecessary triggering.

[0104] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0105] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0106] 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 inventive 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.

[0107] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0108] 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.

[0109] 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 hydrogen energy power supply integrated control system, characterized in that, The system includes the following modules: a cold start timing monitoring and analysis module, used to collect cold start operating condition characteristic signals during the cold start phase, and extract indicator light status and fan speed from the cold start operating condition characteristic signals to determine the initial green light window; within the initial green light window, it identifies the exhaust action interval, load ramp-up segment, and smooth segment, constructing a sequence of exhaust timing evidence objects, which includes various evidence objects characterizing the timing relationship between exhaust and load, voltage behavior, and capacity characteristics; wherein, the starting point of the initial green light window is defined as the rising edge time of the indicator light from red to green; the ending point is the time when the fan speed reaches a fixed upper limit; the available net capacity of the smooth segment is obtained by subtracting the valve response protection time slot from the duration of the segment; The intervention assessment and management module is used to extract three risk features—overlap rate, voltage valley depth, and recovery constant—based on the sequence of evidence objects in the exhaust time series. These features are compared using a lexicographical order and filtered for dominance relationships to obtain the maximum risk set. The maximum risk set is the set of evidence objects not dominated by any other evidence object. Specifically, the overlap rate is the temporal overlap ratio between the exhaust action interval and the load ramp-up segment with the closest time distance; the voltage valley depth is the difference between the valley value of the stack-end voltage during exhaust and the baseline voltage; and the recovery constant is the time required for the voltage to recover from the valley to a preset baseline ratio. The maximum risk set is obtained by comparing and filtering evidence in lexicographical order and considering dominance relationships. Specifically, if the corresponding values ​​of any evidence object on the three risk characteristics are not less than the corresponding values ​​of another evidence object on the three risk characteristics, and the corresponding value on at least one risk characteristic is greater than the corresponding value of the other evidence object on that risk characteristic, then the evidence object is determined to dominate the other evidence object. All evidence objects are traversed and dominated objects are removed. The remaining non-dominated evidence objects constitute the maximum risk set, and their evidence objects are marked as the main risk evidence objects. The comparison of the three risk characteristics is performed in lexicographical order. The peak-shifting and execution management module is used to determine whether to enter the intervention mode based on whether there is evidence of an object with a negative overlap rate greater than zero in the maximum risk set. If it does, it further determines whether to enter the side-by-side suggestion mode or the limited takeover mode based on the comparison of various risk characteristics with the corresponding preset thresholds, and generates the corresponding execution token. Based on the generated execution token, and based on the available net capacity of the smooth segment, a rearrangement plan is generated through a priority strategy, and the exhaust valve is controlled to perform the rearrangement operation within the initial green light window. Among them, the side-by-side suggestion mode refers to the mode of generating optimization suggestions for the main controller to refer to without direct control. Its specific execution process includes: generating a side-by-side suggestion construction drawing based on the available net capacity of the smooth segment through a capacity-distance-translation matching algorithm. The capacity-distance-translation matching algorithm filters out smooth segments with sufficient available net capacity to accommodate the exhaust action, selects the segment with the largest time distance from the load ramp-up segment to maximize spatial isolation, and determines the new exhaust section within this segment and adds defined valve response protection time slots at both ends of it; finally, the construction drawing containing the new exhaust section, expected overlap rate and expected voltage valley depth is written into the execution token and marked as executed by the main controller. Limited takeover mode refers to the mode of directly controlling the exhaust valve to perform rearrangement operations. Its specific execution process includes: generating rearrangement demand boundaries, which include risk control boundaries extracted based on the main risk evidence objects in the maximum risk set and execution constraints defined based on the mechanical characteristics of the exhaust valve and the safety requirements of the fuel cell stack; encapsulating the rearrangement demand boundaries into an execution token and marking it as a direct execution path; wherein, the risk control boundaries include safety thresholds for exhaust overlap rate, voltage valley depth, and recovery constant; the execution constraints include the minimum and maximum duration of a single exhaust action, the time slot interval of the protection time slot, and the adaptation relationship between the total exhaust duration and the total available net capacity of the smooth segment; priority strategies include avoidance strategy, merging strategy, and fragmentation strategy. The selection logic is to prioritize the avoidance strategy, and if the avoidance strategy conditions are not met, the merging strategy is used; if neither the avoidance strategy nor the merging strategy conditions are met, the fragmentation strategy is used.

2. The hydrogen energy power supply integrated control system according to claim 1, characterized in that, Cold start characteristic signals include stack-end voltage, output current, exhaust valve status, load connection marker, indicator light status, and fan speed, all acquired through the controller area network and various sensors.

3. The hydrogen energy power supply integrated control system according to claim 1, characterized in that, The green light initial window indicates the time period in the early stages of a cold start where negative overlap is likely to occur.

4. The hydrogen energy power supply integrated control system according to claim 2, characterized in that, Within the initial green light window, the exhaust action interval, load ramp-up segment, and smooth segment are identified to construct a sequence of exhaust timing evidence objects. Specifically, the exhaust action interval is determined by monitoring the exhaust valve status signal and using an edge-triggered algorithm to identify its rising and falling edges. When the signal maintains a level change at multiple consecutive sampling points, it is determined to be a valid transition. After debouncing, the exhaust action interval is obtained. Where i represents the i-th exhaust action, used to distinguish multiple exhaust events that occur during the cold start process, such as... For the i-th exhaust section, Let i be the starting time of the i-th actual exhaust, where i is a positive integer and i = 1, 2, ..., n, and n is the total number of exhausts within the initial green light window W; where multiple consecutive sampling points of the signal are defined as 3 consecutive sampling points. The load ramp-up phase is calculated by differentiating the output current and determining the current change rate. When the change rate exceeds the set load ramp-up threshold and the duration is not less than the minimum duration threshold, the start and end times are recorded to obtain the load ramp-up phase. Where j represents the j-th load ramp-up, For the j-th load ramp-up segment, Let j be the starting time of the j-th load ramp-up, where j is a positive integer and j = 1, 2, ..., m, and m is the total number of load ramp-ups within the initial green light window W. The smooth segment is defined as the section within the initial green light window W where the absolute value of the current change rate does not exceed a set current slope threshold and the duration is not less than the minimum duration. The output current within the initial green light window W is iterated through, and when both conditions are met simultaneously in a continuous time period, the start and end times are recorded. and To obtain a smooth segment Where k represents the k-th smooth segment, Let p be the available net capacity of the kth smooth segment, where k is a positive integer and k = 1, 2, ..., p, and p is the total number of smooth segments within the initial green light window W. For each exhaust action, the load ramp-up segment with the closest time interval is selected. This, combined with the corresponding reference window, exhaust overlap rate, voltage valley depth, recovery constant, and available net capacity, forms an evidence object encompassing the timing relationship between exhaust and load, voltage behavior, and smooth capacity characteristics. All evidence objects are arranged chronologically to constitute the exhaust timing evidence object sequence. A reference window is defined within this sequence. Δ represents the control time width.

5. The hydrogen energy power supply integrated control system according to claim 1, characterized in that, The determination of whether to enter the intervention mode specifically includes: the intervention mode specifically includes the side-by-side suggestion mode and the limited takeover mode; the preset thresholds corresponding to each risk characteristic are the severe critical threshold and the moderate critical threshold of the negative overlap rate; the fuel cell voltage valley depth corresponding to the fuel cell undervoltage associated risk threshold and the recovery constant corresponding to the fan heat dissipation associated risk threshold; among them, the evidence objects that remain after the dominance relationship screening and constitute the largest risk set are marked as the main risk evidence objects; Based on the obtained maximum risk set, if the maximum risk set is empty, or if the maximum risk set is not empty but the negative overlap rate of all primary risk evidence objects within it is zero, then it is determined that no intervention will be initiated; if the maximum risk set is not empty and there are primary risk evidence objects with a negative overlap rate greater than zero, then it is determined that intervention mode will be entered and the mode subdivision process will begin; in the mode subdivision process, if the negative overlap rate of any primary risk evidence object is greater than the severe critical threshold for negative overlap rate, or if at least two primary risk evidence objects simultaneously satisfy the negative overlap rate being greater than the moderate critical threshold for negative overlap rate and there is an auxiliary risk exceeding the limit, then it is determined that a limited takeover mode will be implemented; if all primary risk evidence objects satisfy the negative overlap rate being greater than zero but not greater than the moderate critical threshold for negative overlap rate and there is no auxiliary risk exceeding the limit, then it is determined that a side-by-side recommendation mode will be implemented; wherein, the auxiliary risk exceeding the limit refers to the stack voltage valley depth being greater than the stack undervoltage associated risk threshold or the recovery constant being greater than the fan heat dissipation associated risk threshold.

6. The hydrogen energy power integrated control system according to claim 1, characterized in that, Priority strategies include avoidance, merging, and fragmentation. Specifically: the selection logic prioritizes the avoidance strategy. If the available net capacity of a single smooth segment is greater than or equal to the total duration of the current exhaust action, the smooth segment with the largest distance from the associated load ramp-up period is selected as the target segment, and a new exhaust interval is set in the center of it. If the avoidance strategy condition is not met, the merging strategy is adopted. That is, if the time interval between the current exhaust action and the adjacent exhaust action is less than the set action interval threshold, and there is a smooth segment that can accommodate the total duration of the merged exhaust, and this total duration satisfies the constraint on the maximum duration of a single exhaust in the rearrangement requirement boundary, then the two exhaust actions are merged into one continuous action for execution. If none of the above priority strategies are met, the fragmentation strategy is adopted, that is, the current exhaust action is divided into multiple segments, based on the constraint on the maximum duration of a single exhaust in the rearrangement requirement boundary. and maximum duration Map the maximum available net capacity to Where k represents the k-th smooth segment; The available net capacity of the k-th smooth segment; Give the lower bound of the partition. ;in, Let i be the total duration of the i-th exhaust action. For each smooth segment Maximum capacity of a single chip The minimum number of chips determined by the maximum capacity of a single chip; Upper Bound of Fragments The maximum number of plates is determined by the minimum exhaust time; The sharding search range is obtained by combining the lower and upper bounds of the sharding. From the lower bound of the fragment Start by successively probing the first minimum feasible number of segments f and output the position of each segment; the setting of the new exhaust interval and the segmented execution of the exhaust action must meet the valve response protection time slot constraint.

7. The hydrogen energy power supply integrated control system according to claim 1, characterized in that, The module following the peak-shifting elimination and execution management module also includes a closed-loop evaluation module. Its steps are as follows: collect execution result data, calculate the overlap rate, voltage valley depth, and recovery time after execution in real time, and compare them with the baseline before the window to obtain the overlap clearance rate, voltage valley depth improvement rate, and recovery acceleration rate. Mark the execution as effective, partially effective, or ineffective, and generate an adoption template or non-adoption list for reuse or avoidance in subsequent cold start processes.

Images