A fault early warning method for energy storage converter based on intelligent sensor

Abstract

This invention discloses a fault early warning method for energy storage converters based on intelligent sensors, belonging to the field of energy storage fault early warning technology. The method includes: collecting branch currents, power device temperatures, and operating parameters of each parallel branch to form an original multi-source time-series dataset; calculating the current imbalance sequence and power device temperature rise rate sequence of each parallel branch based on the original multi-source time-series dataset, and extracting ambient temperature, cumulative charge / discharge cycles, and real-time load rate as operating condition correction factors from the original multi-source time-series dataset; calculating the cumulative current deviation and cumulative heat growth within the corresponding monitoring period based on the current imbalance sequence and power device temperature rise rate sequence, and combining the cumulative current deviation and cumulative heat growth according to the branch number to generate a branch-level coupling feature sequence. This invention achieves dual-closed-loop adaptive adjustment and self-evolution of the early warning boundary.

Description

Technical Field

[0001] This invention relates to the field of energy storage fault early warning technology, and in particular to a fault early warning method for energy storage converters based on intelligent sensors. Background Technology

[0002] In the development of new energy power systems, the operational reliability of energy storage converters is crucial. With the advancement of intelligent sensing technology, fault early warning methods based on multi-source data fusion have become a research hotspot. Existing technologies mainly employ fixed threshold monitoring, signal processing, and time-series classification models: early thresholding methods are simple to implement but struggle to capture early, weak faults; wavelet transform and other signal processing methods improve feature extraction capabilities but still rely on manual judgment; while deep learning models can identify complex patterns, their black-box nature and high computing power requirements limit real-time applications at the edge.

[0003] Existing technologies for early warning of parallel branches in energy storage converters have two shortcomings: First, most methods analyze current imbalance and power device temperature rise as independent features in parallel, ignoring the nonlinear coupling and time accumulation effect of the two in the fault evolution, resulting in insufficient sensitivity to identify early faults such as progressive thermal damage caused by circulating current; Second, early warning thresholds mostly adopt fixed benchmarks or linear corrections under single operating conditions, lacking multi-dimensional adaptive learning capabilities for ambient temperature, equipment aging, and historical early warning accuracy, which easily leads to false alarms or missed alarms under complex operating conditions, making it difficult to balance system robustness and sensitivity. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a fault early warning method for energy storage converters based on intelligent sensors to solve the problems of low early fault identification rate and difficulty in balancing false alarms and missed alarms under complex operating conditions caused by insufficient mining of electro-thermal coupling features and weak threshold adaptive capability.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a fault early warning method for energy storage converters based on intelligent sensors, comprising: collecting branch currents, power device temperatures, and operating parameters of each parallel branch to form an original multi-source time-series dataset; calculating the current imbalance sequence and power device temperature rise rate sequence of each parallel branch based on the original multi-source time-series dataset, and extracting ambient temperature, cumulative charge-discharge cycles, and real-time load rate as operating condition correction factors from the original multi-source time-series dataset; calculating the cumulative current deviation and cumulative heat growth within the corresponding monitoring period based on the current imbalance sequence and power device temperature rise rate sequence, and accumulating the current deviation... Accumulated heat and thermal growth are combined in a one-to-one correspondence according to branch numbers to generate a branch-level coupling characteristic sequence. Based on the branch-level coupling characteristic sequence, a circulation-thermal coupling correlation function is constructed to generate a coupling risk index. The preset baseline warning threshold is dynamically corrected based on the operating condition correction factor and historical warning response feedback data to obtain a dynamic warning threshold. The coupling risk index and the dynamic warning threshold are compared in real time, and combined with a duration filtering mechanism, a graded warning command is generated. Based on the graded warning command, linkage control actions are executed and operation monitoring data is collected. The correction parameters in the dynamic warning threshold are updated based on the operation monitoring data feedback.

[0007] As a preferred embodiment of the fault early warning method for energy storage converters based on intelligent sensors described in this invention, the specific steps for forming the original multi-source time-series dataset are as follows: The energy storage converter is divided into branches according to the parallel branches, and the corresponding branch current, power device temperature and operating condition parameters are collected in sequence based on each branch to obtain the original sampling parameters of each parallel branch. The original sampling parameters of each parallel branch are integrated to form the original multi-source time series dataset.

[0008] As a preferred embodiment of the fault early warning method for energy storage converters based on intelligent sensors described in this invention, the specific steps for calculating the current imbalance sequence and power device temperature rise rate sequence of each parallel branch based on the original multi-source time-series dataset are as follows: The total current sequence is obtained by summarizing the current values ​​of each parallel branch from the original multi-source time series dataset, and the total current sequence is averaged to obtain the theoretical average current sequence. Obtain the deviation ratio between the current value of each parallel branch and the theoretical average current sequence, obtain the current imbalance at each time point, and form a current imbalance sequence. Based on the current imbalance sequence, the corresponding power device temperature data is extracted from the original multi-source time series dataset, and the temperature change slope is obtained by fitting through a sliding window to form the power device temperature rise rate sequence.

[0009] As a preferred embodiment of the fault early warning method for energy storage converters based on intelligent sensors described in this invention, the specific steps for extracting ambient temperature, cumulative charge-discharge cycles, and real-time load rate as operating condition correction factors from the original multi-source time-series dataset are as follows: Ambient temperature, cumulative charge / discharge cycles, and real-time load rate are read from the original multi-source time-series dataset, and the real-time load rate is normalized to form a working condition correction source data group. The environmental degradation factor, aging degradation factor, and load deviation factor are calculated based on the operating condition correction source data set, and the operating condition correction factor is generated by aligning and encapsulating the data according to the time index.

[0010] As a preferred embodiment of the fault early warning method for energy storage converters based on intelligent sensors described in this invention, the specific steps for calculating the cumulative current deviation and cumulative heat growth within the corresponding monitoring period based on the current imbalance sequence and the power device temperature rise rate sequence are as follows: Each monitoring period is determined by combining the current imbalance sequence and the power device temperature rise rate sequence, and the current imbalance data and power device temperature rise rate data corresponding to each monitoring period are extracted. By using current imbalance data and power device temperature rise rate data, the branch interaction change trajectory is constructed, and the interaction characteristic quantity between current imbalance change and temperature rise change is obtained. Interactive features are used to identify continuous thermal growth closed sections within each monitoring period. Based on current imbalance data, power device temperature rise rate data, interactive features, and continuous thermal growth closed sections, the cumulative current deviation and cumulative thermal growth are calculated.

[0011] As a preferred embodiment of the fault early warning method for energy storage converters based on intelligent sensors described in this invention, the specific steps for generating the branch-level coupling feature sequence are as follows: Branch numbers are extracted from the cumulative current deviation and cumulative heat growth, and the cumulative current deviation and cumulative heat growth are paired one-to-one under electrothermal correlation constraints according to the branch numbers to generate an electrothermal coupling feature vector. Based on the electrothermal coupling feature vectors arranged continuously in the time sequence of the monitoring cycle, a branch-level coupling feature sequence is generated.

[0012] As a preferred embodiment of the fault early warning method for energy storage converters based on intelligent sensors described in this invention, the specific steps for generating the coupling risk index are as follows: Electrothermal coupling feature vectors are extracted from the branch-level coupling feature sequence, and causal relationships between branches are established based on the changes in electrothermal coupling features of each parallel branch in adjacent monitoring periods. By utilizing the causal relationships between branches, the electrothermal coupling feature vector is continuously mapped to generate the circulating-thermal coupling response trajectory. Based on the circulation-thermal coupling response trajectory, a circulation-thermal coupling correlation function is constructed, and the coupling risk index is calculated based on the circulation-thermal coupling correlation function.

[0013] As a preferred embodiment of the fault early warning method for energy storage converters based on intelligent sensors described in this invention, the specific steps for obtaining the dynamic early warning threshold are as follows: The operating condition correction factor is subjected to parameter mapping and time serialization to obtain the operating condition adjustment parameter sequence. Using the sequence of operating condition adjustment parameters, a threshold drift function is constructed and an adaptive threshold component for the current operating condition is generated. The adaptive threshold component of the operating condition is corrected by using historical early warning and response feedback data to obtain historical feedback correction values. The historical feedback correction values ​​are then superimposed with the adaptive threshold component of the operating condition to obtain the dynamic early warning threshold.

[0014] As a preferred embodiment of the fault early warning method for energy storage converters based on intelligent sensors described in this invention, the specific steps for generating hierarchical early warning commands are as follows: The risk index is compared with the dynamic early warning threshold in real time to obtain the risk exceedance state sequence; By using a duration filtering mechanism to count and maintain continuous states within a time window of the risk exceedance state sequence, an effective warning persistence indicator is obtained. The risk level is determined by matching the valid warning continuity indicator with the preset multi-level confidence threshold, and a graded warning instruction is generated.

[0015] As a preferred embodiment of the energy storage converter fault early warning method based on intelligent sensors described in this invention, the specific steps of executing linkage control actions according to the hierarchical early warning command and collecting operation monitoring data, and updating the correction parameters in the dynamic early warning threshold based on the feedback of the operation monitoring data, are as follows: Based on the graded early warning instructions, the linkage control actions corresponding to each risk level are determined, and the corresponding parallel branches are subjected to de-rate control and flow equalization adjustment in accordance with the linkage control actions to obtain the disposal execution status; Based on the execution status of the response, the parallel branches after the linkage control action is executed are continuously monitored, and the corresponding operating status parameters and response information are collected to form operating monitoring data. The degree to which coordinated control actions suppress risk changes is determined by using operational monitoring data, and the correction parameters in the dynamic early warning threshold are updated based on the degree of suppression.

[0016] The beneficial effects of this invention are as follows: by calculating the cumulative amount of electrical and thermal stress and constructing a coupled correlation function, the nonlinear coupling characterization of electrical and thermal stress is realized, thereby improving the sensitivity of early hidden fault identification, extending the effective warning window, and avoiding fault missed detection. By correcting the threshold through dual closed-loop correction of operating conditions and historical feedback, the dual closed-loop adaptive adjustment and self-evolution of the warning boundary are realized, thereby achieving the effect of effectively balancing robustness and sensitivity and reducing false alarm and false alarm rates in variable operating condition scenarios. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a fault early warning method for energy storage converters based on smart sensors.

[0019] Figure 2 A flowchart for generating the cumulative quantity and branch-level coupling feature sequence.

[0020] Figure 3 This is a flowchart for hierarchical early warning linkage control and feedback closed loop.

[0021] Figure 4 This is a graph showing the changes in dynamic early warning thresholds under different operating conditions.

[0022] Figure 5 This is a time series diagram of the early warning indicators.

[0023] Figure 6 This is a diagram illustrating the linkage control effect. Detailed Implementation

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0025] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0026] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0027] Reference Figures 1-6 As one embodiment of the present invention, this embodiment provides a fault early warning method for energy storage converters based on smart sensors, comprising the following steps: S1. Collect the branch current, power device temperature and operating parameters of each parallel branch to form the original multi-source time series dataset.

[0028] The energy storage converter is divided into branches according to parallel branches, and the corresponding branch current, power device temperature and operating condition parameters are collected sequentially based on each branch to obtain the original sampling parameters of each parallel branch.

[0029] The specific process includes dividing the conductive loops on the DC or AC side into several independent parallel branches based on the physical topology of the power circuit inside the energy storage converter; deploying high-precision current sensors and temperature sensing elements for each parallel branch; using the current sensors to acquire the instantaneous value of the branch current flowing through the loop in real time; simultaneously acquiring the real-time temperature readings of the power devices through temperature sensing elements attached to the surface of the insulated gate bipolar transistor or heat sink; and combining this with the operating condition detection circuit to obtain the current operating condition parameters; sending the analog signals, such as the instantaneous value of the branch current, the real-time temperature readings of the power devices, and the operating condition parameters, into a signal conditioning circuit for filtering and amplification; converting them into digital signals through an analog-to-digital converter; and aligning and packaging them according to a unified timestamp to obtain the original sampling parameters of each parallel branch that reflect the electrical and thermal state of each parallel branch at a specific moment.

[0030] The original sampling parameters of each parallel branch are integrated to form the original multi-source time series dataset.

[0031] The specific process includes integrating the original sampling parameters of each parallel branch according to a unified timestamp, arranging and reorganizing the instantaneous values ​​of branch current, real-time readings of power device temperature, and operating condition parameters collected at the same time in the order of time series, and collecting the electrical and thermal state parameters of different parallel branches at the same time to form the original multi-source time series dataset.

[0032] S2. Based on the original multi-source time series dataset, calculate the current imbalance sequence and power device temperature rise rate sequence of each parallel branch, and extract the ambient temperature, cumulative charge and discharge times and real-time load rate from the original multi-source time series dataset as operating condition correction factors.

[0033] The total current sequence is obtained by summarizing the current values ​​of each parallel branch from the original multi-source time series dataset, and then averaging the total current sequence to obtain the theoretical average current sequence.

[0034] The specific process includes extracting the instantaneous values ​​of the branch current of each parallel branch from the original multi-source time-series dataset, summing the instantaneous values ​​of the branch current of all parallel branches at the same time to obtain the total current sequence reflecting the overall current change trend of the energy storage converter, and performing an arithmetic average calculation on the total current values ​​within the continuous time window in the total current sequence to obtain the theoretical average current sequence.

[0035] Obtain the deviation ratio between the current value of each parallel branch and the theoretical average current sequence, obtain the current imbalance at each time point, and form a current imbalance sequence.

[0036] The specific process includes extracting the instantaneous branch current values ​​of each parallel branch from the original multi-source time series dataset; for each time stamp, obtaining the difference between the instantaneous branch current value of each parallel branch and the theoretical average current value at the same time in the theoretical average current sequence; performing a ratio calculation between the difference and the theoretical average current value at that time in the theoretical average current sequence to obtain the current imbalance degree at each time that reflects the deviation of the current of a single parallel branch from the overall average level; and then arranging the current imbalance degrees of all parallel branches at each time according to the time series order of the original multi-source time series dataset to form a current imbalance degree sequence that can completely record the changes in the current imbalance state of each parallel branch over time.

[0037] Based on the current imbalance sequence, the corresponding power device temperature data is extracted from the original multi-source time series dataset, and the temperature change slope is obtained by fitting through a sliding window to form the power device temperature rise rate sequence.

[0038] The specific process includes: based on the current imbalance state recorded in the current imbalance sequence at each moment, extracting the real-time readings of the power device temperature at the corresponding moment from the original multi-source time-series dataset; setting a fixed-length time sliding window for each parallel branch; performing linear fitting on the continuous real-time readings of the power device temperature within the window in chronological order; obtaining the temperature change value per unit time by the slope of the fitted straight line; and arranging all the temperature change values ​​per unit time in a time series to form a power device temperature rise rate sequence that characterizes the speed of the power device temperature rise.

[0039] Ambient temperature, cumulative charge / discharge cycles, and real-time load rate are read from the original multi-source time-series dataset, and the real-time load rate is normalized to form a working condition correction source data group.

[0040] The specific process includes extracting the ambient temperature value, cumulative charge / discharge count value, and real-time load rate obtained by the operating condition detection circuit from the original multi-source time-series dataset; determining the maximum rated power reference for the real-time load rate; performing a ratio calculation between the real-time load rate and the maximum rated power reference; converting the real-time load rate into a dimensionless normalized real-time load rate; and combining the ambient temperature value, cumulative charge / discharge count value, and normalized real-time load rate according to the same timestamp to form an operating condition correction source data group that includes environmental thermal conditions, equipment usage frequency, and current load intensity.

[0041] The environmental degradation factor, aging degradation factor, and load deviation factor are calculated based on the operating condition correction source data set, and the operating condition correction factor is generated by aligning and encapsulating the data according to the time index.

[0042] The specific process includes: based on the ambient temperature values ​​in the operating condition correction source data set, calculating the absolute value of the difference between the real-time ambient temperature and the reference ambient temperature; taking the negative value of the product of the ratio of the absolute value of the difference to the reference ambient temperature and the environmental sensitivity coefficient; and obtaining the environmental degradation factor through a natural exponential function. Based on the cumulative charge / discharge cycles in the operating condition correction source data set, calculating the ratio of the cumulative charge / discharge cycles to the designed rated cycle life; taking the negative value of the product of the ratio and the aging sensitivity coefficient; and obtaining the aging degradation factor through a natural exponential function. Based on the normalized real-time load rate in the operating condition correction source data set, calculating the absolute value of the difference between the normalized real-time load rate and the optimal operating power; adding one to the product of the ratio of the absolute value of the difference to the rated power and the load sensitivity coefficient to obtain the load deviation factor. Finally, aligning and encapsulating the environmental degradation factor, aging degradation factor, and load deviation factor according to the time index to generate the operating condition correction factor.

[0043] The expressions for calculating the environmental degradation factor, aging degradation factor, and load deviation factor are as follows: ; ; ; in, Indicates the environmental degradation factor. This represents the natural exponential function. Indicates the environmental sensitivity coefficient. Indicates the real-time ambient temperature. Indicates the reference ambient temperature. Indicates the aging degradation factor. Indicates the aging sensitivity coefficient. Indicates the cumulative number of charge and discharge cycles. Indicates the design rated cycle life. Indicates the load deviation factor. Indicates the load sensitivity coefficient. Indicates real-time load power. Indicates the optimal operating power. Indicates the rated power.

[0044] It should be noted that the environmental sensitivity coefficient is obtained by adjusting the degree of influence of changes in ambient temperature on the warning threshold; the aging sensitivity coefficient is obtained by adjusting the degree of influence of the cumulative number of charge and discharge cycles on the warning threshold; and the load deviation factor is obtained by normalizing the deviation of the real-time load rate, the optimal operating power value, and the rated power value.

[0045] S3. Calculate the cumulative current deviation and cumulative heat growth within the corresponding monitoring period based on the current imbalance sequence and the power device temperature rise rate sequence, and combine the cumulative current deviation and cumulative heat growth according to the branch number to generate a branch-level coupling characteristic sequence.

[0046] Each monitoring period is determined by combining the current imbalance sequence and the power device temperature rise rate sequence, and the current imbalance data and power device temperature rise rate data corresponding to each monitoring period are extracted.

[0047] The specific process includes determining each monitoring cycle based on the current imbalance sequence and the power device temperature rise rate sequence; aligning the two sets of sequence data using timestamps by synchronously acquiring the current imbalance sequence and the power device temperature rise rate sequence; identifying the current imbalance mutation point and the power device temperature rise rate inflection point based on the fluctuation characteristics of the current imbalance sequence and the changing trend of the power device temperature rise rate sequence; dividing the time interval between adjacent mutation points and inflection points into independent monitoring cycles; and extracting the current imbalance sequence data segment and the power device temperature rise rate sequence data segment within the corresponding time range for each monitoring cycle, which are respectively used as the current imbalance data and the power device temperature rise rate data for each monitoring cycle.

[0048] By using current imbalance data and power device temperature rise rate data, the branch interaction change trajectory is constructed, and the interaction characteristic quantity between current imbalance change and temperature rise change is obtained.

[0049] The specific process includes constructing a branch interaction change trajectory in a two-dimensional coordinate system using current imbalance data and power device temperature rise rate data. The current imbalance data is used as the horizontal axis variable, and the power device temperature rise rate data is used as the vertical axis variable. Data points are plotted in the coordinate system according to the time synchronization relationship and connected to form a continuous trajectory. The trajectory characteristics are determined by obtaining the curvature, tangent slope change rate, and enclosing area of ​​the trajectory. The trajectory characteristics are quantified into interactive characteristic quantities that reflect the coordinated evolution law of current imbalance and power device temperature rise rate, thus obtaining the interactive characteristic quantities between current imbalance change and temperature rise change.

[0050] Interactive features are used to identify continuous thermal growth closed segments within each monitoring period. Based on current imbalance data, power device temperature rise rate data, interactive features, and continuous thermal growth closed segments, the cumulative current deviation and cumulative thermal growth are calculated and expressed as follows: ; ; in, This indicates the cumulative amount of current deviation. Indicates the current calculation time. Indicates the length of the monitoring time window. Indicating a historical moment The current imbalance sequence, This represents any historical moment within the window. Represents the current memory time constant. This represents the cumulative amount of heat growth. Indicating a historical moment The power device temperature rise rate sequence, Indicates the electro-thermal coupling coefficient. This represents the thermal memory time constant.

[0051] It should be noted that, It refers to a quantitative parameter that reflects the degree of influence of current imbalance on the temperature rise rate of power devices. It is obtained by analyzing the linear regression relationship between current imbalance and temperature rise rate in historical monitoring data. It refers to a parameter that reflects the degree to which the current imbalance retains historical states over time. It refers to the parameter that reflects the degree to which the temperature rise rate of a power device retains its historical state over time. Both are usually determined by analyzing the exponential decay characteristics of the current imbalance sequence and the power device temperature rise rate sequence in historical data (the decay characteristics refer to the exponential convergence law that the numerical change of the current imbalance sequence and the power device temperature rise rate sequence gradually decreases and approaches a stable value over time), using curve fitting or correlation function analysis.

[0052] The specific process includes: analyzing the time series of interactive characteristic quantities to identify continuous time intervals where the interactive characteristic quantities are consistently higher than the baseline threshold, defining these continuous time intervals as continuous thermal growth closed segments; based on current imbalance data, power device temperature rise rate data, interactive characteristic quantities, and continuous thermal growth closed segments, within the monitoring time window, for the historical current imbalance sequence, combined with the current memory time constant, the cumulative current deviation is obtained through integration calculation; for the historical power device temperature rise rate sequence, combined with the electrothermal coupling coefficient, current imbalance sequence, and thermal memory time constant, the cumulative thermal growth is obtained through integration calculation. The cumulative current deviation reflects the cumulative effect of current imbalance within the time window, and the cumulative thermal growth reflects the cumulative effect of power device temperature rise rate under the influence of electro-thermal coupling.

[0053] It should be noted that the baseline threshold is determined based on the mean and standard deviation of the interaction features within the historical monitoring period, with an exemplary range of values ​​between 1.5 and 2.5 times the standard deviation of the mean of the interaction features.

[0054] Branch numbers are extracted from the cumulative current deviation and cumulative heat growth, and the cumulative current deviation and cumulative heat growth are paired one-to-one under electrothermal correlation constraints according to the branch numbers to generate electrothermal coupling feature vectors.

[0055] The specific process includes extracting the branch number from the cumulative current deviation and the cumulative heat growth, pairing the cumulative current deviation and the cumulative heat growth corresponding to the same branch number one by one under the electrothermal correlation constraint, ensuring that each branch number corresponds to only one set of cumulative current deviation and cumulative heat growth, and combining each set of paired cumulative current deviation and cumulative heat growth into a triplet containing the branch number, cumulative current deviation, and cumulative heat growth. This triplet is the electrothermal coupling feature vector.

[0056] Based on the electrothermal coupling feature vectors arranged continuously in the time sequence of the monitoring cycle, a branch-level coupling feature sequence is generated.

[0057] The specific process includes arranging the electrothermal coupling feature vectors corresponding to different monitoring cycles in chronological order according to the start time of the monitoring cycle, forming an ordered sequence containing electrothermal coupling feature vectors of multiple monitoring cycles. This ordered sequence is the branch-level coupling feature sequence, which fully records the temporal evolution of the cumulative current deviation and cumulative heat growth of a single branch in multiple consecutive monitoring cycles.

[0058] S4. Construct a circulation-thermal coupling correlation function based on the branch-level coupling characteristic sequence, generate a coupling risk index, and dynamically correct the preset benchmark early warning threshold based on the operating condition correction factor and historical early warning response feedback data to obtain the dynamic early warning threshold.

[0059] Electrothermal coupling feature vectors are extracted from the branch-level coupling feature sequence, and causal relationships between branches are established based on the changes in electrothermal coupling features of each parallel branch in adjacent monitoring periods.

[0060] The specific process includes extracting electrothermal coupling feature vectors from the branch-level coupling feature sequence, and based on the changing trend of the electrothermal coupling feature vectors of each parallel branch in adjacent monitoring periods, identifying parallel branch pairs with consistent changing trends by comparing the direction and magnitude of changes in the cumulative current deviation and cumulative heat growth of different parallel branches in adjacent monitoring periods. The correlation between parallel branch pairs with consistent changing trends is defined as a causal relationship, which characterizes the mutual influence between parallel branches formed by the coordinated changes in electrothermal coupling state.

[0061] By utilizing the causal relationships between branches, the electrothermal coupling feature vector is continuously mapped to generate the circulating-thermal coupling response trajectory.

[0062] The specific process includes determining the mutual influence and connection of parallel branches based on the causal relationship between branches, mapping the electrothermal coupling feature vectors corresponding to the parallel branches with causal relationships onto a two-dimensional coordinate system in the time sequence of the monitoring cycle, using the cumulative current deviation as the horizontal axis variable and the cumulative heat growth as the vertical axis variable, plotting the data points corresponding to each monitoring cycle in the two-dimensional coordinate system, and connecting them sequentially to form a continuous curve, which is the circulation-thermal coupling response trajectory.

[0063] Based on the circulation-thermal coupling response trajectory, a circulation-thermal coupling correlation function is constructed, and the coupling risk index is calculated according to the circulation-thermal coupling correlation function. The expression is as follows: ; in, This represents the coupling risk index. This represents the heat-electric weighting coefficient. This represents the mismatch penalty coefficient. This represents the numerical stability constant.

[0064] It should be noted that, It refers to the adjustment parameter used to determine the relative importance of the cumulative heat growth and the cumulative current deviation, which is obtained based on statistical correlation analysis of historical data; It refers to the adjustment parameter used in the circulating current-thermal coupling correlation function to amplify the relative imbalance between the cumulative current deviation and the cumulative heat growth, which is determined based on statistical analysis of historical imbalance data.

[0065] The specific process includes, based on the temporal evolution of the cumulative current deviation and the cumulative heat growth in the circulating current-thermal coupling response trajectory, incorporating the normalized ratio of the cumulative current deviation and the cumulative heat growth, the normalized product of the thermal-electric weighting coefficient and the cumulative current deviation, the imbalance penalty coefficient, and the numerical stability constant into the calculation framework. The circulating current-thermal coupling correlation function is formed by combining the square root of the sum of squares and multiplication operations. The coupling risk index is then calculated based on the circulating current-thermal coupling correlation function. The coupling risk index comprehensively reflects the degree of operational risk of the parallel branch under the synergistic effect of current deviation and heat growth. The higher the value, the greater the risk of electrothermal coupling imbalance.

[0066] It should be noted that the circulating current-thermal coupling correlation function is used to quantify the nonlinear coupling relationship between the cumulative current deviation and the cumulative heat growth. The circulating current-thermal coupling correlation function continuously maps the electrothermal coupling feature vectors in the branch-level coupling feature sequence through causal relationships, reflecting the changes in electrothermal coupling features of parallel branches in adjacent monitoring cycles, and characterizing the risk intensity of circulating current-thermal coordinated faults in energy storage converters. The application of the circulating current-thermal coupling correlation function realizes the nonlinear coupling characterization of the cumulative current deviation and the cumulative heat growth in the time domain, breaking through the limitations of traditional single-parameter independent analysis. It uses the integral accumulation mechanism to capture the memory effect of fault evolution, quantifies the progressive thermal damage evolution process caused by circulating current, improves the sensitivity of early hidden fault identification, extends the effective warning window, avoids fault missed detection, and solves the problem of insufficient electrothermal coupling feature mining.

[0067] The operating condition correction factor is subjected to parameter mapping and time serialization to obtain the operating condition adjustment parameter sequence.

[0068] The specific process includes determining the range of values ​​for the operating condition correction factor based on the operating condition type of the parallel branch, mapping the operating condition correction factors corresponding to different operating condition types to a one-dimensional numerical sequence in the order of the monitoring cycle, using the monitoring cycle number as the sequence index and the specific value of the operating condition correction factor as the sequence element, to form a sequence of operating condition adjustment parameters arranged in chronological order.

[0069] Using the sequence of operating condition adjustment parameters, a threshold drift function is constructed and an adaptive threshold component for the current operating condition is generated.

[0070] The specific process includes extracting the operating condition correction factor corresponding to the current monitoring cycle from the operating condition adjustment parameter sequence, multiplying the operating condition correction factor by the thermal-electric weighting coefficient to obtain the weighted correction value, multiplying the coupling risk index by the imbalance penalty coefficient to obtain the risk weighting value, adding the weighted correction value and the risk weighting value to obtain the output value of the threshold drift function, and linearly superimposing the output value of the threshold drift function with the preset base threshold to obtain the operating condition adaptive threshold component at the current moment.

[0071] It should be noted that the base threshold is preset based on the statistical average of the historical electrothermal coupling characteristics of the parallel branches under standard operating conditions, with an exemplary value range of 1.5 to 3.0. The threshold drift function is used to quantify the influence of ambient temperature, equipment aging, and load changes on the preset baseline warning threshold. The threshold drift function performs parameter mapping and time serialization processing on the operating condition correction factor, converting the environmental attenuation factor, aging attenuation factor, and load deviation factor into the current operating condition adaptive threshold component. The threshold drift function enables the warning threshold to be dynamically adjusted according to the operating conditions, overcoming the rigidity problem of fixed thresholds under complex operating conditions, giving the warning model the ability to adaptively learn from environmental changes and equipment aging, balancing the robustness and sensitivity of the warning in variable operating condition scenarios, reducing false alarms and false negatives, and ensuring that the dynamic warning threshold reflects the equipment operating status in real time.

[0072] The adaptive threshold component is corrected using historical early warning response feedback data to obtain historical feedback correction values. These historical feedback correction values ​​are then superimposed with the adaptive threshold component to obtain the dynamic early warning threshold, expressed as: ; in, Indicates the dynamic early warning threshold. This indicates the preset baseline warning threshold. Indicates the feedback gain coefficient. This represents the feedback integral time constant. Indicating a historical moment The coupling risk index, This represents the feedback memory time constant.

[0073] It should be noted that, It is preset based on the historical operating data of the equipment under standard operating conditions. It is obtained by collecting the cumulative current deviation and cumulative heat growth of multiple monitoring cycles under standard operating conditions, calculating the product of the mean of the normalized ratio of the two and the reference coefficient. The exemplary value range is 1.5 to 3.0. It refers to the numerical value representing the fault risk intensity at that time, obtained through the circulating current-thermal coupling correlation function based on the cumulative current deviation and cumulative heat growth at the corresponding historical moment. It is a configuration parameter that characterizes the attenuation rate of the weight of historical early warning response feedback data, and is tuned according to dynamic characteristics and the validity period of historical data; the feedback gain coefficient is a dimensionless parameter that characterizes the weight of the impact of historical early warning response feedback data on the dynamic early warning threshold correction, and is preset according to the response sensitivity requirements of early warning equipment to historical errors or obtained through iterative optimization based on operational monitoring data feedback.

[0074] The specific process includes: extracting the coupled risk index corresponding to the historical early warning and response records from the historical early warning and response feedback data; calculating the difference between the coupled risk index and the preset benchmark early warning threshold; using the reciprocal of the feedback integral time constant and the difference to calculate the normalized deviation; multiplying the normalized deviation with the feedback gain coefficient to obtain the instantaneous feedback quantity; using the exponential decay function to weight the time interval from the time point of the historical early warning and response records to the current time; multiplying the instantaneous feedback quantity with the weighting coefficient to obtain the weighted feedback value; integrating the weighted feedback values ​​corresponding to all historical early warning and response records in the time dimension to obtain the historical feedback correction value; and adding the historical feedback correction value with the working condition adaptive threshold component to obtain the dynamic early warning threshold.

[0075] like Figure 4 The graph showing the dynamic warning threshold changes under different operating conditions illustrates that the dynamic warning thresholds drift differently with the change of the operating condition correction factor under standard operating conditions, high ambient temperature conditions, high cumulative charge-discharge cycles conditions, high real-time load rate conditions, and composite variable operating conditions. The threshold adjustment is more pronounced under composite variable operating conditions, while the thresholds under high ambient temperature conditions, high cumulative charge-discharge cycles conditions, and high real-time load rate conditions also show an orderly change trend matching the corresponding operating states. This graph demonstrates that the dynamic warning threshold in this invention is not a fixed constant, but rather can be adaptively corrected based on ambient temperature, cumulative charge-discharge cycles, and real-time load rate, and continuously adjusted by combining historical warning handling feedback data. This verifies the technical effectiveness of this invention in balancing robustness and sensitivity, and reducing false alarms and false negatives in variable operating condition scenarios.

[0076] S5. The coupling risk index is compared with the dynamic early warning threshold in real time, and combined with the duration filtering mechanism, a graded early warning instruction is generated.

[0077] By comparing the coupled risk index with the dynamic early warning threshold in real time, a risk exceedance state sequence is obtained.

[0078] The specific process includes comparing the coupled risk index with the dynamic early warning threshold in real time to obtain a risk exceedance state sequence. By collecting the coupled risk index of the monitored object in real time, the coupled risk index is compared with the dynamic early warning threshold in each data collection cycle. When the coupled risk index is greater than the dynamic early warning threshold, it is determined that the current moment is in a risk exceedance state and marked as the first state identifier. When the coupled risk index is less than or equal to the dynamic early warning threshold, it is determined that the current moment is in a safe state and marked as the second state identifier. The first state identifier or the second state identifier corresponding to each moment is arranged and combined in chronological order to form a risk exceedance state sequence containing continuous time dimension state identifiers, thereby reflecting the changes in the risk exceedance status of the monitored object at different times.

[0079] By using a duration filtering mechanism to count and maintain continuous states within a time window of the risk exceedance state sequence, an effective warning continuity indicator is obtained.

[0080] The specific process includes: sliding the risk exceedance state sequence through a fixed-length time window; counting the occurrences of the first state identifier within each time window to obtain a continuous state count value; comparing the continuous state count value with a preset duration threshold; when the continuous state count value is greater than or equal to the preset duration threshold, determining that the risk exceedance state in the current time window meets the continuous warning condition and generating a valid warning duration flag; when the continuous state count value is less than the preset duration threshold, determining that the risk exceedance state in the current time window is an instantaneous fluctuation and ignoring the state; repeating the continuous state counting and holding processing operation for each sliding time window; and outputting a sequence of valid warning duration flags for each time window.

[0081] It should be noted that the duration filtering mechanism refers to the logical rule of counting and maintaining continuous states in the risk exceedance state sequence by setting a time window to identify valid warning duration indicators that meet the continuous warning conditions. The duration threshold is preset based on the duration of valid warning events in historical warning response feedback data. By analyzing the false alarm rate and false negative rate corresponding to different durations in historical warning response feedback data, the duration with the optimal combination of false alarm rate and false negative rate is selected as the preset duration threshold. An exemplary value range is 3 to 7 times. The warning condition refers to the state indicated by the valid warning duration indicator meeting the requirements for triggering an alarm. The duration filtering mechanism counts and maintains continuous states in the risk exceedance state sequence within a time window. When the continuous state count value reaches the preset duration threshold, a valid warning duration indicator is generated, thereby establishing the warning condition.

[0082] The risk level is determined by matching the valid warning continuity indicator with the preset multi-level confidence threshold, and a graded warning instruction is generated.

[0083] The specific process includes reading the continuous status count value corresponding to the valid warning continuity indicator, comparing the continuous status count value with each threshold level in the preset multi-level confidence threshold, determining the current risk status as the highest risk level and generating a first-level warning instruction when the continuous status count value is greater than or equal to the highest level confidence threshold, determining the current risk status as the highest risk level and generating a second-level warning instruction when the continuous status count value is between the intermediate level confidence threshold, determining the current risk status as the medium risk level and generating a second-level warning instruction when the continuous status count value is between the lowest level confidence threshold and the intermediate level confidence threshold, determining the current risk status as the low risk level and generating a third-level warning instruction, and outputting a graded warning instruction containing risk level identification and handling suggestions according to the handling priority corresponding to different risk levels.

[0084] It should be noted that the multi-level confidence thresholds are preset based on the urgency and false alarm tolerance corresponding to different risk levels in historical early warning and response feedback data. By analyzing the response time and false alarm frequency of risk events at all levels in historical early warning and response feedback data, risk levels with high urgency and low false alarm tolerance are set as high-level confidence thresholds, and risk levels with low urgency and high false alarm tolerance are set as low-level confidence thresholds. The exemplary value ranges are 3 to 5 times for low-level confidence thresholds, 6 to 8 times for medium-level confidence thresholds, and 9 to 12 times for high-level confidence thresholds.

[0085] like Figure 5 The time-series diagram of the early warning indicators shows that as the fault branch gradually evolves from early current deviation to thermal anomaly, the current imbalance sequence first shows a continuous increase, followed by a synchronous increase in the power device temperature rise rate sequence. The coupling risk index rapidly approaches and exceeds the dynamic early warning threshold, thus forming an effective and continuous early warning indicator. This diagram intuitively reflects that the present invention does not rely on a single instantaneous signal for judgment, but rather achieves early identification of progressively hidden faults through the continuous evolution of the current imbalance sequence, power device temperature rise rate sequence, cumulative current deviation, cumulative thermal growth, and coupling risk index. This verifies that the present invention has the technical effect of improving the sensitivity of early fault identification and extending the effective early warning window.

[0086] S6. Execute linkage control actions according to the graded early warning instructions and collect operation monitoring data, and update the correction parameters in the dynamic early warning threshold based on the feedback of operation monitoring data.

[0087] Based on the graded early warning instructions, the linkage control actions corresponding to each risk level are determined, and the corresponding parallel branches are subjected to de-rate control and flow equalization adjustment in accordance with the linkage control actions to obtain the disposal execution status.

[0088] The specific process includes: determining the linkage control actions corresponding to each risk level based on the graded early warning instructions; performing derating control and current sharing adjustment on the corresponding parallel branches according to the linkage control actions to obtain the disposal execution status; parsing the risk level identifier in the graded early warning instructions and searching for the linkage control action matching the risk level identifier from the preset control action mapping table; when the linkage control action is derating control, reducing the output power of the corresponding parallel branch according to the preset derating ratio; when the linkage control action is current sharing adjustment, adjusting the current distribution ratio of the corresponding parallel branch to make the current of each branch tend to be balanced; collecting the output power and current distribution parameters of the parallel branches in real time; judging the execution effect of derating control and current sharing adjustment based on the changes in output power and current distribution parameters; marking the execution effect as the disposal execution status and outputting it.

[0089] It should be noted that the control action mapping table is preset based on historical fault handling records. By statistically analyzing the most effective linkage control actions under different risk levels, a key-value pair mapping relationship between risk levels and optimal linkage control actions is established, forming a control action mapping table containing risk level identifiers and linkage control actions. The derating ratio is preset based on the safe operation parameters and handling effect data in the historical fault handling records. By analyzing the maximum allowable power reduction within the safe operation range of parallel branches under each risk level, and combining the fault suppression success rate, the optimal derating ratio that balances safe operation and fault suppression is selected as the corresponding derating ratio.

[0090] Based on the execution status of the response, the parallel branches after the linkage control action is executed are continuously monitored, and the corresponding operating status parameters and response information are collected to form operation monitoring data.

[0091] The specific process includes: continuously collecting the output power, current distribution, and voltage fluctuation amplitude of the parallel branch as operating status parameters by reading the parallel branch operating parameters in real time, and collecting the power change rate and current balance change trend of the parallel branch after the execution of the linkage control action as handling response information. The operating status parameters and handling response information are packaged and integrated according to a preset time interval to generate and output operating monitoring data containing timestamps, operating status parameters, and handling response information.

[0092] It should be noted that the preset time interval is based on the frequency of change of parallel branch operating status parameters and the update cycle of response information in historical operation monitoring data. By analyzing the fluctuation cycle of operating status parameters and the effective feedback time of response information in historical operation monitoring data, the optimal acquisition cycle that can fully capture parameter changes and avoid data redundancy is taken as the preset time interval.

[0093] The degree to which coordinated control actions suppress risk changes is determined by using operational monitoring data, and the correction parameters in the dynamic early warning threshold are updated based on the degree of suppression.

[0094] The specific process includes extracting operational status parameters and response information from operational monitoring data, obtaining the difference in risk indicators before and after the execution of the linkage control action, comparing the difference with the preset risk suppression benchmark value to obtain the degree of suppression of risk changes by the linkage control action, adjusting the correction parameter in the dynamic early warning threshold when the suppression degree is higher than the risk suppression benchmark value, and adjusting the correction parameter in the dynamic early warning threshold when the suppression degree is lower than the risk suppression benchmark value, thus completing the feedback update of the correction parameter in the dynamic early warning threshold.

[0095] It should be noted that the risk suppression reference value is preset based on the ideal suppression effect data of linkage control actions under different risk levels in historical operation monitoring data. By analyzing the distribution of the difference in risk indicators corresponding to each risk level in historical operation monitoring data, a critical difference value that can effectively distinguish whether the linkage control action has achieved the expected suppression effect is selected as the risk suppression reference value.

[0096] For example Figure 6 The linkage control effect diagram shows that after the hierarchical early warning instruction is triggered, the power reduction control and current sharing regulation are sequentially executed, the output power enters the controlled decline range, the change trend of the current balance degree continues to improve, and the coupling risk index gradually drops from a high level to near the safe zone; at the same time, the operating state parameters and disposal response information after disposal continue to be used as operation monitoring data to participate in the update of the correction parameters in the dynamic early warning threshold, forming a closed-loop of early warning, control and feedback. This diagram shows that the present invention can not only identify fault risks in advance, but also effectively suppress the risk changes through linkage control actions after early warning, and achieve continuous correction of the threshold through feedback update, further verifying the effect of the double closed-loop adaptive adjustment and self-evolution of the present invention.

[0097] In summary, the present invention realizes the non-linear coupling characterization of electrical and thermal stresses by calculating the electro-thermal cumulative amount and constructing a coupling correlation function, achieving the effects of improving the sensitivity of early hidden fault identification, extending the effective early warning window and avoiding missed fault detection; through the double closed-loop correction of the threshold by working condition correction and historical feedback, the double closed-loop adaptive adjustment and self-evolution of the early warning boundary are realized, achieving the effects of effectively balancing robustness and sensitivity and reducing the false alarm and missed alarm rates in variable working condition scenarios.

[0098] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A fault early warning method for energy storage converters based on intelligent sensors, characterized in that, include: Collect the branch current, power device temperature and operating parameters of each parallel branch to form the original multi-source time series dataset; Based on the original multi-source time series dataset, the current imbalance sequence and power device temperature rise rate sequence of each parallel branch are calculated, and the ambient temperature, cumulative charge and discharge cycles and real-time load rate are extracted from the original multi-source time series dataset as operating condition correction factors. Based on the current imbalance sequence and the power device temperature rise rate sequence, the cumulative current deviation and cumulative heat growth within the corresponding monitoring period are calculated, and the cumulative current deviation and cumulative heat growth are combined according to the branch number to generate a branch-level coupling characteristic sequence. Based on the branch-level coupling characteristic sequence, a circulation-thermal coupling correlation function is constructed to generate a coupling risk index. The preset benchmark early warning threshold is dynamically corrected based on the operating condition correction factor and historical early warning response feedback data to obtain the dynamic early warning threshold. The coupling risk index is compared with the dynamic early warning threshold in real time, and combined with the duration filtering mechanism, a graded early warning instruction is generated. Based on the graded early warning instructions, the system executes linkage control actions and collects operational monitoring data. Based on the feedback from the operational monitoring data, it updates the correction parameters in the dynamic early warning threshold.

2. The fault early warning method for energy storage converter based on intelligent sensors as described in claim 1, characterized in that, The specific steps for forming the original multi-source time-series dataset are as follows: The energy storage converter is divided into branches according to the parallel branches, and the corresponding branch current, power device temperature and operating condition parameters are collected in sequence based on each branch to obtain the original sampling parameters of each parallel branch. The original sampling parameters of each parallel branch are integrated to form the original multi-source time series dataset.

3. The fault early warning method for energy storage converters based on intelligent sensors as described in claim 2, characterized in that, The specific steps for calculating the current imbalance sequence and power device temperature rise rate sequence of each parallel branch based on the original multi-source time series dataset are as follows: The total current sequence is obtained by summarizing the current values ​​of each parallel branch from the original multi-source time series dataset, and the total current sequence is averaged to obtain the theoretical average current sequence. Obtain the deviation ratio between the current value of each parallel branch and the theoretical average current sequence, obtain the current imbalance at each time point, and form a current imbalance sequence. Based on the current imbalance sequence, the corresponding power device temperature data is extracted from the original multi-source time series dataset, and the temperature change slope is obtained by fitting through a sliding window to form the power device temperature rise rate sequence.

4. The fault early warning method for energy storage converter based on intelligent sensors as described in claim 3, characterized in that, The specific steps for extracting ambient temperature, cumulative charge / discharge cycles, and real-time load rate as operating condition correction factors from the original multi-source time-series dataset are as follows: Ambient temperature, cumulative charge / discharge cycles, and real-time load rate are read from the original multi-source time-series dataset, and the real-time load rate is normalized to form a working condition correction source data group. The environmental degradation factor, aging degradation factor, and load deviation factor are calculated based on the operating condition correction source data set, and the operating condition correction factor is generated by aligning and encapsulating the data according to the time index.

5. The fault early warning method for energy storage converter based on intelligent sensors as described in claim 4, characterized in that, The specific steps for calculating the cumulative current deviation and cumulative heat growth within the corresponding monitoring period based on the current imbalance sequence and the power device temperature rise rate sequence are as follows: Each monitoring period is determined by combining the current imbalance sequence and the power device temperature rise rate sequence, and the current imbalance data and power device temperature rise rate data corresponding to each monitoring period are extracted. By using current imbalance data and power device temperature rise rate data, the branch interaction change trajectory is constructed, and the interaction characteristic quantity between current imbalance change and temperature rise change is obtained. Interactive features are used to identify continuous thermal growth closed sections within each monitoring period. Based on current imbalance data, power device temperature rise rate data, interactive features, and continuous thermal growth closed sections, the cumulative current deviation and cumulative thermal growth are calculated.

6. The fault early warning method for energy storage converter based on intelligent sensors as described in claim 1 or 5, characterized in that, The specific steps for generating the branch-level coupled feature sequence are as follows: Branch numbers are extracted from the cumulative current deviation and cumulative heat growth, and the cumulative current deviation and cumulative heat growth are paired one-to-one under electrothermal correlation constraints according to the branch numbers to generate an electrothermal coupling feature vector. Based on the electrothermal coupling feature vectors arranged continuously in the time sequence of the monitoring cycle, a branch-level coupling feature sequence is generated.

7. The fault early warning method for energy storage converter based on intelligent sensors as described in claim 1, characterized in that, The specific steps for generating the coupling risk index are as follows: Electrothermal coupling feature vectors are extracted from the branch-level coupling feature sequence, and causal relationships between branches are established based on the changes in electrothermal coupling features of each parallel branch in adjacent monitoring periods. By utilizing the causal relationships between branches, the electrothermal coupling feature vector is continuously mapped to generate the circulating-thermal coupling response trajectory. Based on the circulation-thermal coupling response trajectory, a circulation-thermal coupling correlation function is constructed, and the coupling risk index is calculated based on the circulation-thermal coupling correlation function.

8. The fault early warning method for energy storage converter based on intelligent sensors as described in claim 1 or 7, characterized in that, The specific steps for obtaining the dynamic early warning threshold are as follows: The operating condition correction factor is subjected to parameter mapping and time serialization to obtain the operating condition adjustment parameter sequence. Using the sequence of operating condition adjustment parameters, a threshold drift function is constructed and an adaptive threshold component for the current operating condition is generated. The adaptive threshold component of the operating condition is corrected by using historical early warning and response feedback data to obtain historical feedback correction values. The historical feedback correction values ​​are then superimposed with the adaptive threshold component of the operating condition to obtain the dynamic early warning threshold.

9. The fault early warning method for energy storage converter based on intelligent sensors as described in claim 1, characterized in that, The specific steps for generating tiered early warning instructions are as follows: The risk index is compared with the dynamic early warning threshold in real time to obtain the risk exceedance state sequence; By using a duration filtering mechanism to count and maintain continuous states within a time window of the risk exceedance state sequence, an effective warning persistence indicator is obtained. The risk level is determined by matching the valid warning continuity indicator with the preset multi-level confidence threshold, and a graded warning instruction is generated.

10. The fault early warning method for energy storage converter based on intelligent sensors as described in claim 9, characterized in that, The steps for executing linkage control actions based on graded early warning instructions and collecting operational monitoring data, and updating the correction parameters in the dynamic early warning threshold based on the feedback from the operational monitoring data, are as follows: Based on the graded early warning instructions, the linkage control actions corresponding to each risk level are determined, and the corresponding parallel branches are subjected to de-rate control and flow equalization adjustment in accordance with the linkage control actions to obtain the disposal execution status; Based on the execution status of the response, the parallel branches after the linkage control action is executed are continuously monitored, and the corresponding operating status parameters and response information are collected to form operating monitoring data. The degree to which coordinated control actions suppress risk changes is determined by using operational monitoring data, and the correction parameters in the dynamic early warning threshold are updated based on the degree of suppression.

Images