A detection analysis method and system for an energy storage system

Abstract

The application relates to a detection analysis method and system for an energy storage system, aiming to improve system state monitoring accuracy, fault early warning capability and response intelligent level. The method first synchronously collects and transmits multi-source data in real time through a sensor network deployed in a battery module, a power conversion unit and a cooling system. Then, key dynamic characteristics are extracted from the multi-source data to comprehensively form dynamic characteristic state information of the energy storage system. The state information is input into a pre-trained deep learning model to output a health state score and a fault probability of the system. Based on the evaluation results and fault tracing information, the system triggers a multi-level early warning mechanism and executes an active response strategy. After the response is executed, the system re-evaluates the health state in a closed loop to verify the response effect. If the improvement does not reach the expectation, an artificial diagnosis protocol is started. The application realizes a complete process from dynamic perception, intelligent diagnosis, accurate response to closed loop verification, and improves the safety and reliability of the energy storage system operation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric energy storage, in particular to a detection analysis method and system for an energy storage system. BACKGROUND

[0002] As a key link of energy network, the safe, stable and efficient operation of the energy storage system is of great importance. The current state monitoring and fault diagnosis technology for the energy storage system generally has the problems of single monitoring dimension and delayed response. Traditional methods mostly rely on single-point sensor data or preset static thresholds for judgment, which is difficult to effectively capture the rapidly changing dynamic characteristics in the charging and discharging process, and is also unable to perform real-time synchronization and collaborative analysis on multi-source data, resulting in insufficient recognition accuracy of early potential faults and complex failure modes, and making it difficult to achieve accurate health state evaluation.

[0003] In terms of health state diagnosis, the existing technology mostly adopts methods based on empirical rules or simple statistical models, which have limited representation ability of dynamic changing characteristics. These methods are difficult to fully exploit and utilize the high-dimensional, nonlinear dynamic characteristic information captured by the sensor network, and are unable to establish deep mapping relationships between these dynamic characteristics and internal degradation states and potential fault types of the system, which restricts the accuracy of health state evaluation and the early fault warning capability, and makes it difficult to support accurate fault tracing.

[0004] In addition, the existing fault response mechanism of the energy storage system is often passive and isolated. There is a lack of close closed-loop linkage and effect verification between the diagnosis results and subsequent control strategies and protection actions. After the system issues a warning or performs a protection action, there is a lack of real-time evaluation mechanism for the effectiveness of the response measures, and the actual effect after the strategy execution cannot be dynamically adjusted or decision upgraded, which affects the overall self-healing ability and operation reliability of the system, and there is an urgent need for a detection analysis method that can realize a complete chain of "dynamic perception-intelligent diagnosis-precise response-closed-loop verification". SUMMARY

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0006] According to the first aspect of the present application, a detection analysis method for an energy storage system is claimed, comprising:

[0007] S1, acquiring a sensor network of the energy storage system, and synchronizing multi-source energy storage data in real time in the sensor network;

[0008] S2, extracting dynamic characteristic data of the multi-source energy storage data, and obtaining dynamic characteristic state information of the energy storage system based on the dynamic characteristics;

[0009] S3, inputting the dynamic characteristic state information into a pre-trained deep learning model to obtain a health state of the energy storage system;

[0010] S4, obtaining fault tracing information of the energy storage system based on the health state and prewarning the energy storage system to perform active response;

[0011] S5, reevaluating the health state of the energy storage system after the active response and deciding whether to start an artificial diagnosis protocol.

[0012] Further, the S1 further comprises:

[0013] deploying a sensor network on a battery module, a power conversion unit and a cooling system of the energy storage system;

[0014] synchronizing current data, voltage data and temperature data collected in the sensor network in real time, and a sampling frequency is not less than 1 kHz;

[0015] The S2 further comprises:

[0016] extracting high-frequency fluctuation characteristics in a current waveform of the current data to identify abnormal oscillation modes in a charging and discharging process of the energy storage system;

[0017] calculating an instantaneous drop depth of the voltage data when a load mutates and marking a drop duration;

[0018] monitoring temperature field distribution of the temperature data to generate a thermal gradient space map;

[0019] The S3 further comprises:

[0020] inputting the dynamic characteristic state information into a pre-trained deep learning model to output a health state score SOH and a failure probability;

[0021] if the failure probability exceeds a preset threshold, triggering a three-level prewarning mechanism;

[0022] The S4 further comprises:

[0023] locating a power converter defect according to the high-frequency fluctuation characteristics to generate a power reduction operation instruction;

[0024] locating a battery aging area in combination with voltage drop and the thermal gradient space map to dynamically adjust a charging curve;

[0025] The S5 further comprises:

[0026] closed-loop verification and optimization, reevaluating the SOH score after executing a response strategy, and starting an artificial diagnosis protocol if improvement does not reach an expectation.

[0027] Further, the S1 further comprises:

[0028] At the switching moment of the charging and discharging state, a transient process capture mode is started, the sampling frequency is increased to more than 10 kHz, and the whole switching process is continuously covered;

[0029] A transient capture trigger condition is added, and the charging and discharging state switching signal is monitored in real time, including the action current of the contactor;

[0030] The 10 kHz high-speed sampling is started 5 ms before the switching occurs and lasts for 100 ms after the switching is completed;

[0031] The transient data segment is stored and the transient data segment is marked as a switching event data set.

[0032] Further, the S2 extracts the high-frequency fluctuation characteristics in the current waveform of the current data, identifies the abnormal oscillation mode in the charging and discharging process of the energy storage system, and further comprises:

[0033] A multi-scale sliding window analysis method is used to identify the current abnormal oscillation in the charging and discharging process, and the size of the window is adaptively adjusted according to the charging and discharging rate;

[0034] The multi-scale sliding window is analyzed, and three window sizes are set, including:

[0035] Small window, 10 ms, used to capture microsecond-level peaks;

[0036] Medium window, 100 ms, used to analyze fluctuation period;

[0037] Large window, 1 s, used to count oscillation energy;

[0038] According to the real-time charging and discharging rate, a window combination is dynamically selected, when the charging and discharging rate is high >1, the small window is emphasized, and when the charging and discharging rate is low <0.5, the large window is enabled.

[0039] Further, the S2 monitors the temperature field distribution of the temperature data, and generates a thermal gradient space map, further comprising:

[0040] A three-dimensional thermal map is constructed by a distributed temperature sensor array, and the coordinates of the region where the temperature gradient exceeds the safety threshold are labeled;

[0041] When constructing the three-dimensional thermal map, the temperature sensor array is arranged in the XYZ direction of the battery module, and the interval is ≤2 cm;

[0042] A continuous temperature field cloud image is generated by a Kriging interpolation algorithm, and the region where the temperature difference between adjacent sensors is ≥5℃ is automatically circled as a high-risk area, and the center coordinates are output;

[0043] The deep learning model of S3 further comprises:

[0044] The architecture adopts a fusion of a time convolution network and an attention mechanism, and the input features include current fluctuation patterns, voltage drop sequences, and thermal gradient distributions;

[0045] The architecture of the deep learning model further comprises a time convolution network layer for extracting time-varying features of the current or voltage of the energy storage system;

[0046] A spatial attention module is used to focus on the gradient mutation area in the three-dimensional thermal map;

[0047] According to the correlation of the space-time abnormal patterns by the feature fusion layer, an SOH score and a fault classification label are output.

[0048] Further, the method further comprises:

[0049] When the three-dimensional thermal map shows a local high-temperature area and a space-time correlation of voltage drop events, the deep learning model outputs a battery internal short-circuit fault label of the energy storage system and starts an emergency isolation program;

[0050] When the three-dimensional thermal map shows a local temperature rise rate > 1℃ / s and is accompanied by a voltage instantaneous drop > 10%, it is marked as a thermal-electric coupling abnormal event;

[0051] The deep learning model outputs an internal short-circuit diagnosis result, and controls the relay to isolate the corresponding module within 500ms.

[0052] Further, the dynamic adjustment charging curve of S4 further comprises:

[0053] A segmented slow charging strategy is implemented for the battery in the aging area, and a constant current is charged to 70% capacity for the first stage;

[0054] A constant voltage mode is switched for the second stage until full charging;

[0055] A segmented slow charging strategy is adopted, a constant current is charged for the first stage, the current value is set to 0.3 times the rated capacity, and switching is performed when 70% state of charge is reached;

[0056] For the second stage, constant voltage charging is performed, and the cut-off voltage is maintained until the current decays to a preset current value.

[0057] Further, the three-level early warning mechanism of S3 further comprises:

[0058] For the first level of warning, the fault probability is 30%-50%, the feature data is recorded and the risk point is marked, the abnormal feature segment is stored locally, and the system status light turns yellow;

[0059] For the second level of early warning, the failure probability is 50%-70%, the maximum charging and discharging rate is limited, the charging and discharging current is limited to not more than 0.5 times the rated value, and the early warning information is pushed to the monitoring platform;

[0060] For the third level of early warning, the failure probability is >70%, the fault unit is cut off and the alarm information is pushed, the DC contactor is disconnected, the sound and light alarm is activated, and the fault positioning coordinates are synchronously sent.

[0061] Further, the S5 further comprises:

[0062] A strategy effect evaluation matrix is established, the SOH change rate and the fault recurrence frequency index before and after optimization are compared, and the response strategy parameters are automatically corrected;

[0063] The construction elements of the strategy effect evaluation matrix include core indexes and automatic correction rules;

[0064] The core indexes include the SOH score change amount, the fault code recurrence times, and the temperature extreme value drop amplitude;

[0065] The automatic correction rule includes increasing the current limit value of the charging strategy by 5% if the SOH improvement is <3% in two consecutive evaluations.

[0066] According to the second aspect of the present application, the present application claims to protect a detection analysis system for an energy storage system, comprising:

[0067] One or more processors;

[0068] A memory having one or more programs stored thereon, when the one or more programs are executed by the one or more processors, so that the one or more processors implement the detection analysis method for the energy storage system.

[0069] The present application relates to a detection analysis method and system for an energy storage system, aiming to improve the system state monitoring accuracy, fault early warning capability and response intelligent level. The method first synchronously collects and transmits multi-source data in real time through the sensor network deployed in the battery module, power conversion unit and cooling system. Then, the key dynamic features are extracted from the multi-source data to form the dynamic feature state information of the energy storage system. The state information is input into the pre-trained deep learning model to output the health state score and the failure probability of the system. Based on the evaluation results and the fault tracing information, the system triggers a multi-level early warning mechanism and executes an active response strategy. After the response is executed, the system re-evaluates the health state in a closed loop to verify the response effect; if the improvement does not meet the expectation, the artificial diagnosis protocol is started. The present application realizes a complete process from dynamic perception, intelligent diagnosis, accurate response to closed loop verification, and significantly improves the safety and reliability of the energy storage system operation. BRIEF DESCRIPTION OF DRAWINGS

[0070] Figure 1 A workflow diagram of a detection analysis method for a storage system as claimed in the embodiments of the present application;

[0071] Figure 2 A structural diagram of a detection analysis system for a storage system as claimed in the embodiments of the present application. DETAILED DESCRIPTION

[0072] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0073] The terms "first", "second", "third" in the present application are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second", "third" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality of" is at least two, such as two, three, etc., unless otherwise explicitly and specifically limited. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present application are only used to explain the relative position relationship, movement condition, etc. between components in a certain posture (as shown in the drawings), and if the certain posture changes, the directional indications also change accordingly. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, product or device.

[0074] In this document, the term "embodiment" means that the specific features, structures or characteristics described in connection with the embodiment can be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily mean that the same embodiment is referred to, nor does it mean that independent or alternative embodiments are mutually exclusive or alternative to other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0075] According to the first embodiment of the present application, the present application claims a detection analysis method for a storage system, with reference to Figure 1 , comprising:

[0076] S1, acquire a sensor network of the energy storage system, in which multi-source energy storage data are synchronously collected in real time;

[0077] S2, extract dynamic characteristic data of the multi-source energy storage data, and obtain dynamic characteristic state information of the energy storage system based on the dynamic characteristic data;

[0078] S3, input the dynamic characteristic state information into a pre-trained deep learning model to obtain a health state of the energy storage system;

[0079] S4, acquire fault tracing information of the energy storage system based on the health state, and prewarn the energy storage system to perform active response;

[0080] S5, reevaluate the health state of the energy storage system after the active response, and decide whether to start an artificial diagnosis protocol.

[0081] In this embodiment, a 20 kWh lithium battery energy storage system is detected and analyzed;

[0082] The system is configured to include 16 groups of 3.2 V / 100 Ah lithium iron phosphate battery modules, and form a 400 V direct current system.

[0083] A zero-flux Hall sensor LEM CKSR 200-P is arranged at the positive and negative poles of the bus, a double-layer electromagnetic shielding shell is used, the accuracy is ±0.1%, and current monitoring is performed

[0084] Eight optical fiber sensors Opsens OTG-M are attached to the surface of each module, wave division multiplexing single fiber transmission is used, the spatial positioning accuracy is 0.3 cm, and temperature monitoring is performed

[0085] The control host uses an industrial-grade industrial computer Intel i7 / 16 GB RAM to run the detection system.

[0086] Further, the S1 further includes:

[0087] The sensor network is arranged on the battery module, the power conversion unit and the cooling system of the energy storage system;

[0088] The current data, the voltage data and the temperature data collected in the sensor network are synchronously collected in real time, and the sampling frequency is not less than 1 kHz;

[0089] The S2 further includes:

[0090] The high-frequency fluctuation characteristics in the current waveform of the current data are extracted, and abnormal oscillation modes in the charging and discharging process of the energy storage system are identified;

[0091] The instantaneous drop depth of the voltage data when the load suddenly changes is calculated, and the drop duration is marked.

[0092] monitoring the temperature field distribution of the temperature data, generating a thermal gradient spatial map;

[0093] The S3 further comprises:

[0094] inputting the dynamic feature state information into a pre-trained deep learning model, outputting a health state score SOH and a failure probability;

[0095] If the failure probability exceeds a preset threshold, triggering a three-level early warning mechanism;

[0096] The S4 further comprises:

[0097] According to the positioning power converter defect of the high-frequency fluctuation feature, generating a reduced power operation instruction;

[0098] Combine the voltage drop with the thermal gradient spatial map to locate the battery aging area, and dynamically adjust the charging curve;

[0099] The S5 further comprises:

[0100] Closed-loop verification optimization, after executing the response strategy, re-evaluate the SOH score, if the improvement does not reach the expected, start the artificial diagnosis protocol.

[0101] Further, the S1 further comprises:

[0102] At the moment of switching the charging and discharging state, start the transient process capture mode, increase the sampling frequency to more than 10kHz, and continuously cover the whole switching process;

[0103] Add transient capture trigger conditions, real-time monitor the charging and discharging state switching signal, including the action current of the contactor;

[0104] Start 10kHz high-speed sampling 5ms before switching occurs, and continue for 100ms after switching is completed;

[0105] Store the transient data segment and associate the transient data segment with the switching event data set.

[0106] Further, in the S2, the high-frequency fluctuation feature in the current waveform of the current data is extracted, and the abnormal oscillation mode in the charging and discharging process of the energy storage system is identified, further comprising:

[0107] Adopting multi-scale sliding window analysis method to identify the current abnormal oscillation in the charging and discharging process, the size of the window is adaptively adjusted according to the charging and discharging rate;

[0108] Analyzing the multi-scale sliding window, setting three window sizes, including:

[0109] Small window, 10 ms, for capturing microsecond-level spikes;

[0110] Medium window, 100 ms, for analyzing fluctuation periods;

[0111] Large window, 1 s, for statistical oscillation energy;

[0112] According to the real-time charge / discharge rate, the window combination is dynamically selected, when the charge / discharge rate is high rate > 1, the small window is focused, and when the charge / discharge rate is low rate < 0.5, the large window is enabled.

[0113] In this embodiment, the current, voltage, and temperature data are continuously collected during the synchronous collection of multi-source data, and the reference sampling rate is 1 kHz;

[0114] The transient capture trigger is used to monitor the contactor control signal, and the 10 kHz high-speed sampling is automatically switched to 5 ms before the charge / discharge switching and lasts for 105 ms;

[0115] The switching event data set is stored.

[0116] During dynamic feature extraction, the current analysis includes high rate working condition (1.5 times discharge), and the 10 ms small window is enabled: the oscillation spike of period 0.5 ms is captured, and the amplitude is 15 A;

[0117] Medium window 100 ms analysis: identify abnormal fluctuation frequency group 210 Hz / 430 Hz;

[0118] For voltage analysis, when the load suddenly increases from 50 kW to 80 kW, record the starting time T1 and the voltage transient drop of 42 V at the recovery time T1+120 ms.

[0119] Further, the temperature field distribution of the temperature data monitored in S2 is generated to generate a thermal gradient space map, which further comprises:

[0120] A three-dimensional thermal map is constructed by a distributed temperature sensor array, and the coordinates of the area where the temperature gradient exceeds the safety threshold are labeled;

[0121] When constructing the three-dimensional thermal map, the temperature sensor array is arranged in the XYZ direction of the battery module, and the interval is ≤2 cm;

[0122] A continuous temperature field cloud picture is generated by the Kriging interpolation algorithm, the area where the temperature difference of adjacent sensors is ≥5℃ is automatically circled as a high-risk area, and the center coordinates are output;

[0123] In this embodiment, when constructing the three-dimensional thermal map of the module, the sensor array interval is 1.8c, the high temperature area coordinates X3-Y7-Z1 are circled, and the temperature difference is 7.2℃;

[0124] The thermal gradient direction is marked as decreasing from the cooling inlet to the outlet.

[0125] The deep learning model of S3 further comprises:

[0126] An architecture that fuses a time convolution network and an attention mechanism, input features including current fluctuation patterns, voltage drop sequences, and thermal gradient distributions;

[0127] The architecture of the deep learning model further comprises a time convolution network layer for extracting time-varying features of the current or voltage of the energy storage system;

[0128] A spatial attention module is used to focus on the gradient mutation area in the three-dimensional thermal map;

[0129] According to the spatiotemporal abnormal pattern associated by the feature fusion layer, an SOH score and a fault classification label are output.

[0130] In this embodiment, the input of the deep learning model is:

[0131] Current: 210Hz oscillation pattern, extracted by a time convolution network;

[0132] Voltage: 42V drop event sequence;

[0133] Temperature: X3-Y7-Z1 region thermal map, focused by a spatial attention module;

[0134] Output results:

[0135] SOH score: 83%, downward trend;

[0136] Fault probability: 68%, triggering a secondary warning.

[0137] Further, the method further comprises:

[0138] When the three-dimensional thermal map shows a local high-temperature area and a spatiotemporal correlation with a voltage drop event, the deep learning model outputs a battery internal short circuit fault label of the energy storage system and starts an emergency isolation program;

[0139] When the three-dimensional thermal map shows a local temperature rise rate >1℃ / s and is accompanied by a voltage instantaneous drop >10%, it is marked as a thermal-electric coupling abnormal event;

[0140] The deep learning model outputs an internal short circuit diagnosis result, and controls a relay to isolate the corresponding module within 500ms.

[0141] Further, the dynamic adjustment charging curve of S4 further comprises:

[0142] A segmented slow charging strategy is implemented for the battery in the aging area, and a constant current is charged to 70% capacity for the first stage;

[0143] For the second stage, switch to constant voltage mode until full charge;

[0144] Adopting the segmented slow charging strategy, for the first stage, constant current charging, the current value is set to 0.3 times the rated capacity, and switching is performed when reaching 70% state of charge;

[0145] For the second stage, constant voltage charging, maintain the cutoff voltage until the current decays to a preset current value.

[0146] In this embodiment, the power converter defect response current high frequency oscillation matches the IGBT aging characteristics, and the load is reduced to 70% of the rated power by sending instructions to the PCS;

[0147] For battery aging response, locate the X3-Y7-Z1 module, the temperature rise is 2.1℃ / s during voltage drop period, and the segmented slow charging strategy is executed;

[0148] Wherein stage 1: 0.3A constant current charging to SOC 70%, time-consuming 142min; stage 2: 3.65V constant voltage charging to current≤5A cutoff.

[0149] Further, the S3 three-level early warning mechanism further includes:

[0150] For the first level of warning, the fault probability is 30%-50%, the characteristic data is recorded and the risk point is marked, the abnormal characteristic segment is stored locally, and the system status light turns yellow;

[0151] For the second level of warning, the fault probability is 50%-70%, the maximum charge and discharge rate is limited, the charge and discharge current is limited to not more than 0.5 times the rated value, and the warning information is pushed to the monitoring platform;

[0152] For the third level of warning, the fault probability is >70%, the fault unit is cut off and the alarm information is pushed, the DC contactor is disconnected, the audible and visual alarm is activated, and the fault positioning coordinates are sent synchronously.

[0153] In this embodiment, the detection condition is that the module X2-Y5-Z3 temperature rise rate is 1.8℃ / s, accompanied by a voltage instantaneous drop of 14.3% (>10% threshold);

[0154] The response action is that the model outputs the "internal short circuit" label with a confidence of 92%;

[0155] For the third level of warning activation, the DC contactor is disconnected in this embodiment, and the action time is 380ms; the alarm information "module X2-Y5-Z3 isolation" is sent;

[0156] The audible and visual alarm is started, using 105dB buzzer + red flashing.

[0157] Further, the S5 further includes:

[0158] A policy effect evaluation matrix is established, and the SOH change rate and the fault recurrence frequency index before and after optimization are compared to automatically correct the response strategy parameters;

[0159] The construction elements of the policy effect evaluation matrix include core indexes and automatic correction rules;

[0160] The core indexes include the SOH score change amount, the fault code recurrence times, and the temperature extreme value drop amplitude;

[0161] The automatic correction rules include increasing the current limit value of the charging strategy by 5% if the SOH improvement is less than 3% in two consecutive evaluations.

[0162] According to the second embodiment of the present application, the present application claims to protect a detection analysis system for an energy storage system, comprising:

[0163] One or more processors;

[0164] A memory having one or more programs stored thereon, when the one or more programs are executed by the one or more processors, so that the one or more processors implement the detection analysis method for the energy storage system.

[0165] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units or components shown or discussed can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0166] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The above integrated unit can be realized in the form of hardware or in the form of a software functional unit. The above is only an embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process conversion using the content of the present application specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

[0167] The above detailed description of the application is merely exemplary in nature and, thus, does not limit the application. Any modification and the like that do not depart from the spirit and principles of the application are intended to be within the scope of the application.

Claims

1. A method for detecting analysis of an energy storage system, characterized by, The method comprises the following steps: S1, acquiring a sensor network of the energy storage system, and synchronizing multi-source energy storage data in real time in the sensor network; S2, extracting dynamic characteristic data of the multi-source energy storage data, and obtaining dynamic characteristic state information of the energy storage system based on the dynamic characteristic data; S3, inputting the dynamic characteristic state information into a pre-trained deep learning model to obtain a health state of the energy storage system; S4, obtaining fault tracing information of the energy storage system based on the health state, and prewarning the energy storage system to perform active response; S5, reevaluating the health state of the energy storage system after the energy storage system performs active response, and deciding whether to start an artificial diagnosis protocol; The S1 further comprises the following steps: deploying a sensor network on a battery module, a power conversion unit and a cooling system of the energy storage system; synchronizing current data, voltage data and temperature data collected in the sensor network in real time, and the sampling frequency is not less than 1 kHz; The S2 further comprises the following steps: extracting high-frequency fluctuation characteristics in a current waveform of the current data, and identifying abnormal oscillation modes in a charging and discharging process of the energy storage system; calculating an instantaneous drop depth of the voltage data when a load suddenly changes, and marking a drop duration; monitoring a temperature field distribution of the temperature data, and generating a thermal gradient space map; The S3 further comprises the following steps: inputting the dynamic characteristic state information into a pre-trained deep learning model to output a health state score SOH and a failure probability; if the failure probability exceeds a preset threshold, triggering a three-level prewarning mechanism; The S4 further comprises the following steps: locating a power converter defect based on the high-frequency fluctuation characteristics, and generating a power reduction operation instruction; locating a battery aging area based on the thermal gradient space map and the voltage drop, and dynamically adjusting a charging curve; The S5 further comprises the following steps: closed-loop verification and optimization, reevaluating the SOH score after executing a response strategy, and starting an artificial diagnosis protocol if the improvement does not meet the expectation.

2. A detection analysis method for an energy storage system according to claim 1, characterized in that, The S1 further comprises the following steps: starting a transient process capture mode at a moment when a charging and discharging state is switched, increasing the sampling frequency to greater than 10 kHz, and continuously covering the whole switching process; adding a transient capture trigger condition, monitoring a charging and discharging state switching signal in real time, and including an action current of a contactor; starting 10 kHz high-speed sampling 5 ms before the switching occurs, and continuing to 100 ms after the switching is completed; storing a transient data segment and associating the transient data segment with a switching event data set.

3. A detection analysis method for an energy storage system according to claim 2, characterized in that, The step of extracting high-frequency fluctuation characteristics in a current waveform of the current data in the S2, and identifying abnormal oscillation modes in a charging and discharging process of the energy storage system, further comprises the following steps: identifying current abnormal oscillation in the charging and discharging process by using a multi-scale sliding window analysis method, and adaptively adjusting the size of the window according to a charging and discharging rate; analyzing the multi-scale sliding window, and setting three window sizes, including: a small window, 10 ms, for capturing microsecond-level sharp peaks; a medium window, 100 ms, for analyzing fluctuation periods; a large window, 1 s, for counting oscillation energy; According to the real-time charge-discharge rate, the window combination is dynamically selected, when the charge-discharge rate is high rate >1, the small window is focused, and when the charge-discharge rate is low rate <0.5, the large window is enabled.

4. The detection analysis method for an energy storage system according to claim 2, wherein, The temperature field distribution of the temperature data monitored in S2 further comprises: A three-dimensional thermal map is constructed by a distributed temperature sensor array, and the coordinates of the region where the temperature gradient exceeds the safety threshold are marked; When constructing the three-dimensional thermal map, the temperature sensor array is arranged in the XYZ direction of the battery module, and the interval is ≤2 cm; A continuous temperature field cloud image is generated by the Kriging interpolation algorithm, and the region where the temperature difference between adjacent sensors is ≥5℃ is automatically circled as a high-risk area, and the center coordinates are output; The deep learning model of S3 further comprises: An architecture that fuses a time series convolution network with an attention mechanism is adopted, and the input features include current fluctuation patterns, voltage drop sequences, and thermal gradient distributions; The architecture of the deep learning model further comprises a time series convolution network layer that extracts time-varying features of the current or voltage of the energy storage system; A spatial attention module is used to focus on the gradient mutation area in the three-dimensional thermal map; According to the spatiotemporal anomaly pattern associated by the feature fusion layer, an SOH score and a fault classification label are output.

5. A method of detecting and analyzing for a storage system according to claim 4, wherein, Further comprising: When the three-dimensional thermal map shows a spatiotemporal correlation between local high-temperature areas and voltage drop events, the deep learning model outputs a battery internal short circuit fault label for the energy storage system and starts an emergency isolation program; When the three-dimensional thermal map shows that the local temperature rise rate is >1℃ / s and is accompanied by a voltage instantaneous drop of >10%, it is marked as a thermal-electric coupling abnormal event; The deep learning model outputs an internal short circuit diagnosis result, and controls the relay to isolate the corresponding module within 500 ms.

6. A detection analysis method for an energy storage system according to claim 2, characterized in that, The dynamic adjustment of the charging curve of S4 further comprises: Segmented slow charging strategy is implemented for the battery in the aging area, constant current charging to 70% capacity for the first stage; Constant voltage mode is switched for the second stage until full charge; Segmented slow charging strategy is adopted, constant current charging for the first stage, current value is set to 0.3 times the rated capacity, and switching is made when 70% state of charge is reached; Constant voltage charging for the second stage, maintaining the cut-off voltage until the current decays to the preset current value.

7. A detection analysis method for an energy storage system according to claim 2, characterized by, The three-level early warning mechanism of S3 further comprises: For the first level of warning, the fault probability is 30%-50%, the feature data is recorded and the risk point is marked, the abnormal feature segment is stored locally, and the system status light turns yellow; For the second level of warning, the fault probability is 50%-70%, the maximum charge-discharge rate is limited, the charge-discharge current is limited to not more than 0.5 times the rated value, and the warning information is pushed to the monitoring platform; For the third level of warning, the fault probability is >70%, the fault unit is cut off and the alarm information is pushed, the direct current contactor is disconnected, the audible and visual alarm is activated, and the fault positioning coordinates are sent synchronously.

8. The detection analysis method for an energy storage system according to claim 2, wherein, S5 further comprises: An evaluation matrix of the effect of the strategy is established, the SOH change rate and the fault recurrence frequency indicators before and after optimization are compared, and the response strategy parameters are automatically corrected; The construction elements of the strategy effect evaluation matrix include core indicators and automatic correction rules; The core indicators include SOH score change, fault code recurrence times, and temperature extreme value drop amplitude; The automatic correction rule includes increasing the current limit value of the charging strategy by 5% if the SOH is improved by less than 3% in two consecutive evaluations.

9. A detection analysis system for an energy storage system, characterized by, Comprise: One or more processors; A memory having one or more programs stored thereon, which, when executed by the one or more processors, cause the one or more processors to implement the detection analysis method for the energy storage system according to any one of claims 1 to 8.

Images