Energy storage cabinet multi-parameter intelligent monitoring device based on MEMS sensor array

Abstract

The invention relates to the technical field of energy storage cabinet safety monitoring, in particular to an energy storage cabinet multi-parameter intelligent monitoring device based on an MEMS sensor array, a sensor array unit in the energy storage cabinet multi-parameter intelligent monitoring device comprises distributed MEMS temperature and vibration sensors, the temperature sensors are installed in a non-vertical mode, detection deviation caused by direct blowing of airflow is avoided, and the energy storage cabinet multi-parameter intelligent monitoring device based on the MEMS sensor array is obtained. The airflow disturbance correction unit corrects temperature original data in real time and eliminates interference of airflow on module body temperature detection by means of a heat dissipation airflow vector field dynamic model and a compensation algorithm, and the multi-parameter coupling analysis unit processes vibration and temperature signals in a double-channel mode, extracts tab loosening high-frequency micro-vibration characteristics and correlates the temperature rising rate. Early fault logic is triggered through two-stage judgment, the fault decision and output unit fuses the correction temperature, the coupling strength coefficient and the speed-up ratio, graded early warning is generated through a fault tree model and transmitted to an external monitoring system, and accurate monitoring of early faults of the energy storage cabinet is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy storage cabinet safety monitoring, in particular to an energy storage cabinet multi-parameter intelligent monitoring device based on a MEMS sensor array. BACKGROUND

[0002] Energy storage cabinet safety monitoring is an important technology, which is specifically applied to the precise monitoring link of early failure of the energy storage cabinet battery module, and the core is to realize early identification of the operation risk of the energy storage cabinet by accurately capturing the correlation characteristics of temperature and vibration, so as to ensure the safe and stable operation of the energy storage system. In the existing MEMS monitoring scheme of the energy storage cabinet, the MEMS temperature sensor is easily disturbed by the non-steady-state interference of the cabinet internal heat dissipation airflow, the temperature detection value deviates from the real temperature of the module body due to the direct blowing or oblique sweeping of the airflow on the probe surface, and it is easy to misjudge as temperature anomaly, and only temperature or vibration is monitored, the coupling law of the unique high-frequency micro-vibration and temperature rising rate of the tab loosening is not associated, the absolute value of the early temperature of the tab loosening often does not break through the safety threshold, but after the high-frequency micro-vibration appears, the temperature rising rate will be accelerated synchronously, the existing scheme cannot capture this coupling characteristic, or the failure disposal is delayed due to the missed early coupling characteristic, and the precise and timely early warning of the tab loosening cannot be realized, in order to solve this technical problem, the present application provides an energy storage cabinet multi-parameter intelligent monitoring device based on a MEMS sensor array. SUMMARY

[0003] The present application aims to provide an energy storage cabinet multi-parameter intelligent monitoring device based on a MEMS sensor array to solve the problems in the background art.

[0004] To achieve the above-mentioned purpose, an energy storage cabinet multi-parameter intelligent monitoring device based on a MEMS sensor array is provided, which comprises: The sensor array unit 1 is composed of distributed MEMS temperature sensors and MEMS vibration sensors, wherein the MEMS temperature sensors adopt a non-vertical installation mode, so that the probe surface and the direction of the cabinet internal heat dissipation airflow form an included angle, thereby avoiding the temperature detection deviation caused by the direct blowing of the airflow; The airflow disturbance correction unit 2 is internally provided with a compensation algorithm based on the heat dissipation airflow dynamics characteristics, which can correct the temperature raw data collected by the sensor array unit 1 in real time, and eliminate the influence of airflow disturbance on the authenticity of the module body temperature; The multi-parameter coupling analysis unit 3 synchronously processes the vibration signal and the temperature change data stream, extracts the high-frequency micro-vibration characteristics unique to the tab loosening through real-time frequency domain analysis, and associates the synchronous change law of the temperature rising rate after the appearance of the characteristics, when the two satisfy the preset coupling condition, the early failure judgment logic is automatically triggered; The fault decision and output unit 4 fuses the corrected temperature data and the multi-parameter coupling analysis result, generates a hierarchical early warning signal, and transmits the signal to an external monitoring system through a communication interface.

[0005] Compared with the prior art, the present application has the following advantages: The MEMS temperature sensor is installed in a non-vertical direction to avoid direct blowing by air flow, a heat dissipation air flow vector field dynamic model and a compensation algorithm built in the air flow disturbance correction unit are combined to correct temperature raw data in real time, a multi-parameter coupling analysis unit extracts high-frequency micro-vibration characteristics unique to the tab loosening and correlates synchronous change rules of temperature rise rate, and two-stage fault determination logic is used, so that the non-steady-state interference of heat dissipation air flow on temperature detection is eliminated, and the vibration and temperature rise coupling characteristics of the early tab loosening are accurately captured, thereby effectively solving the problems of inaccurate temperature detection caused by air flow interference and missed early tab loosening fault caused by not correlating multi-parameter coupling characteristics in the existing MEMS monitoring scheme of the energy storage cabinet. BRIEF DESCRIPTION OF DRAWINGS

[0006] Figure 1 The present application is a whole block diagram.

[0007] The meanings of the various labels in the figure are as follows: 1, sensor array unit; 2, air flow disturbance correction unit; 3, multi-parameter coupling analysis unit; 4, fault decision and output unit. DETAILED DESCRIPTION

[0008] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0009] The present application provides a multi-parameter intelligent monitoring device for an energy storage cabinet based on a MEMS sensor array, as shown in Figure 1 The present application provides a multi-parameter intelligent monitoring device for an energy storage cabinet based on a MEMS sensor array, as shown in The sensor array unit 1 is composed of distributed MEMS temperature sensors and MEMS vibration sensors, wherein the MEMS temperature sensors are installed in a non-vertical direction, so that the probe surface forms an angle with the direction of the heat dissipation air flow in the cabinet, thereby avoiding temperature detection deviation caused by direct blowing by air flow; The air flow disturbance correction unit 2 is built in a compensation algorithm based on the heat dissipation air flow dynamics, which corrects the temperature raw data collected by the sensor array unit 1 in real time, and eliminates the influence of air flow interference on the temperature authenticity of the module body; The multi-parameter coupling analysis unit 3 synchronously processes the vibration signal and the temperature change data stream, extracts the high-frequency micro-vibration characteristics specific to the tab loosening through real-time frequency domain analysis, and correlates the synchronous change rule of the temperature rising rate after the characteristics appear, and when both satisfy the preset coupling condition, automatically triggers the early fault judgment logic; The fault decision and output unit 4 fuses the corrected temperature data and the multi-parameter coupling analysis result, generates a graded early warning signal, and transmits it to the external monitoring system through the communication interface.

[0010] The compensation algorithm built in the airflow disturbance correction unit 2 is realized by establishing a dynamic model of the heat dissipation airflow vector field. Based on the spatial distribution parameters of the sensor array unit 1 and the operation parameters of the axial flow fan, the airflow disturbance action matrix is constructed, and the airflow vector intensity and direction on the surface of each MEMS temperature sensor probe are calculated in real time. The algorithm generates a temperature compensation factor according to the vector field model and dynamically superimposes it on the original temperature data stream of the sensor in a sliding time window manner, thereby eliminating the non-steady-state interference of the heat dissipation airflow on the heat conduction process of the probe surface and ensuring that the corrected temperature value accurately reflects the temperature of the module body.

[0011] The real-time dynamic correction process is as follows: The airflow disturbance correction unit 2 first inputs the original temperature data collected by the sensor array unit 1 into the vector field dynamic model, calculates the heat exchange intensity deviation coefficient between the current airflow vector and the probe surface, generates a temperature correction gradient with phase compensation characteristics, and then integrates the gradient into the original data stream in the form of time series, thereby synchronously eliminating the temperature detection lag effect and amplitude attenuation effect caused by airflow transient disturbance, and finally outputting a continuous temperature signal consistent with the thermal inertia characteristics of the module body.

[0012] The synchronous processing process of the multi-parameter coupling analysis unit 3 adopts a dual-channel signal fusion architecture: The vibration signal channel extracts the original vibration waveform in the preset frequency band through adaptive band-pass filtering, generates a vibration energy spectrum through time-frequency transformation, and the temperature change channel performs first-order differentiation on the corrected continuous temperature signal to extract the temperature rising rate curve. The two-channel data streams are synchronized through the time stamp alignment module to provide a time sequence reference for subsequent feature correlation.

[0013] The extraction of high-frequency micro-vibration characteristics includes: The multi-parameter coupling analysis unit 3 identifies the energy surge frequency band in the vibration energy spectrum, establishes a resonance peak matching template based on the mechanical resonance characteristics of tab loosening failure, compares the envelope similarity of the vibration energy spectrum and the resonance peak template, selects the vibration events that satisfy the feature matching condition, and marks their starting time point and duration.

[0014] The high-frequency micro-vibration characteristics specific to tab loosening are as follows: The mechanical resonance phenomenon caused by the periodic change of the contact resistance at the battery tab junction, the vibration energy forms a shock-like waveform with a steep rising edge and a slow decay tail in a certain frequency band, and the time domain interval of adjacent vibration events has a strong correlation with the phase of the battery charging and discharging current.

[0015] The synchronous change rule of the temperature rise rate is as follows: After detecting the starting time point of the high-frequency micro-vibration feature, the multi-parameter coupling analysis unit 3 starts the trend tracking window of the temperature rise rate curve, and determines whether the vibration feature and the temperature rise rate exist spatio-temporal coupling by analyzing the monotonic increasing characteristics of the temperature rise rate in the window and the acceleration change. The coupling is manifested as the positive correlation between the vibration energy spectrum amplitude increment and the temperature rise rate increment.

[0016] The early fault determination logic includes two-stage triggering mechanism: The primary determination is based on the number of high-frequency micro-vibration features exceeding the set threshold, and the secondary determination verifies whether the temperature rise rate shows a continuous acceleration trend in the corresponding period. Only when both levels of determination are met, the early fault warning signal is triggered, and false triggering caused by mechanical noise or transient heat dissipation fluctuations is suppressed.

[0017] The hierarchical early warning signal generation process is as follows: The fault decision and output unit 4 takes the corrected temperature absolute value, the coupling strength coefficient output by the multi-parameter coupling analysis unit 3, and the temperature rise rate acceleration ratio as three-dimensional decision inputs, and maps the input parameters to the "normal", "early warning" and "emergency alarm" three states through the fault tree model. The early warning state corresponds to the high-frequency micro-vibration and temperature rise acceleration coupling event, and the emergency alarm state corresponds to the temperature absolute value breaking through the safety threshold.

[0018] The core decision rules of the fault tree model include: When the high-frequency micro-vibration and temperature rise acceleration coupling event is detected, the early warning signal is triggered regardless of whether the temperature absolute value breaks through the safety threshold. If the temperature rise rate acceleration ratio breaks through the critical value during the early warning period, it is immediately upgraded to emergency alarm. At the same time, the system continues to track the temperature rise rate returning to normal characteristics after the coupling event is removed, realizing the fault self-recovery discrimination.

[0019] It needs to be further explained that although the sensor array unit 1 avoids partial temperature detection deviation caused by direct blowing of heat dissipation airflow through non-vertical installation, the airflow distribution in the energy storage cabinet is non-steady-state due to the influence of the operating state of the axial flow fan, the shielding of the components in the cabinet, etc. Even if the probe is not directly opposite the airflow, the change of local airflow velocity will still interfere with the heat conduction process on the probe surface, resulting in that the temperature raw data cannot accurately reflect the temperature of the battery module body. Therefore, the airflow disturbance correction unit 2 needs to correct the temperature data in real time by relying on the heat dissipation airflow vector field dynamic model through the built-in compensation algorithm, and completely eliminate the non-steady-state interference of airflow. The specific implementation is as follows: The core carrier of the compensation algorithm built in the airflow disturbance correction unit 2 is the heat dissipation airflow vector field dynamic model. This model is not a fixed parameter model, but is dynamically constructed based on the spatial distribution parameters of the sensor array unit 1 and the operating parameters of the axial flow fan, to ensure that it can match the airflow changes in the cabinet in real time. Among them, the spatial distribution parameters of the sensor array include the three-dimensional coordinates, installation angle and probe surface area of all MEMS temperature sensors in the energy storage cabinet. These parameters determine the spatial position and flow area of each sensor probe in the airflow field, and are the basis for calculating the interference differences of airflow on different probes. The operating parameters of the axial flow fan include the real-time speed of the fan, the wind speed distribution of the outlet and the main direction of the airflow. These parameters directly determine the overall strength and flow trend of the airflow in the cabinet. When constructing the model, first, input the two types of parameters into the airflow dynamics calculation module. Through the mapping relationship between the parameters based on the basic principles of fluid mechanics, the model further constructs an airflow disturbance action matrix. The matrix has dimensions of sensor number, airflow parameters and interference degree. The row represents each MEMS temperature sensor, and the column includes the airflow vector intensity, i.e. the airflow volume acting on the probe surface per unit time, the airflow direction, i.e. the angle with the normal direction of the probe surface, and the corresponding heat conduction interference coefficient. The value of each element in the matrix is calculated by a spatial interpolation algorithm, to accurately describe the global airflow state in the cabinet.

[0020] With the support of the airflow disturbance action matrix, the heat dissipation airflow vector field dynamic model can calculate the airflow vector intensity and direction acting on the surface of each MEMS temperature sensor probe in real time. For example, when the speed of the axial flow fan increases from 1500 rpm to 2000 rpm, the model first calculates the wind speed increment of the outlet according to the fan operating parameters, and then calculates the airflow vector intensity at the probe by combining the sensor spatial distribution parameters through the airflow attenuation formula. At the same time, the actual acting angle of the airflow and the probe surface is determined through directional vector decomposition, to ensure that the airflow action state of each probe can be accurately quantified, avoiding the deviation caused by uniform parameter correction of different position sensors.

[0021] After the model completes the calculation of the airflow vector parameters of each probe, the compensation algorithm generates corresponding temperature compensation factors based on these parameters. The temperature compensation factor is a parameter that quantifies the degree of influence of airflow disturbance on temperature detection. Its numerical value is positively correlated with the strength of the airflow vector and is related to the direction of the airflow. When the airflow is obliquely across the probe, the heat exchange efficiency is lower than that of the direct blowing, and the compensation factor is relatively small. For example, the airflow vector intensity at a certain probe is 2.6 m / s, and the corresponding compensation factor is 0.8°C. Another probe far from the fan has an airflow intensity of 1.2 m / s, and the compensation factor is 0.3°C. This ensures that the compensation factor can accurately match the degree of airflow disturbance for different probes. To avoid overcompensation or undercompensation caused by instantaneous airflow fluctuations, the algorithm uses a sliding time window to dynamically superimpose the temperature compensation factor on the original temperature data stream of the sensor. The sliding time window refers to a fixed time interval, and the average airflow vector intensity within this interval is calculated in real time. The compensation factor corresponding to the average intensity is used as the correction value at the current time, rather than directly using the compensation factor corresponding to the instantaneous airflow. For example, if the airflow momentarily increases to 3 m / s due to fan speed fluctuations, directly using this instantaneous value to calculate the compensation factor may lead to excessive correction. However, the sliding time window generates the compensation factor by combining the airflow average values within the previous 5 seconds, which responds to the overall trend of airflow changes and filters out instantaneous disturbances. During the superimposition process, the algorithm synchronously associates the compensation factor with the original temperature data at the corresponding time. For example, if the original temperature value is 32.5°C and the compensation factor is 0.8°C, the corrected temperature value is 33.3°C. This process is executed in real time, and the compensation factor and the corrected value are updated every 0.1 second. Through the above complete process, the compensation algorithm can accurately eliminate the non-steady-state interference of the cooling airflow on the heat conduction process of the probe surface. Whether it is the airflow intensity fluctuation caused by the change of the axial flow fan speed or the local airflow direction deviation caused by the shielding of the cabinet components, the model can be adapted in real time, and the compensation factor can be adjusted dynamically to achieve correction. The final output temperature data is no longer affected by the non-steady-state airflow, but is consistent with the thermal inertia characteristics of the battery module itself. This ensures that the temperature data used by the subsequent multi-parameter coupling analysis unit 3 can truly reflect the heating state of the module, providing a reliable temperature reference for fault diagnosis.

[0022] The specific implementation of the real-time dynamic correction process of the airflow disturbance correction unit is as follows: The real-time dynamic correction of the air flow disturbance correction unit 2 starts with the temperature raw data input of the sensor array unit 1, which is collected by the distributed MEMS temperature sensor in real time. Each set of data is accompanied by a collection timestamp, sensor number and spatial coordinate information of the corresponding probe, ensuring that when the input vector field dynamic model is matched, the sensor in the model can quickly match the air flow parameters. The unit first classifies these raw data according to the sensor number, and imports them into the pre-set mapping channel of the heat dissipation air flow vector field dynamic model, avoiding confusion of data from different sensors, and laying a foundation for subsequent accurate deviation coefficient calculation. After the data input is completed, the model immediately starts the heat exchange intensity deviation coefficient calculation. The heat exchange intensity deviation coefficient is a parameter that quantifies the deviation degree of the current air flow vector from the actual state of heat exchange on the probe surface compared to the ideal state without air flow disturbance. In the ideal state without air flow, the probe surface only exchanges heat with the module body through heat conduction, and the temperature detection value can accurately reflect the body temperature. When the air flow acts on the probe, the air flow will carry away part of the heat on the probe surface, resulting in actual heat exchange intensity higher than the ideal state, and the detection value lower than the body temperature. The deviation coefficient is used to measure the difference between the actual heat exchange and the ideal heat exchange. During the calculation process, the model will call the current air flow vector intensity, air flow direction and probe heat conduction coefficient of the sensor, and calculate the deviation coefficient corresponding to each raw data collection time through the heat exchange formula simplification, ensuring that the deviation coefficient can match the instantaneous change of the air flow in real time. Based on the heat exchange intensity deviation coefficient obtained by calculation, the air flow disturbance correction unit 2 further generates a temperature correction gradient with phase compensation characteristics. The temperature correction gradient refers to the magnitude of the correction of the original temperature value per unit time, and its value is positively related to the deviation coefficient. The larger the deviation coefficient, the more significant the detection deviation caused by the air flow, and the larger the correction gradient needs to be. The phase compensation characteristic is to solve the time lag problem between air flow instantaneous disturbance and temperature detection. Since air flow changes will first act on the probe surface and then affect the temperature detection value through heat conduction, the change of the detection value lags behind the air flow disturbance. Phase compensation can adjust the time phase of the correction gradient to align the correction action with the actual influence time of the air flow disturbance, avoiding the lagging correction situation where the air flow has weakened but the correction is still increasing. For example, when the air flow suddenly increases at a certain time, the phase compensation will adjust the correction gradient to the corresponding magnitude in advance, ensuring that the detection value is pre-corrected before being affected by the air flow, thereby eliminating the deviation caused by the time difference.After generating the temperature correction gradient, the air flow disturbance correction unit 2 applies the gradient to the temperature original data stream in the form of time series integration, which means that the temperature correction gradient in a continuous time period is accumulated and superimposed, rather than isolatedly correcting the original data at a single time point. Since the thermal influence of air flow disturbance on the probe is continuous, if only the data at a certain time point is corrected, the corrected temperature signal will have a fault-like fluctuation, which does not conform to the law of continuous change of the module body temperature. However, through time series integration, the correction gradient at each time point is gradually accumulated in chronological order, making the correction process continuous and smooth. For example, in a 5-second time series, the first-second correction gradient is 0.2℃ / s, the second-second is 0.3℃ / s, and the fifth-second is 0.25℃ / s. The total correction value in this time period is the cumulative sum of the gradient of each second, which is superimposed on the original temperature data of the corresponding time period, ensuring that the trend of the corrected temperature is coherent. This integral correction process can simultaneously eliminate the temperature detection lag effect and amplitude decay effect caused by air flow transient disturbance. For the lag effect, the phase compensation characteristic has aligned the correction gradient with the air flow disturbance time, so the integral correction can respond to the air flow change in real time, avoiding the detection value lagging behind the body temperature. For the amplitude decay effect, the accumulation of the correction gradient can compensate for the low detection value caused by the heat carried away by the air flow. Finally, the output of the air flow disturbance correction unit 2 is no longer a discrete temperature point disturbed by air flow, but a continuous temperature signal consistent with the thermal inertia characteristic of the battery module body. The thermal inertia characteristic means that the module body has a large thermal capacity and the temperature changes relatively smoothly. The corrected continuous temperature signal also shows this smooth change trend, without sudden rise and fall caused by air flow transient fluctuation, accurately reflecting the real temperature state of the module body, and providing a reliable temperature data basis for the subsequent multi-parameter coupling analysis unit 3 to associate the vibration signal with the temperature change rule.

[0023] After the air flow disturbance correction unit 2 outputs the continuous temperature signal consistent with the thermal inertia characteristic of the battery module body, the multi-parameter coupling analysis unit 3 needs to process the vibration signal collected by the MEMS vibration sensor and the temperature signal synchronously. Since the physical dimensions and change rules of the two types of signals are significantly different, unified channel processing may lead to feature loss or time sequence misalignment. Therefore, the unit adopts a dual-channel signal fusion architecture to achieve accurate synchronization through the logic of separate channel feature extraction and time sequence alignment, providing a reliable data basis for the subsequent feature association of tab loosening failure. The specific implementation is as follows: The dual-channel signal fusion architecture of the multi-parameter coupling analysis unit 3 first starts the vibration signal channel, and the core task of the channel is to extract the high-frequency micro-vibration characteristics unique to the lug loosening from the complex background noise. The first step of the channel processes the original vibration signal through adaptive band-pass filtering. The adaptive band-pass filtering is different from the fixed frequency band filtering method. It can dynamically adjust the filtering frequency band range according to the real-time vibration environment in the energy storage cabinet. When the energy storage cabinet is running, various background noises will be generated, and the high-frequency micro-vibration of the lug loosening is concentrated in a specific preset frequency band. The adaptive band-pass filtering will first identify the main frequency band of the current background noise, and then automatically shrink the filter passband to the preset high-frequency interval, while suppressing the background noise outside the passband, avoiding the noise from masking the weak vibration signal of the lug loosening. For example, when the fan speed is increased, causing the low-frequency noise to increase, the lower limit of the filter passband will be fine-tuned from 1000Hz to 1050Hz, further isolating the low-frequency interference and ensuring that the extracted original vibration waveform only contains the effective signal of the preset high-frequency band.

[0024] After completing the adaptive band-pass filtering, the vibration signal channel performs time-frequency transformation on the extracted high-frequency original vibration waveform. The time-frequency transformation is to convert the vibration waveform in the time domain into a three-dimensional distribution of vibration energy spectrum of time, frequency and energy. The purpose is to intuitively present the vibration energy intensity at different times and different frequencies. The high-frequency micro-vibration caused by the loose tab appears as a short pulse signal in the time domain, which is difficult to identify directly. After time-frequency transformation, it will form a clear feature of sudden increase of energy in a specific frequency band in the energy spectrum, and can accurately mark the starting time, duration and peak intensity of the sudden increase of energy, providing a clear vibration feature anchor point for the subsequent correlation of temperature change. The temperature change channel processed in parallel with the vibration signal channel focuses on extracting the temperature rise rate feature from the continuous temperature signal that can reflect the abnormal heating of the module. The channel first receives the continuous temperature signal output by the airflow disturbance correction unit 2. This kind of signal has eliminated the airflow interference and can truly reflect the smooth change trend of the module body temperature. The channel performs first-order differential processing on it. The first-order differential processing here is used to calculate the temperature change per unit time, i.e. the temperature rise rate. Compared with the absolute value of temperature, the temperature rise rate can better capture the heating abnormality caused by the loose tab in the early stage. The loose tab will increase the contact resistance, and the heat generated by the current will increase. However, this heat generation only shows a slight acceleration of the temperature rise rate in the early stage, and the absolute value of temperature may still be within the normal range. The temperature rise rate curve generated by the first-order differential processing can amplify this subtle rate change, making it easier to identify the coupling rule of rate acceleration after vibration in the subsequent correlation of vibration features. After the vibration energy spectrum and the temperature rise rate curve are output by the two channels respectively, the multi-parameter coupling analysis unit 3 synchronizes the data streams of the two channels through the time stamp alignment module. The core logic of the time stamp alignment module is to eliminate the micro-time sequence deviation in the data collection and transmission process based on the time stamp reference of the two types of signal collection. Although the MEMS temperature sensor and the MEMS vibration sensor are started simultaneously for collection, due to the slight difference in signal transmission path length and sensor response speed, the vibration data and temperature data at the same time may have a time difference of milliseconds. If directly correlated, it will cause the time sequence of the vibration features and the temperature change to be misaligned, misjudging the coupling relationship between the two. When the module is working, it will first read the collection time stamp of each set of vibration energy spectrum data and temperature rise rate data, and then correct the temperature data time stamp to be consistent with the vibration data through linear interpolation algorithm, ensuring that each vibration feature node can accurately match the temperature rise rate at the same time, forming a synchronized data set with one-to-one correspondence of vibration and temperature time sequence. Through the above double-channel fusion process, the multi-parameter coupling analysis unit 3 not only extracts the core features of the two types of signals, but also eliminates the data deviation through time sequence alignment. The finally output synchronized data set can accurately reflect the correlation of vibration and temperature at the same time, providing a strict time sequence reference for the subsequent identification of the temperature rate acceleration coupling rule after the appearance of the unique vibration feature of the loose tab, and triggering the early fault judgment logic.

[0025] The specific implementation of the multi-parameter coupling analysis unit high-frequency micro-vibration feature extraction is as follows: The multi-parameter coupling analysis unit 3 first identifies the energy burst frequency band in the vibration energy spectrum. The energy burst frequency band refers to the region in the vibration energy spectrum where the energy value of a certain high-frequency band significantly exceeds the surrounding background energy in a short time. Its identification needs to be combined with the high-frequency characteristics of the tab loosening and the dynamic threshold determination. The unit first calculates the average background energy of the preset high-frequency interval based on historical normal operation data, that is, the 1000-3000 Hz frequency band where tab loosening faults often occur. This interval is determined by the mechanical properties of the tab material, length, and connection structure. The 1.5 times of this average value is set as the dynamic burst threshold. The threshold will be fine-tuned with the running time to avoid misjudgment. Then, the 1000-3000 Hz frequency band of the vibration energy spectrum is scanned segment by segment. If the energy value of a certain frequency band exceeds the dynamic burst threshold for 3 consecutive sampling points, and the frequency band width is within the range of 50-100 Hz, then this frequency band is marked as an energy burst frequency band to be verified, excluding interference caused by electromagnetic noise.

[0026] After identifying the energy surge frequency band to be verified, the unit establishes a resonance peak matching template based on the mechanical resonance characteristics of the tab looseness fault. The mechanical resonance characteristics of the tab looseness fault are derived from the nature of the fault. When the tab and the battery module connecting piece are loose, the intermittent contact failure caused by the periodic change of the contact resistance of the two with the periodic change of the charging and discharging current will cause a small mechanical collision and vibration at the contact point. This vibration is affected by the elastic modulus of the tab copper alloy material and the constraint strength of the connecting structure, and will form mechanical resonance at a specific frequency. The vibration waveform has an impact-like characteristic. Based on these characteristics, the resonance peak matching template is constructed as a three-dimensional structure, in which the frequency axis locks the resonance interval of 1800-2200 Hz, the time axis marks the time window of 0.05 seconds rising and 0.2 seconds decaying, and the energy axis sets the template envelope according to the peak energy proportion, while embedding the non-periodic interval characteristics of the tab looseness vibration. The time interval of adjacent vibration events is related to the phase of the charging and discharging current. The template does not preset a fixed interval, but only focuses on the envelope shape of a single vibration, ensuring that the template can accurately match the vibration rules of the real fault. Subsequently, the unit compares the actual energy envelope of the verified frequency band with the envelope similarity of the resonance peak matching template to screen vibration events that meet the feature matching conditions. The envelope similarity refers to the degree of coincidence between the actual energy envelope and the template envelope in shape, time parameter, and energy distribution. The unit first extracts the actual energy envelope of the verified energy surge frequency band, i.e., the curve formed by connecting the energy values of each time node in the frequency band, and then compares it with the envelope curve of the resonance peak template at each time node. If the rising time of the actual envelope deviates from the template rising time by less than or equal to 0.01 seconds, the decay time deviates from the template decay time by less than or equal to 0.03 seconds, and the difference between the energy values of each time node of the actual envelope and the template energy values is less than or equal to 20%, it is determined that the verified frequency band meets the feature matching conditions, and the corresponding signal is the high-frequency micro-vibration generated by the tab looseness. If the deviation exceeds the above range, it is determined as an interference signal and is excluded. Finally, the unit labels the time information of the vibration events that meet the feature matching conditions, aligns the global time reference through the time stamp alignment module, determines the starting time point of the vibration event, i.e., the time when the energy surge frequency band first exceeds the dynamic surge threshold, and calculates the duration from the starting time to the time when the energy falls below the dynamic surge threshold. The starting time, duration, and actual energy envelope are stored together with the vibration energy spectrum data of the vibration event.

[0027] The specific embodiment of the unique high-frequency micro-vibration characteristics of the tab loosening is screened out by the resonance peak matching template in the multi-parameter coupling analysis unit 3, and after the vibration event of the suspected tab loosening is screened out, in order to further confirm that the event is not the interference signal of the fan high-frequency harmonic and electromagnetic radiation, the unique performance of the high-frequency micro-vibration of the tab loosening needs to be analyzed from the essence of the fault, which is the physical and electrical coupling effect after the battery tab loosening, and has clear mechanism support and quantifiable form law. The specific embodiment is as follows: The core cause of the unique high-frequency micro-vibration of the tab loosening is the periodic change of the contact resistance at the battery tab connection, which directly leads to mechanical resonance. When the battery tab and the module connecting piece in the energy storage cabinet are loose, the contact surface cannot maintain close contact, and a small physical gap is formed. During the charging and discharging process of the battery, the current is not a constant value, but a periodic fluctuation with the charging and discharging cycle. The periodic change of the current leads to the synchronous fluctuation of the contact resistance. When the current is at the peak value, the Joule heat effect of the contact point is enhanced, the metal contact surface slightly expands due to heating, the gap temporarily shrinks, and the contact resistance decreases. When the current is at the valley value, the Joule heat decreases, the contact surface cools and shrinks, the gap recovers or even expands, and the contact resistance rises again. This cycle causes intermittent mechanical collision between the tab and the connecting piece. The tab itself is made of copper alloy material, and its inherent mechanical properties determine that the vibration caused by the collision will form mechanical resonance at a specific high-frequency band, rather than irregular chaotic vibration. This is the primary characteristic of tab loosening vibration that distinguishes it from other interference signals.

[0028] From the vibration waveform form, the energy of the high-frequency micro-vibration in a certain frequency band will form a shock-like waveform with steep rising edge and slow decay tail. The steep rising edge is due to the instantaneous nature of mechanical collision. The collision process of the lug and the connecting piece only lasts for a very short time. The kinetic energy at the moment of collision is quickly converted into vibration energy, making the vibration amplitude rapidly rise to a peak value in a short time. There is no gradual process of energy accumulation. In the vibration energy spectrum, the slope from the background energy level to the peak energy is very large, forming a steep shape. The slow decay tail is related to the elastic reset characteristics of the lug. After the collision ends, the lug does not immediately stop vibrating, but performs a small amplitude reciprocating reset movement by relying on the elasticity of the copper alloy. The vibration energy is slowly released with the gradual recovery of the elastic deformation, rather than instantaneously dissipated. Therefore, the duration of the decay section is much longer than that of the rising section, and the energy falls exponentially to the background level. Finally, a shock-like waveform with steep front and slow back is formed. This waveform form is significantly different from the sharp peak and tail of electromagnetic interference and the gentle fluctuation of high-frequency harmonics of axial flow fans, which can be used as a key form basis for feature recognition. In addition, another unique performance of the loose lug high-frequency micro-vibration is the strong correlation between the time domain interval of adjacent vibration events and the phase of the battery charging and discharging current. The fluctuation of the contact resistance is determined by the phase of the charging and discharging current, and the generation of the vibration event depends on the mechanical collision caused by the fluctuation of the contact resistance. Therefore, the change of the current phase will directly affect the frequency of the vibration event. When the current is in the phase peak interval, the fluctuation of the contact resistance is the most severe, the collision frequency of the lug and the connecting piece is the highest, and the time domain interval of the adjacent two vibration events is the shortest, usually 0.5-1 seconds. When the current is in the phase valley interval, the contact resistance fluctuation is gentle, the collision frequency is reduced, and the time domain interval of the adjacent vibration events is extended to 2-3 seconds. When the current is in the stable transition stage, the interval is maintained at an intermediate level of 1-1.5 seconds. This regular change of the interval with the current phase is completely different from random noise and cabinet structure resonance. The matching degree of the interval and the current phase can be monitored to further verify whether the vibration event is caused by the lug loosening. Through the above three unique performances, the high-frequency micro-vibration of the lug loosening can be accurately defined, which provides clear feature basis for the subsequent multi-parameter coupling analysis unit 3 to associate the temperature rise rate and trigger the early fault judgment logic, avoiding the false triggering of the early warning caused by mistakenly identifying the interference signal as the fault feature.

[0029] After the multi-parameter coupling analysis unit 3 accurately extracts the high-frequency micro-vibration characteristics unique to the lug loosening and marks the starting time point of the vibration event through the resonance peak template matching, only relying on the vibration characteristics cannot completely determine the early fault. Occasionally, there may be slight vibrations in the energy storage cabinet without actual hazards. Such vibrations will not cause module temperature rise anomalies. If misjudgment occurs, unnecessary maintenance costs will be increased. Therefore, the unit needs to further associate the synchronous change rule of the temperature rise rate to verify the time and space coupling relationship and confirm whether the vibration is caused by the lug loosening, which will cause temperature rise risks. The specific implementation is as follows:The multi-parameter coupling analysis unit 3 starts the trend tracking window of the temperature rise rate curve immediately after detecting the starting time point of the high-frequency micro-vibration feature, i.e., the time when the marked vibration event first exceeds the dynamic surge threshold. The trend tracking window is a fixed-length analysis interval set for the vibration starting time point, and its length is not arbitrarily set but determined based on a large amount of fault simulation test data. In a laboratory environment, after artificially creating a tab looseness fault, it takes 3-8 seconds of lag time from the appearance of the vibration feature to the appearance of the significant change in the temperature rise rate. Therefore, the window length is set to 10 seconds, and the window starting time is completely aligned with the vibration starting time, ensuring that the change process of the temperature rise rate after vibration occurs can be completely captured, while avoiding missing key changes due to too short window or including irrelevant temperature fluctuation data due to too long window. After the trend tracking window is started, the unit first analyzes the monotonic increasing property of the temperature rise rate curve in the window. The monotonic increasing property refers to the fact that, within the entire tracking window, the temperature rise rate, generated by the temperature change channel through first-order differentiation processing, gradually increases from the initial value at the window starting time to the final value at the window ending time, without any significant decrease node. The physical logic of this property is derived from the development law of tab looseness fault. The increase in contact resistance caused by tab looseness is not instantaneous and stable, but gradually intensified with vibration (continuous collision of contact points). The continuous rise in contact resistance leads to a synchronous increase in heat generation power, which in turn causes the temperature rise rate to show a "gradually accelerated" monotonic increasing trend. If the rate decreases within the window, it means that vibration does not cause continuous increase in contact resistance, and it is likely to be a non-hazardous transient vibration. The unit compares the rate values at each sampling point. If the rate values of the last 5 consecutive sampling points do not decrease, it is determined that the monotonic increasing property is satisfied. While verifying the monotonic increasing property, the unit simultaneously analyzes the acceleration change of the temperature rise rate. The acceleration of the temperature rise rate refers to the change in rate per unit time, reflecting the degree of rate acceleration. The more serious the tab looseness, the faster the increase in contact resistance, the more rapid the increase in heat generation, and the larger the value of the rate acceleration. If the acceleration is 0 or negative, it means that the rate growth is stagnant or slowing down, which does not conform to the fault development law. During the analysis process, the unit calculates the average value of the rate acceleration every 1 second in the window. If the average value is always positive and stable at 0.02℃ / min² or above, which is calibrated by normal operation data, then it is further confirmed that the temperature rising rate is in an accelerating state, which is consistent with the failure mechanism of tab loosening. Through the double analysis of the monotonic increasing characteristics and the acceleration change, the unit can determine whether the high-frequency micro-vibration characteristics and the temperature rising rate exist in spatiotemporal coupling. The spatiotemporal coupling refers to that the vibration characteristics and the temperature rising rate change are synchronized in time and logically related. The rate change conforms to the increasing and accelerating rules caused by the failure, rather than the accidental superposition without any relevance. This coupling is finally reflected by the positive correlation characteristics between the vibration energy spectrum amplitude increment and the temperature rising rate increment. The vibration energy spectrum amplitude increment refers to the maximum energy amplitude of the vibration event within the window, and the increase compared to the normal vibration amplitude outside the window. Its size directly reflects the severity of tab loosening. The temperature rising rate increment is the difference between the rate value at the end of the window and the rate value at the beginning of the window. The positive correlation characteristics are that the greater the vibration energy spectrum amplitude increment, the greater the temperature rising rate increment. This is because more serious loosening will produce stronger vibration energy, and at the same time, it will also cause the contact resistance to increase significantly, thereby triggering more obvious temperature rising rate acceleration. The unit compares the change trends of the two, and if they both show a synchronous increasing rule in 3 groups of continuous vibration and temperature rising data, then it is finally confirmed that there is spatiotemporal coupling, which provides a key basis for the subsequent early failure determination logic.

[0030] The specific implementation of the two-level triggering mechanism of the early failure determination logic of the multi-parameter coupling analysis unit is as follows: The first level of the early fault determination logic is the primary determination, which is based on whether the number of continuous occurrences of high-frequency micro-vibration features exceeds a set threshold. The high-frequency micro-vibration features refer to vibration events that are filtered by a resonance peak matching template and conform to the unique shape of tab loosening, rather than all high-frequency signals. The number of continuous occurrences is counted within a preset statistical time window, and the length of the time window is set according to the development law of tab loosening failure. Laboratory fault simulation shows that tab loosening from initial slight loosening to forming stable failure features requires a development period of about 60 seconds, during which 3 or more high-frequency micro-vibrations are continuously generated. Therefore, the statistical time window is set to 60 seconds to ensure complete coverage of the continuous occurrence process of failure features. The threshold is determined by comparing normal operation data and fault data. When the energy storage cabinet is in normal operation without failure, the number of vibration events that meet the tab loosening feature within 60 seconds is not more than 2, while in the fault state, the number is greater than or equal to 3. Therefore, the threshold is set to 3. When the primary determination is executed, the unit will count the number of high-frequency micro-vibration events that meet the conditions within each 60-second window in real time. If the number exceeds 3 in two consecutive statistical windows, or the number exceeds 5 in a single window, it represents rapid failure development, and the primary determination passes. If the number does not reach the threshold, or it only meets the single window requirement and then falls back, it is determined as not passing the primary determination, and the subsequent determination is temporarily excluded to exclude single and accidental vibration interference. After the primary determination passes, the logic automatically enters the secondary determination, which focuses on verifying whether the temperature rise rate within the primary determination statistical period shows a continuous acceleration trend. The 60-second window triggered by the primary determination is strictly aligned with the statistical window of the primary determination to avoid false positives due to time window misalignment. The verification of continuous acceleration trend needs to be combined with the analysis logic of temperature rise rate and developed from two dimensions. First, whether the acceleration of temperature rise rate is continuously positive. The unit extracts the average acceleration of rate every 10 seconds within the corresponding period. The acceleration calculation method is consistent with the trend tracking window, which is the change of rate per unit time. If all average values are positive, allowing a small fluctuation of ±0.01℃ / min² to exclude sensor errors, it means that the rate is always accelerating. Second, whether the rate increment reaches the failure level standard. The difference between the temperature rise rates at the beginning and end of the calculation period is calculated. If the difference is greater than or equal to 0.3℃ / min (this value is derived from fault data: the rate increment is usually ≤0.1 ℃ / min, and the increment will increase significantly due to the continuous increase in contact resistance), the acceleration trend is further confirmed to comply with the failure law. Only when both dimensions are verified, i.e., the acceleration is continuously positive and the rate increment meets the standard, is the secondary determination considered to pass. If the acceleration is negative, the rate increment does not meet the standard, or only a single dimension meets the standard, the secondary determination is determined to fail, even if the primary determination has passed, no early warning is triggered. The early failure determination logic only triggers the early failure warning signal when the primary determination and the secondary determination pass at the same time. The warning signal contains specific failure-related information, such as the number of vibration events that meet the primary determination, the rate increment of the secondary determination, the corresponding statistical time window, and the associated sensor number, which facilitates the positioning of the failure module by the operation and maintenance personnel. The signal is transmitted in real time to the fault decision and output unit (4) to provide a basis for subsequent hierarchical warning. At the same time, the two-level mechanism can effectively suppress various false triggers. When mechanical noise occurs, although more than 3 high-frequency signals may be generated within a single window, these signals are mostly not in the form of loose tabs and are difficult to pass the primary determination. Even if there are signals that meet the characteristics, the secondary determination will not pass due to the lack of temperature rise acceleration trend. When there is a transient heat dissipation fluctuation, the temperature rise rate may temporarily increase, but the primary determination will not pass directly due to the lack of corresponding high-frequency micro-vibration, and the warning will not be triggered. Through the dual verification of the continuous occurrence of vibration characteristics and the synchronous acceleration of temperature rise trend, the early failure determination logic can accurately capture the early failure of loose tabs and avoid false judgments caused by interference signals to the greatest extent, ensuring the accuracy and reliability of the warning.

[0031] After the multi-parameter coupling analysis unit 3 completes the early failure determination through the two-level triggering mechanism and clearly defines the coupling relationship between high-frequency micro-vibration and temperature rise acceleration, the fault decision and output unit 4 needs to further integrate the core data of the previous units to convert the scattered temperature and vibration coupling characteristics into hierarchical warning signals with clear priorities. If only a single warning type is output, it is easy to confuse the early failure of loose tabs with the temperature overrun emergency, leading to improper operation and maintenance response. Therefore, the unit realizes accurate hierarchical warning through three-dimensional decision input and fault tree model mapping logic. The specific implementation is as follows: The fault decision and output unit 4 first determines the specific source and physical meaning of the three-dimensional decision input, ensures that the input data can fully reflect the operation state of the energy storage cabinet and has no interference, the first dimension input is the corrected temperature absolute value, which is not the original collected data of the MEMS temperature sensor, but the result corrected by the heat dissipation airflow vector field dynamic model of the airflow disturbance correction unit 4, which has eliminated the non-steady-state interference of the heat dissipation airflow and can accurately reflect the real temperature of the battery module body, which is the direct basis for judging whether the module is facing the risk of overheating. The value will be synchronized in real time from the continuous temperature signal output end of the airflow disturbance correction unit 4, the second dimension input is the coupling strength coefficient output by the multi-parameter coupling analysis unit 4, which is the core index of quantifying the coupling degree of high-frequency micro-vibration characteristics and temperature rise rate acceleration trend, the stronger the coupling, the larger the coefficient value, the value range is calibrated by fault simulation as 0-1, 0 represents no coupling, and 1 represents complete coupling, the third dimension input is the temperature rise rate acceleration ratio, which refers to the ratio of the acceleration of the current temperature rise rate to the average acceleration of the energy storage cabinet in normal operation.

[0032] After obtaining the three-dimensional decision input, the fault decision and output unit 4 maps the input parameters to the "normal", "early warning" and "emergency alarm" three-level states through the preset fault tree model. Based on the energy storage cabinet safety operation standard and fault data, the triggering conditions of each state are determined. The "normal" state corresponds to the three-dimensional input being within the safe range, the modified temperature absolute value being lower than the safe threshold, the coupling strength coefficient being less than 0.3, representing no significant vibration and temperature rise coupling, and the temperature rise acceleration ratio being less than 1.5, representing that the temperature rise acceleration does not exceed the normal fluctuation range. The three conditions need to be met at the same time. At this time, the unit determines that the energy storage cabinet is running normally, does not trigger any early warning signal, and only records the regular operation data. The "early warning" state corresponds to the high-frequency micro-vibration and temperature rise acceleration coupling event. The triggering conditions are that the coupling strength coefficient exceeds 0.3 (representing significant coupling), the temperature rise acceleration ratio exceeds 1.5 (representing obvious temperature rise acceleration), and the modified temperature absolute value is still lower than the safe threshold. This state is aimed at the early failure of the tab loosening. At this time, although the module temperature does not exceed the limit, the vibration and temperature rise have formed a fault-level coupling. If no intervention is made, it may develop into temperature overrun. Therefore, the early warning signal will mark the early warning and tab loosening risk, and will be accompanied by the specific values of the coupling strength coefficient and the acceleration ratio, reminding the operation and maintenance personnel to check the tab connection of the corresponding module within 24 hours. The "emergency alarm" state corresponds to the temperature absolute value breaking through the safe threshold. The triggering condition is that the modified temperature absolute value exceeds the safe threshold. Whether the coupling strength coefficient and the temperature rise acceleration ratio are in the normal range or not, it is directly determined as an emergency alarm. Because the temperature overrun has directly threatened the safety of the energy storage cabinet, it needs to be disposed first. The early warning signal will mark the emergency alarm and temperature overrun, and highlight the difference between the current temperature value and the safe threshold. At the same time, it automatically associates with the emergency measure suggestion. After the fault tree model completes the state determination, the fault decision and output unit 4 automatically generates a graded early warning signal. In addition to the three-level state identification, the signal also contains key traceability information, ensuring that the operation and maintenance personnel can quickly locate the fault point and the fault reason. The unit transmits the early warning signal to the external monitoring system in real time through the preset communication interface, and stores the early warning log locally, which is convenient for subsequent fault tracing and system optimization. Through the comprehensiveness of the three-dimensional input and the logical rigor of the fault tree model, the early failure is avoided, and the delay of the emergency situation is prevented, realizing the precision and practicality of the graded early warning, and providing reliable signal support for the safe operation of the energy storage cabinet.

[0033] The specific implementation of the core decision rule of the fault decision and output unit fault tree model is as follows: after the fault decision and output unit 4 completes the preliminary mapping of the three-dimensional decision input to the three-level state through the fault tree model, in order to further solve the actual needs of "coupling event priority early warning", "early warning state dynamic adjustment" and "fault recovery automatic discrimination", avoid the early warning lag or misjudgment caused by fixed threshold determination, the fault tree model is built-in a set of logically rigorous core decision rules, which deeply combines the feature data output by each unit in the early stage with the operation and maintenance requirements, and the specific implementation is as follows: The primary logic of the core decision rule of the fault tree model is the priority early warning mechanism for high-frequency micro-vibration and temperature rise acceleration coupling events. The high-frequency micro-vibration and temperature rise acceleration coupling events refer to the events verified by the trend tracking window and meeting the vibration and temperature rise space coupling by the multi-parameter coupling analysis unit 3. The judgment basis is that the coupling strength coefficient exceeds the preset threshold, and the temperature rise rate acceleration ratio exceeds 1.5, which represents that the current temperature rise acceleration amplitude is more than 1.5 times of that in normal operation. The rule clearly stipulates that as long as such a coupling event is detected, an early warning signal is directly triggered regardless of whether the modified temperature absolute value breaks through the safety threshold. The core logic of this design is that the coupling event itself indicates the early failure of the lug loosening which will gradually deteriorate. Even if the current temperature does not exceed the limit, if early warning is not timely, the temperature will inevitably rise rapidly with the continuous increase of the contact resistance. After the early warning signal is triggered, the core decision rule enters the dynamic upgrading logic, which focuses on monitoring whether the temperature rise rate acceleration ratio breaks through the critical value during the early warning period. The early warning period refers to the entire period from the early warning trigger time to the warning release or upgrade. The model reads the latest data of the temperature rise rate acceleration ratio at a high frequency of once per second. The critical value is set based on the failure deterioration rate analysis. In the laboratory simulation of the lug loosening failure deterioration process, when the acceleration ratio exceeds 3, the temperature will break through the safety threshold within 10 minutes. If early warning is still maintained at this time, the best disposal time will be missed. Therefore, the critical value is set to 3. The rule stipulates that if the temperature rise rate acceleration ratio of any three consecutive sampling points breaks through 3 during the early warning period, the model will immediately upgrade the warning state from "early warning" to "emergency alarm", and update the warning information to emergency alarm and lug loosening deterioration, and automatically supplement the difference between the current temperature absolute value and the safety threshold. Through the communication interface, the sound and light alarm instructions are sent to the external monitoring system to forcibly remind the operation and maintenance personnel to take emergency measures immediately to avoid further temperature rise and cause thermal runaway risk. In addition, the core decision rule also includes fault self-recovery judgment logic, which realizes the automatic release of the warning state by continuously tracking the regression normal characteristics of the temperature rise rate after the coupling event is released. The judgment standard for the release of the coupling event is that the coupling strength coefficient output by the multi-parameter coupling analysis unit 3 is less than 0.3 for 5 minutes, and the temperature rise rate acceleration ratio is less than 1 for 5 minutes.5, both need to meet at the same time, represent the tab loose fault may have been repaired by operation and maintenance, or due to the decrease of charge and discharge current, the contact resistance fluctuation slows down, and the temperature rising rate returns to the normal range of the energy storage cabinet during normal operation, which is obtained by statistical analysis of historical normal data, and lasts more than 10 minutes, the model will start the regression tracking window after the coupling event is removed, and record the temperature rising rate every 2 minutes, if all the recorded values are within the normal range within 10 minutes, and there is no sign of coupling event again, it is determined that the fault is self-recovered, the early warning state is automatically removed, the system state is updated to normal, and is marked in the early warning log, if the temperature rising rate is still higher than the normal range after the coupling event is removed, it is determined that the coupling is removed but the temperature rising is not restored, the early warning is still retained, until the rate returns or the operation and maintenance confirms that the fault is eliminated, through the above decision rules, the fault tree model not only realizes the accurate capture of early faults, but also solves the dynamic adjustment of early warning state and the automatic management of fault recovery, avoids the lag problem of not warning when the temperature is not exceeded, and the inefficient problem of manual removal after early warning, and at the same time ensures that each early warning adjustment is supported by the data output of the previous unit, and logically connects with the function depth of each unit, and finally provides an early warning management scheme for the operation and maintenance of the energy storage cabinet.

[0034] In the application, the sensor array unit 1 contains distributed MEMS temperature and vibration sensors, the temperature sensor is not vertically installed, avoiding detection deviation caused by direct blowing of airflow, the airflow disturbance correction unit 2 uses a dynamic model of heat dissipation airflow vector field and a compensation algorithm to correct the temperature raw data in real time, eliminate the interference of airflow on the temperature detection of the module body, the multi-parameter coupling analysis unit 3 processes vibration and temperature signals in double channels, extracts the high-frequency micro-vibration characteristics of tab looseness, and correlates the temperature rising rate, triggers the early fault logic through two-stage judgment, the fault decision and output unit 4 fuses the corrected temperature, coupling strength coefficient and rate acceleration ratio, generates a hierarchical early warning through a fault tree model, and transmits to an external monitoring system, realizing accurate monitoring of early faults of the energy storage cabinet.

[0035] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above examples, the above examples and descriptions in the specification are only preferred examples of the present application, and are not intended to limit the present application, various changes and improvements can be made without departing from the spirit and scope of the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A multi-parameter intelligent monitoring device for energy storage cabinets based on MEMS sensor array, characterized in that, The application relates to a battery module temperature monitoring system, which comprises the following units: a sensor array unit (1) is composed of distributed MEMS temperature sensors and MEMS vibration sensors, wherein the MEMS temperature sensors adopt a non-vertical installation mode, so that the surface of the probe forms an angle with the direction of the heat dissipation airflow in the cabinet, thereby avoiding the temperature detection deviation caused by the direct blowing of the airflow; an airflow disturbance correction unit (2) is internally provided with a compensation algorithm based on the heat dissipation airflow dynamic characteristics, the original temperature data collected by the sensor array unit (1) are corrected in real time, and the influence of the airflow disturbance on the authenticity of the temperature of the module body is eliminated; a multi-parameter coupling analysis unit (3) synchronously processes vibration signals and temperature change data streams, extracts the high-frequency micro-vibration characteristics of the tab loosening through real-time frequency domain analysis, and correlates the synchronous change rule of the temperature rising rate after the characteristics appear; when the two meet the preset coupling conditions, early fault judgment logic is automatically triggered; and a fault decision and output unit (4) fuses the corrected temperature data and the multi-parameter coupling analysis results, generates a graded early warning signal, and transmits the signal to an external monitoring system through a communication interface. The compensation algorithm internally provided in the airflow disturbance correction unit (2) is realized by establishing a heat dissipation airflow vector field dynamic model, the model constructs an airflow disturbance action matrix based on the spatial distribution parameters of the sensor array unit (1) and the operation parameters of the axial flow fan, calculates the airflow vector intensity and direction suffered by the surface of each MEMS temperature sensor probe in real time, the algorithm generates a temperature compensation factor according to the vector field model, and dynamically superimposes the temperature compensation factor on the original temperature data stream of the sensor in a sliding time window mode, so that the non-steady-state disturbance of the heat dissipation airflow on the heat conduction process of the probe surface is eliminated, and the corrected temperature value accurately reflects the temperature of the module body. The real-time dynamic correction process is as follows: the airflow disturbance correction unit (2) firstly inputs the original temperature data collected by the sensor array unit (1) into the vector field dynamic model, calculates the heat exchange intensity deviation coefficient of the current airflow vector and the probe surface, generates a temperature correction gradient with phase compensation characteristics, and then integrates the gradient into the original data stream in the form of time series, thereby synchronously eliminating the temperature detection lag effect and amplitude attenuation effect caused by airflow instantaneous disturbance, and finally outputting a continuous temperature signal consistent with the thermal inertia characteristics of the module body. The synchronous processing process of the multi-parameter coupling analysis unit (3) adopts a double-channel signal fusion architecture: the vibration signal channel extracts the original vibration waveform in the preset frequency band through adaptive band-pass filtering, generates a vibration energy spectrum through time-frequency transformation, the temperature change channel carries out first-order differential processing on the corrected continuous temperature signal, extracts a temperature rising rate curve, and the data streams of the two channels are synchronized through a time stamp alignment module, thereby providing a time sequence reference for subsequent feature correlation. The extraction of the high-frequency micro-vibration characteristics specifically includes the following steps: the multi-parameter coupling analysis unit (3) identifies an energy sudden increase frequency band in the vibration energy spectrum, establishes a resonance peak matching template based on the mechanical resonance characteristics of the tab loosening fault, compares the envelope similarity of the vibration energy spectrum and the resonance peak template, screens out vibration events meeting the feature matching conditions, and marks the starting time point and the duration of the vibration events. 2.The multi-parameter intelligent monitoring device for energy storage cabinet based on MEMS sensor array according to claim 1, characterized in that: ​ 3.The multi-parameter intelligent monitoring device for energy storage cabinet based on MEMS sensor array according to claim 1, characterized in that: ​ ​ 4. The multi-parameter intelligent monitoring device for energy storage cabinet based on MEMS sensor array according to claim 3, characterized in that: ​ ​ 5. The MEMS sensor array-based multi-parameter intelligent monitoring device for energy storage cabinet according to claim 4, characterized in that: ​ ​ 6.The MEMS sensor array-based multi-parameter intelligent monitoring device for energy storage cabinet according to claim 5, characterized in that: The high-frequency micro-vibration characteristic unique to the tab looseness is characterized by: The mechanical resonance phenomenon caused by the periodic change of the contact resistance at the battery tab junction, the vibration energy of which forms a shock-like waveform with a steep rising edge and a slow decay tail in a specific frequency band, and the time domain interval of adjacent vibration events has a strong correlation with the phase of the battery charging and discharging current. 7.The MEMS sensor array based multi-parameter intelligent monitoring device for energy storage cabinet according to claim 4, characterized in that: The synchronous change rule of the associated temperature rise rate is specifically: After detecting the starting time point of the high-frequency micro-vibration characteristic, the multi-parameter coupling analysis unit (3) starts the trend tracking window of the temperature rise rate curve, and determines whether the vibration characteristic and the temperature rise rate exist spatio-temporal coupling by analyzing the monotonic increasing characteristic of the temperature rise rate and the acceleration change in the window. The coupling is manifested as a positive correlation between the vibration energy spectrum amplitude increment and the temperature rise rate increment. 8.The MEMS sensor array based multi-parameter intelligent monitoring device of energy storage cabinet according to claim 7, characterized in that: The early fault determination logic includes a two-stage triggering mechanism: The primary determination is based on the number of times the high-frequency micro-vibration characteristic appears continuously exceeding the set threshold, and the secondary determination verifies whether the temperature rise rate in the corresponding period shows a continuous acceleration trend. Only when both levels of determination are met at the same time, the early fault warning signal is triggered, and false triggering caused by mechanical noise or transient heat dissipation fluctuations is suppressed. 9.The MEMS sensor array based multi-parameter intelligent monitoring device of energy storage cabinet according to claim 1, characterized in that: The hierarchical early warning signal generation process is: The fault decision and output unit (4) takes the corrected temperature absolute value, the coupling strength coefficient output by the multi-parameter coupling analysis unit (3), and the temperature rise rate acceleration ratio as three-dimensional decision inputs, maps the input parameters to the "normal", "early warning", and "emergency alarm" three states through the fault tree model, where the early warning state corresponds to the high-frequency micro-vibration and temperature rise acceleration coupling event, and the emergency alarm state corresponds to the temperature absolute value breaking through the safety threshold.

10. The MEMS sensor array-based multi-parameter intelligent monitoring device for energy storage cabinet according to claim 9, characterized in that: The core decision rules of the fault tree model include: When the high-frequency micro-vibration and temperature rise acceleration coupling event is detected, the early warning signal is triggered regardless of whether the temperature absolute value breaks through the safety threshold. If the temperature rise rate acceleration ratio breaks through the critical value during the early warning period, it is immediately upgraded to emergency alarm. At the same time, the system continues to track the temperature rise rate returning to normal characteristics after the coupling event is removed, realizing fault self-recovery discrimination.

Images