Lithium battery health state monitoring method and system based on internet of things

Abstract

This invention discloses a method and system for monitoring the health status of lithium batteries based on the Internet of Things (IoT), belonging to the field of battery monitoring technology. The method includes the following steps: monitoring the current and voltage of the battery pack through battery management nodes to generate event records; calculating battery health values; calculating the time deviation values ​​of each battery management node relative to a virtual time axis; generating data sections covering the same time nodes across the entire virtual time axis through resampling and missing data extrapolation; calculating state differences to generate a continuous sequence of difference features; and predicting the remaining battery usage time. This invention constructs a virtual time axis using the recording time of charge / discharge state transition events as synchronization anchors and performs dynamic time warping and resampling on the health status sequences reported by each node. This solves the problems in existing technologies where the lack of a unified time reference for each monitoring node leads to time base deviations in health status assessments and inaccurate consistency diagnosis at the convergence point.

Description

Technical Field

[0001] This invention relates to the field of battery monitoring technology, and in particular to a method and system for monitoring the health status of lithium batteries based on the Internet of Things. Background Technology

[0002] With the large-scale deployment of lithium battery energy storage power stations in areas such as power peak shaving and frequency regulation, and new energy consumption, and the need for continuous health management of electric vehicle power batteries throughout their entire life cycle, IoT-based distributed monitoring architecture has become the mainstream solution for battery status awareness. By deploying edge nodes with independent data acquisition and computing capabilities at each branch or module of the battery cluster, real-time operational data such as voltage, current, and temperature are acquired, and the calculation results are uploaded to the aggregation terminal for overall cluster consistency analysis and operation and maintenance decisions.

[0003] Existing IoT-based lithium battery health status monitoring systems independently trigger the health status estimation process through each edge node relying on its local crystal oscillator clock. The estimation results are then sent to the aggregation controller after being timestamped. The aggregation controller calculates the battery pack consistency index, performs equalization control, and predicts the remaining service life based on this information.

[0004] However, in actual engineering operations, due to the lack of a unified time base coordination mechanism among edge nodes, the reported health status estimates often correspond to different physical sections with millisecond to second-level deviations. When the aggregation end performs consistency diagnosis, it forcibly aligns state data from different moments, resulting in a deviation between the constructed battery pack imbalance and the actual situation, easily producing false differences or masking the true imbalance. This problem further leads to erroneous or missed equalization operations, causing unnecessary energy transfer losses and accelerated aging. At the same time, the remaining lifetime prediction based on inaccurate consistency data deviates from the actual degradation trajectory, causing errors in the decision-making regarding tiered utilization grouping and retirement timing, affecting the safety and economic benefits of battery assets. Summary of the Invention

[0005] This application provides a lithium battery health status monitoring method and system based on the Internet of Things, which solves the problems in the prior art where the lack of a unified time reference for each monitoring node leads to time base deviation in health status assessment, inaccurate consistency diagnosis at the aggregation end, and consequently, malfunctions in balance and deviations in lifespan prediction. It achieves accurate time alignment for distributed node health status assessment, ensuring that the assessment results truly reflect the battery pack status under the same physical cross-section.

[0006] This application provides a method for monitoring the health status of lithium batteries based on the Internet of Things, including: monitoring the current and voltage of the battery pack through a battery management node, determining charge / discharge state transition events, and generating event records containing the occurrence time, event type, and channel marker;

[0007] Calculate the battery health value, and combine the local timestamp, battery health value, and unused event log timestamp into a reporting data packet for transmission;

[0008] Receive data packets reported by each battery management node, compare the event record times and determine the synchronization reference point, and calculate the time deviation value of each battery management node relative to the virtual time axis;

[0009] Based on the time deviation value, the local timestamp is converted into a virtual time. Through resampling and extrapolation of missing data, a data profile covering the same time node across the entire virtual time axis is generated.

[0010] Extract health data at the same virtual time in the data section, calculate the state differences, and generate a continuous sequence of difference features;

[0011] Adjust the battery equilibrium state based on the differential feature sequence to predict the remaining battery usage time.

[0012] Furthermore, the steps for determining charge / discharge state transition events and generating event records containing the occurrence time, event type, and channel tag include:

[0013] Monitor the bus current. When the current value crosses a set small threshold and changes from being greater than to being continuously less than the threshold, it is determined that charging and discharging has stopped. When the current value changes from being continuously less than to being greater than the threshold, it is determined that charging and discharging has started.

[0014] When the voltage of any single cell reaches the cutoff boundary, it is determined that the charging and discharging is cut off;

[0015] Upon completion of the determination, an event record is generated containing a timestamp, a type identifier, and a channel label. The type identifier records the current conversion status, and the channel label is used to specify the specific acquisition location where the data change occurred.

[0016] Furthermore, the step of combining the local timestamp, battery health value, and unused event log timestamps into a reporting data packet for transmission includes:

[0017] When the internal computing process in each battery management node completes a battery health value calculation according to its own independent clock, the output data structure, in addition to carrying the local timestamp of the node and the calculated battery health value, also includes the precise timestamp of the latest unused event record generated within the jurisdiction of the node since the last data output.

[0018] If no new event records appear within the current calculation period, then select to append the timestamp of the most recent historical event record.

[0019] The selected event record timestamp and the corresponding event record are written together into the data set to be reported and combined into a data packet. After the reporting action is completed, the timestamp of the event record is marked as used.

[0020] Furthermore, the steps of comparing event recording times and determining synchronization reference points, and calculating the time deviation values ​​of each battery management node relative to the virtual time axis include:

[0021] When the aggregation decision-making terminal receives data packets reported from multiple battery management nodes, it iterates through the event record timestamps in all data packets, searches for similar events whose time precision differs within a preset tolerance range, and identifies the event as a candidate synchronization reference point.

[0022] If the same type of event is reported by at least two different battery management nodes, the earliest event timestamp is selected as the reference time for the synchronization reference point.

[0023] A virtual timeline is constructed based on the reference time, and the zero point of the virtual timeline is bound to the reference time.

[0024] The difference between the timestamps of the same type of events in the other reported data packets and the timestamps of the events corresponding to the reference time is recorded as the inherent time deviation value of the battery management node relative to the virtual time axis.

[0025] Furthermore, the steps to generate data sections covering the same time points across the entire virtual timeline through resampling and extrapolation of missing data include:

[0026] Based on the constructed virtual timeline, the time position is adjusted for the battery health value sequence reported by each battery management node;

[0027] Based on the inherent time deviation value corresponding to each node, the local timestamp carried by each battery health value in the sequence is converted into the corresponding virtual moment on the virtual time axis;

[0028] After the conversion is completed, a unified gridded time reference is established to address the issue of inconsistent values ​​for multiple reported sequences on the virtual time axis, thereby obtaining health alignment data for each battery management node at the same virtual time.

[0029] Furthermore, through resampling and missing data extrapolation, data sections covering the same time nodes across the entire virtual timeline are generated. The missing data extrapolation steps include:

[0030] For a time interval between two adjacent synchronization reference points that does not contain other synchronization reference points, if at least one battery management node in the interval is missing a battery health value, or if it is necessary to take a value at any tiny moment on the virtual time axis in the interval, then the preset mathematical degradation model is called to perform smooth deduction calculation.

[0031] The smooth extrapolation calculation uses the battery health value after the node is aligned with the starting synchronization reference point of the interval as the initial value. Based on the current cycle number and historical degradation rate stored locally, the linear degradation trend is expressed by the first component representing calendar aging, and the non-linear degradation trend is expressed by the second component representing cycle charge and discharge. The two components are added together and extrapolated according to the time span from the starting reference point to the target virtual minute moment to obtain the estimated health value.

[0032] Furthermore, the steps for calculating state differences and generating a continuous sequence of difference features include:

[0033] After obtaining the data cross-section of the same time node that fully covers the entire virtual time axis, the battery consistency index is calculated to form the state difference characteristic data at that specific virtual moment.

[0034] The calculation process of the battery consistency index is automatically re-executed after each virtual time axis update, generating state difference feature data that is continuously distributed over time, and arranged in chronological order of virtual moments to form the difference feature sequence.

[0035] Furthermore, the steps for adjusting the battery equilibrium state based on the differential feature sequence include:

[0036] Obtain the maximum range and standard deviation in the differential feature sequence, and compare the maximum range and standard deviation with the preset multi-level action thresholds to determine the balance type and balance intensity.

[0037] When the maximum range is greater than the set trigger threshold, the two branches with the largest single-cell voltage difference corresponding to the maximum range are automatically selected and used as target equalization branches.

[0038] The corresponding energy flow direction parameters and energy transfer parameters are calculated based on the difference feature sequence, and the direction parameters and energy transfer parameters are combined into an equalization control command.

[0039] The equalization control command is sent directly to the corresponding battery management node, which then adjusts its internal equalization execution logic according to the equalization control command.

[0040] Further steps in predicting the remaining battery life include:

[0041] Obtain the average health value and the corresponding trend decay rate in the differential feature sequence;

[0042] Extract the total length of the virtual timeline accumulated by each battery management node during the virtual time transition process;

[0043] By using the average health value, the trend decay rate and the total length of the virtual time axis, the remaining available virtual time span between the current virtual moment and the battery reaching the preset health termination threshold is calculated.

[0044] The remaining available virtual time span is converted into the corresponding actual calendar time span using a preset time mapping rule, and the converted actual calendar time span is determined as the remaining battery usage time and output.

[0045] This application provides an IoT-based lithium battery health status monitoring system to implement an IoT-based lithium battery health status monitoring method, including: an event record generation module, a health value calculation module, a time deviation value calculation module, a data profile generation module, a difference feature sequence generation module, and a remaining time prediction module.

[0046] The event record generation module is used to monitor the current and voltage of the battery pack through the battery management node, determine the charge / discharge state transition event, and generate an event record containing the occurrence time, event type, and channel marker.

[0047] The health value calculation module is used to calculate the battery health value and combine the local timestamp, the battery health value and the unused event record time into a reporting data packet for transmission.

[0048] The time deviation value calculation module is used to receive the data packets reported by each battery management node, compare the event recording time and determine the synchronization reference point, and calculate the time deviation value of each battery management node relative to the virtual time axis.

[0049] The data profile generation module is used to convert the local timestamp into a virtual time based on the time deviation value, and generate a data profile covering the same time node across the entire virtual time axis by resampling and extrapolating from missing data.

[0050] The differential feature sequence generation module is used to extract health data at the same virtual time in the data section, calculate the state difference, and generate a continuous differential feature sequence.

[0051] The remaining time prediction module is used to adjust the battery balance state according to the difference feature sequence and predict the remaining battery usage time.

[0052] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0053] By using the electrical boundary events during the battery charge / discharge state transition as the time alignment reference across nodes, the monitoring nodes of each independent clock obtain a unified timing reference, eliminating the time base deviation of the health estimate and ensuring that the reported status corresponds to the same physical section.

[0054] Furthermore, in the process of aligning time using event records, a virtual timeline is constructed by determining a synchronization benchmark, and local timestamps are converted into virtual moments based on the deviations of each node. After resampling and extrapolation, isochronous data profiles are generated, ensuring that the health information used for consistency analysis at the aggregation end truly belongs to the same moment. This avoids the occurrence of false imbalance characteristics and the masking of real differences, thus ensuring the accuracy of imbalance state assessment results.

[0055] Furthermore, based on the real difference feature sequence, the grading threshold is automatically matched, and energy transfer is only performed when necessary to prevent unnecessary cycle loss and accelerated aging. At the same time, the remaining usable time is estimated by using the aligned health mean and decay rate, so that the tiered utilization decision is consistent with the actual health status and avoids asset value loss due to misjudgment of status. Attached Figure Description

[0056] Figure 1 A flowchart of an IoT-based lithium battery health status monitoring method provided in an embodiment of this application;

[0057] Figure 2 This is a schematic diagram of the structure of an IoT-based lithium battery health status monitoring system provided in an embodiment of this application. Detailed Implementation

[0058] This application provides an IoT-based lithium battery health status monitoring method and system, which solves the problems in the prior art where the lack of a unified time reference for each monitoring node leads to time base deviation in health status assessment, inaccurate consistency diagnosis at the convergence end, and consequently, malfunctions in equalization and deviations in lifespan prediction. By constructing a virtual time axis with the recording time of charge-discharge state transition events as the synchronization anchor point and dynamically resampling and aligning the health status sequences of each node, the application achieves accurate construction of isochronous sections for distributed node health status assessment, ensuring that the assessment results truly reflect the battery pack status under the same physical section.

[0059] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0060] like Figure 1The diagram shows a flowchart of an IoT-based lithium battery health status monitoring method provided in this application embodiment. The method is applied to an IoT-based lithium battery health status monitoring system and includes the following steps: monitoring the current and voltage of the battery pack through a battery management node, determining charge / discharge state transition events based on changes in current and voltage, and generating event records containing the occurrence time, event type, and channel marker.

[0061] Calculate the battery health value, and combine the local timestamp, battery health value, and unused event log timestamp into a reporting data packet for transmission;

[0062] Receive data packets reported by each battery management node, compare the event record times and determine the synchronization reference point, and calculate the time deviation value of each battery management node relative to the virtual time axis;

[0063] Based on the time deviation value, the local timestamp is converted into a virtual time. Through resampling and extrapolation of missing data, a data profile covering the same time node across the entire virtual time axis is generated.

[0064] Extract health data at the same virtual time in the data section, calculate the state differences, and generate a continuous sequence of difference features;

[0065] Adjust the battery equilibrium state based on the differential feature sequence to predict the remaining battery usage time.

[0066] Furthermore, the steps for determining charge / discharge state transition events and generating event records containing the occurrence time, event type, and channel tag include:

[0067] Monitor the bus current. When the current value crosses a set small threshold and changes from being greater than to being continuously less than the threshold, it is determined that charging and discharging has stopped. When the current value changes from being continuously less than to being greater than the threshold, it is determined that charging and discharging has started.

[0068] When the voltage of any single cell reaches the cutoff boundary, it is determined that the charging and discharging is cut off;

[0069] Upon completion of the determination, an event record is generated containing a timestamp, a type identifier, and a channel label. The type identifier records the current conversion status, and the channel label is used to specify the specific acquisition location where the data change occurred.

[0070] In this embodiment, software data acquisition modules distributed in each battery management node are first used to continuously monitor the bus current signal of the battery pack and the voltage signal of each individual battery at a preset sampling frequency (e.g., 10Hz to 100Hz).

[0071] The specific event determination and record generation logic is as follows: the real-time acquired current value is converted into an absolute value, and a moving average filter is introduced to eliminate instantaneous noise caused by IoT sensor disturbances.

[0072] When the absolute value of the bus current changes from being greater than the calibration threshold to being continuously less than the threshold for a duration exceeding the set filter window (e.g., 3 seconds), it is determined that a "charging cut-off" or "discharging cut-off" event has occurred.

[0073] Conversely, when the current changes from a state where it is consistently below the threshold to an absolute value greater than the threshold, it is determined as a "charging start" or "discharging start" event.

[0074] Simultaneously, boundary crossing monitoring is implemented for the individual cell voltage signals of all branches. Once the voltage of any single cell first triggers the preset maximum charging cutoff voltage boundary (e.g., for ternary lithium batteries or lithium iron phosphate batteries) or minimum discharging cutoff voltage boundary (e.g., 2.5V), even if the current has not completely dropped, the corresponding charging / discharging cutoff event is forcibly determined to have occurred.

[0075] The moment the event determination is completed, a structured event record is immediately constructed in local memory, containing three core data fields:

[0076] High-precision timestamp: Reads the current absolute time value obtained from a local high-precision timer (with a resolution of at least microseconds);

[0077] Event type identifier: Use an enumeration value (such as representing the start of charging, representing the end of charging, etc.) to clearly mark the current transition state;

[0078] Channel label: Records the specific current sensor number or individual voltage acquisition channel number that triggered the event, used to specify the specific acquisition location where the data change occurred.

[0079] Furthermore, the step of combining the local timestamp, battery health value, and unused event log timestamps into a reporting data packet for transmission includes:

[0080] When the internal computing process in each battery management node completes a battery health value calculation according to its own independent clock, the output data structure, in addition to carrying the local timestamp of the node and the calculated battery health value, also includes the precise timestamp of the latest unused event record generated within the jurisdiction of the node since the last data output.

[0081] If no new event records appear within the current calculation period, then select to append the timestamp of the most recent historical event record.

[0082] The selected event record timestamp and the corresponding event record are written together into the data set to be reported and combined into a data packet. After the reporting action is completed, the timestamp of the event record is marked as used.

[0083] In this embodiment, within each battery management node, the internal independent State of Health (SOH) estimation process runs independently according to its local independent crystal oscillator clock and triggering rules (e.g., every hour or after each complete charge-discharge cycle). Due to differences in manufacturing processes and temperature drift among the crystal oscillators of each node, there is an asynchronous deviation between their local clocks on the order of milliseconds to seconds.

[0084] After a node completes a battery health calculation, specific packet assembly logic is executed to ensure the secure and distortion-free transmission of this time-base biased data to the aggregation point:

[0085] In the output data structure, in addition to allocating regular fields to carry the local timestamp when the node completed its calculation and the corresponding battery health percentage (SOH percentage), a pointer is used to retrieve the local event storage area. This searches for the most recently generated and unreferenced event record within the node's jurisdiction since the last data reporting action. If such an unused record exists, its precise timestamp is extracted and written to the data packet as an additional alignment feature field.

[0086] If the system operates smoothly within the current health calculation period and no new state transition events occur, the timestamp of the most recent historical event record is extracted and appended. Finally, the local timestamp, battery health value, selected event record timestamp, and the complete content of the event record itself (including type identifier and channel label) are written into the reporting dataset, combined to form a reporting data packet, and sent out through an IoT communication link (such as MQTT or CoAP protocol).

[0087] Once successfully sent, the status bit of the event record is changed from "unreferenced" to "used" to ensure that it will not be reported as the latest event in subsequent calculation cycles.

[0088] Furthermore, the steps of comparing event recording times and determining synchronization reference points, and calculating the time deviation values ​​of each battery management node relative to the virtual time axis include:

[0089] When the aggregation decision-making terminal receives data packets reported from multiple battery management nodes, it iterates through the event record timestamps in all data packets, searches for similar events whose time precision differs within a preset tolerance range, and identifies the event as a candidate synchronization reference point.

[0090] If the same type of event is reported by at least two different battery management nodes, the earliest event timestamp is selected as the reference time for the synchronization reference point.

[0091] A virtual timeline is constructed based on the reference time, and the zero point of the virtual timeline is bound to the reference time.

[0092] The difference between the timestamps of the same type of events in the other reported data packets and the timestamps of the events corresponding to the reference time is recorded as the inherent time deviation value of the battery management node relative to the virtual time axis.

[0093] In this embodiment, the aggregation decision-making end (such as a cloud server or station control gateway) continuously receives data packets reported from multiple different battery management nodes through a network interface. Upon receiving the data, the aggregation end software initiates a virtual time alignment process to eliminate the impact of inconsistencies in the clocks of the nodes.

[0094] The specific execution steps are as follows: First, all received reported data packets are parsed concurrently to extract the event record timestamps and event type identifiers. Then, these timestamps are traversed within a time sliding window to find events of the same type with completely identical event type identifiers (e.g., both being charging cut-off events) and whose pairwise timestamp values ​​differ within a preset tolerance range. These identified events indicate that they physically belong to the same heap-wide environment-triggered event and are therefore identified as candidate synchronization reference points.

[0095] To establish a unified reference system, if a certain type of event is simultaneously reported by at least two different battery management nodes, the timestamps of the event carried by these nodes are compared, and the earliest numerically occurring timestamp is selected as the reference reference time for the synchronization benchmark (denoted as ). ).

[0096] Next, a one-dimensional linear virtual timeline is constructed in memory, and a binding operation is performed: the reference zero time of this virtual timeline is set... ) and the reference base time Perform a strong binding.

[0097] For each of the remaining battery management nodes, the timestamp of the same type of event recorded in its reported data packet corresponding to the reference base time (denoted as...) The mathematical difference between the node and the reference time is defined as the inherent time deviation of that node relative to the virtual time axis (denoted as ). The calculation formula is as follows:

[0098] ;

[0099] This inherent time deviation value reflects the amount by which the node's local clock leads or lags relative to the time axis of the unified physical section, and is temporarily stored in the node feature matrix in memory.

[0100] Furthermore, the steps to generate data cross-sections covering the same time points across the entire virtual timeline through resampling and extrapolation of missing data include:

[0101] Based on the constructed virtual timeline, the time position is adjusted for the battery health value sequence reported by each battery management node;

[0102] Based on the inherent time deviation value corresponding to each node, the local timestamp carried by each battery health value in the sequence is converted into the corresponding virtual time on the virtual time axis; the specific conversion method is to add the local timestamp to the inherent time deviation value of the node.

[0103] After the conversion is completed, to address the issue of inconsistent values ​​on the virtual time axis for multiple reported sequences, a unified gridded time reference is established. The time step of this grid is equal to the preset health update cycle. The data of each battery management node on the virtual time axis are linearly interpolated according to this grid to obtain the health aligned data of each battery management node at the same virtual time.

[0104] In this embodiment, the inherent time deviation value of each node is obtained. Then, the deviation value is used to reconstruct the time series of all reported health data with local time base.

[0105] First, for any battery management node The reported battery health value sequence is processed by a time conversion module that iterates through each element in the sequence. Let the local timestamp of a given element be... Convert it to the corresponding virtual moment on the virtual timeline according to the following formula. :

[0106] ;

[0107] After this additive transformation, the health values ​​that originally belonged to different time bases were successfully projected onto the same virtual time axis. However, due to the randomness of the calculation triggered by each node or slight fluctuations in the period, the specific values ​​of the multiple data sequences after projection on the virtual time axis are often interleaved and do not overlap, making it impossible to directly compare horizontal consistency.

[0108] To address this issue, a unified gridded time reference is established on a virtual timeline. The time step of this grid ( The value is strictly set to be equal to the preset standard battery health update cycle (e.g., every 24 hours or every 10 virtual time units). Then, for each node's data points scattered on the virtual time axis, a linear numerical interpolation algorithm is used to resample them onto the grid nodes.

[0109] Let the virtual time of the target grid be And this moment happens to fall exactly between two adjacent virtual moments after the transformation. and Between (the corresponding health values ​​are respectively) and Then, the health alignment data at that grid time is calculated using the following linear interpolation formula. :

[0110] ;

[0111] By performing this resampling on all data across all nodes, health-aligned data for each battery management node was obtained on the same virtual time grid.

[0112] Furthermore, through resampling and missing data extrapolation, data sections covering the same time nodes across the entire virtual timeline are generated. The missing data extrapolation steps include:

[0113] For a time interval between two adjacent synchronization reference points that does not contain other synchronization reference points, if at least one battery management node in the interval is missing a battery health value, or if it is necessary to take a value at any tiny moment on the virtual time axis in the interval, then the preset mathematical degradation model is called to perform smooth deduction calculation.

[0114] The smooth extrapolation calculation uses the battery health value after the node is aligned with the starting synchronization reference point of the interval as the initial value. Based on the current cycle number and historical degradation rate stored locally, the linear degradation trend is expressed by the first component representing calendar aging, and the non-linear degradation trend is expressed by the second component representing cycle charge and discharge. The two components are added together and extrapolated according to the time span from the starting reference point to the target virtual minute moment to obtain the estimated health value.

[0115] In this embodiment, in actual IoT communication, due to network packet loss, temporary node offline or sampling anomalies, some nodes often have long data gaps within the time interval between two adjacent synchronization reference points, or the resampling grid cannot obtain enough nearby known points.

[0116] If at least one node in the interval lacks health data, or if it is necessary to take continuous values ​​at an ultra-high resolution for any tiny moment on the virtual timeline, the preset double exponential degradation model will be automatically invoked for smoothing.

[0117] The logical construction and recursive process of the mathematical degradation model are as follows: The deduction engine uses the health value of the missing node after alignment at the starting synchronization reference point of the current time interval as the initial value (denoted as ). Simultaneously, it reads the current cumulative charge / discharge cycles of the battery pack stored locally or in the central database of the battery management node. ) and the historical degradation rate parameter obtained by fitting historical operating data.

[0118] The mathematical analytical expression of the bi-exponential degradation model is defined as follows:

[0119] ;

[0120] In the formula: The estimated health value of the target at a virtual, minute moment. The initial known health value at the starting synchronization reference point of this interval is obtained through the aforementioned resampling or the most recent valid alignment value. This refers to the virtual time span from the initial synchronization reference point of the interval to the target virtual infinitesimal moment, expressed in hours (h), and is directly obtained by subtracting the target moment from the initial moment. In the time span Within this period, the actual increase in the number of charge-discharge cycles for the battery is a real number not less than 0. In the absence of precise counting, it can be calculated by multiplying the average daily cycle frequency in the historical records of this node by the time span. Transformed into The calendar aging degradation coefficient primarily expresses the linear decline in battery health caused by the passage of time and calendar aging; its value range is limited to [specific range missing]. to The method for obtaining the SOH value is as follows: during the long-term storage period when the battery is not being charged or discharged, the SOH value is obtained by performing a univariate linear regression analysis on the rate of change of SOH over time. This is the cyclic aging degradation coefficient, primarily expressed through exponential components to represent the nonlinear accelerated degradation trend caused by frequent charge-discharge cycles and the superposition of electrochemical internal stresses. Its value range is limited to [specific range missing]. to The method of obtaining this information is as follows: under continuous charge and discharge operation conditions, the amount of health lost is deducted from the total decrease in health. After calculating the calendar aging portion, the remaining decay amount is obtained by exponential fitting of the change with respect to the number of cycles.

[0121] The inference engine substitutes the above parameters into the model for recursive calculation, and then outputs the estimated health value of the target at a virtual minute moment. This estimated health value is integrated and supplemented with the aligned data generated by resampling in the previous steps, together forming a cross-sectional health status dataset that is spatially complete, continuously distributed, and temporally isochronous on the virtual time axis.

[0122] Furthermore, the steps for calculating state differences and generating a continuous sequence of difference features include:

[0123] After obtaining the data cross-section of the same time node that fully covers the entire virtual time axis, the battery consistency index is calculated. Specifically, at any selected virtual moment, the health alignment data or estimated health value of all battery management nodes participating in the calculation at that moment is extracted, and the maximum range, standard deviation and distribution skewness of these health values ​​are calculated respectively. The calculated maximum range, standard deviation and distribution skewness are combined together to form the state difference feature data at that specific virtual moment.

[0124] The calculation process of the battery consistency index is automatically re-executed after each virtual time axis update, generating state difference feature data that is continuously distributed over time, and arranged in chronological order of virtual moments to form the difference feature sequence.

[0125] In this embodiment, after obtaining the isochronous cross-sectional health status dataset that fully covers the entire virtual time axis, quantitative calculations are performed on the differences of the entire stack or cluster of batteries.

[0126] For any virtual time point on the virtual timeline, perform horizontal slice extraction, and extract the health-aligned data (from resampling) and the estimated health values ​​(from degradation model deduction) of all battery management nodes participating in the calculation at that time into a single set. Let the total number of nodes participating in the calculation at that time be... indivual.

[0127] Subsequently, through mathematical operations, three key statistical indicators for the health status data within this set were calculated:

[0128] Maximum range ( The formula for calculating the health score is: the maximum health score minus the minimum health score in the set.

[0129] ;

[0130] Standard deviation ( ): Reflects the degree of dispersion of the health status of each node around the average value, and is calculated using the following formula:

[0131] ;

[0132] in, This is the arithmetic mean of the health values ​​of all nodes at that moment.

[0133] Distribution skewness (Skew): Reflects the asymmetry in the distribution of health status, used to capture the "long tail" phenomenon of extreme premature aging in individual battery cells. The calculation formula is:

[0134] ;

[0135] The calculated maximum range, standard deviation, and skewness are combined to construct a three-dimensional mathematical vector: This vector serves as the characteristic data of battery state differences at that specific virtual moment.

[0136] Whenever the virtual timeline grid moves forward or the data is updated, the aforementioned horizontal slicing and statistical indicator calculation process is automatically retried in the background. The resulting series of state difference feature data, continuously distributed over time, are linearly arranged strictly according to the chronological order of the virtual moments, ultimately forming a high-dimensional difference feature sequence. This sequence, as the core feature, is seamlessly passed to subsequent control and prediction logic.

[0137] Furthermore, the steps for adjusting the battery equilibrium state based on the differential feature sequence include:

[0138] Obtain the maximum range and standard deviation in the differential feature sequence, and compare the maximum range and standard deviation with the preset multi-level action thresholds to determine the balance type and balance intensity.

[0139] When the maximum range is greater than the set trigger threshold, the two branches with the largest single-cell voltage difference corresponding to the maximum range are automatically selected and used as target equalization branches.

[0140] The corresponding energy flow direction parameters and energy transfer parameters are calculated based on the difference feature sequence, and the direction parameters and energy transfer parameters are combined into an equalization control command.

[0141] The equalization control command is sent directly to the corresponding battery management node, which then adjusts its internal equalization execution logic according to the equalization control command.

[0142] In this embodiment, after receiving the difference feature sequence, the high-precision time alignment index is used to guide the underlying energy equalization operation in real time, avoiding erroneous active equalization caused by time base deviation.

[0143] The specific execution logic is as follows: Real-time acquisition of the maximum range corresponding to the latest time step in the differential feature sequence (…). ) and standard deviation ( A tiered action threshold table is pre-defined. The threshold for triggering a level one mild imbalance is set as follows: The threshold for triggering a level 2 severe imbalance is .

[0144] The decision logic will use the current maximum range Compare with these thresholds: when Less than When the condition is within a good consistency range, the equilibrium type remains "off," suppressing any energy transfer; when Crossing the trigger threshold to enter and When in between, activate the "mild active balancing" mode; when Exceed At that time, the "high-intensity forced equilibrium" mode is activated.

[0145] Once it is determined that equalization needs to be initiated, the system compares the data from the isochronous cross-section and automatically finds and filters out the two battery paths that are directly related to the generation of the maximum range and have the largest absolute value of the single-cell voltage difference in the current physical cross-section. These two paths are then designated as the target equalization branches (one is the source branch and the other is the destination branch).

[0146] Next, based on the standard deviation The magnitude of the amplitude is used to calculate the specific energy transfer parameters (unit: ampere-hours Ah or watt-hours Wh) through a proportional function, and the energy flow direction parameters are determined based on the relative health of the two branches.

[0147] Subsequently, the direction parameters and energy transfer parameters are encapsulated into a standard low-level control frame, forming a balancing control command. This command is directly sent to the corresponding low-level battery management node via the IoT bus. Upon receiving the command, the hardware driver inside the battery management node immediately drives its internal switching transistors and bidirectional DC-DC converter to precisely execute energy transfer between specified branches, thereby achieving closed-loop regulation of the system's consistency state.

[0148] Further steps in predicting the remaining battery life include:

[0149] Obtain the average health value and the corresponding trend decay rate in the differential feature sequence;

[0150] Extract the total length of the virtual timeline accumulated by each battery management node during the virtual time transition process;

[0151] By using the average health value, the trend decay rate and the total length of the virtual time axis, the remaining available virtual time span between the current virtual moment and the battery reaching the preset health termination threshold is calculated.

[0152] After obtaining the remaining available virtual time span, the remaining available virtual time span is converted into the corresponding actual calendar time span using a preset time mapping rule, and the converted actual calendar time span is determined as the remaining battery usage time and output.

[0153] In this embodiment, the differential feature sequence is used as the input stream to perform long-term asset valuation of battery retirement and cascade utilization.

[0154] The specific prediction and mapping mechanism is as follows: First, extract the average health value from the differential feature sequence ( And by estimating the first or second derivative of the average change trend within a historical window in the differential feature sequence, the trend decay rate of the current stage's health (denoted as ) is calculated. ).

[0155] Simultaneously, the total length of the virtual time axis accumulated by each battery management node during the virtual time transition is extracted from the node feature matrix. This length represents the effective operating span that the battery actually experiences under the aligned reference frame.

[0156] Using the following linear or progressive lifecycle estimation model, calculate the time remaining until the battery reaches the preset threshold for termination of its tiered use health (e.g., ...). or The remaining available virtual time span between () ):

[0157] ;

[0158] In the formula, and All figures are percentages. The unit is "the amount of health decrease per unit of virtual time".

[0159] Since this lifetime is calculated on a virtual timeline after time skew has been eliminated, in order to make it relevant to practical engineering applications, the remaining available virtual time span must be obtained. Afterwards, the preset time mapping rules must be invoked to restore it.

[0160] Specifically, by analyzing historical operation logs, the actual calendar time consumed is calculated. ) and the corresponding total length of the virtual timeline ( The proportional coefficient between ) Then, the remaining available virtual time span is multiplied by this scaling factor to smoothly convert it into the corresponding actual calendar time span. ):

[0161] ;

[0162] Finally, the actual calendar time span obtained from the conversion (e.g., remaining time) is determined as the remaining usage time of the battery array and officially output. This output result is directly used to guide energy storage power station managers to accurately select the decision-making time for future tiered utilization, dismantling and grouping, and decommissioning and reorganization, avoiding misjudgment of asset value.

[0163] like Figure 2 The diagram shown is a structural schematic of an IoT-based lithium battery health status monitoring system provided in this application embodiment. The IoT-based lithium battery health status monitoring system provided in this application embodiment includes: an event record generation module, a health value calculation module, a time deviation value calculation module, a data profile generation module, a differential feature sequence generation module, and a remaining time prediction module.

[0164] The event record generation module is used to monitor the current and voltage of the battery pack through the battery management node, determine the charge / discharge state transition event, and generate an event record containing the occurrence time, event type, and channel marker.

[0165] The health value calculation module is used to calculate the battery health value and combine the local timestamp, the battery health value and the unused event record time into a reporting data packet for transmission.

[0166] The time deviation value calculation module is used to receive the data packets reported by each battery management node, compare the event recording time and determine the synchronization reference point, and calculate the time deviation value of each battery management node relative to the virtual time axis.

[0167] The data profile generation module is used to convert the local timestamp into a virtual time based on the time deviation value, and generate a data profile covering the same time node across the entire virtual time axis by resampling and extrapolating from missing data.

[0168] The differential feature sequence generation module is used to extract health data at the same virtual time in the data section, calculate the state difference, and generate a continuous differential feature sequence.

[0169] The remaining time prediction module is used to adjust the battery balance state according to the difference feature sequence and predict the remaining battery usage time.

[0170] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0171] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0172] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0173] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0174] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0175] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring the health status of lithium batteries based on the Internet of Things, characterized in that, Includes the following steps: By monitoring the current and voltage of the battery pack through the battery management node, the charging and discharging state transition events are determined, and event records containing the occurrence time, event type, and channel tag are generated. Calculate the battery health value, and combine the local timestamp, battery health value, and unused event log timestamp into a reporting data packet for transmission; Receive data packets reported by each battery management node, compare the event record times and determine the synchronization reference point, and calculate the time deviation value of each battery management node relative to the virtual time axis; Based on the time deviation value, the local timestamp is converted into a virtual time. Through resampling and extrapolation of missing data, a data profile covering the same time node across the entire virtual time axis is generated. Extract health data at the same virtual time in the data section, calculate the state differences, and generate a continuous sequence of difference features; Adjust the battery equilibrium state based on the differential feature sequence to predict the remaining battery usage time.

2. The method for monitoring the health status of lithium batteries based on the Internet of Things as described in claim 1, characterized in that, The steps for determining charge / discharge state transition events and generating event records containing the occurrence time, event type, and channel tag include: Monitor the bus current. When the current value crosses a set small threshold and changes from being greater than to being continuously less than the threshold, it is determined that charging and discharging has stopped. When the current value changes from being continuously less than to being greater than the threshold, it is determined that charging and discharging has started. When the voltage of any single cell reaches the cutoff boundary, it is determined that the charging and discharging is cut off; Upon completion of the determination, an event record is generated containing a timestamp, a type identifier, and a channel label. The type identifier records the current conversion status, and the channel label is used to specify the specific acquisition location where the data change occurred.

3. The method for monitoring the health status of lithium batteries based on the Internet of Things as described in claim 1, characterized in that, The steps for combining the local timestamp, battery health value, and unused event log timestamps into a reporting data packet and sending it include: When the internal computing process in each battery management node completes a battery health value calculation according to its own independent clock, the output data structure, in addition to carrying the local timestamp of the node and the calculated battery health value, also includes the precise timestamp of the latest unused event record generated within the jurisdiction of the node since the last data output. If no new event records appear within the current calculation period, then select to append the timestamp of the most recent historical event record. The selected event record timestamp and the corresponding event record are written together into the data set to be reported and combined into a data packet. After the reporting action is completed, the timestamp of the event record is marked as used.

4. The method for monitoring the health status of lithium batteries based on the Internet of Things as described in claim 1, characterized in that, The steps for comparing event recording times and determining the synchronization reference point, and calculating the time deviation value of each battery management node relative to the virtual time axis include: When the aggregation decision-making terminal receives data packets reported from multiple battery management nodes, it iterates through the event record timestamps in all data packets, searches for similar events whose time precision differs within a preset tolerance range, and identifies the event as a candidate synchronization reference point. If the same type of event is reported by at least two different battery management nodes, the earliest event timestamp is selected as the reference time for the synchronization reference point. A virtual timeline is constructed based on the reference time, and the zero point of the virtual timeline is bound to the reference time. The difference between the timestamps of the same type of events in the other reported data packets and the timestamps of the events corresponding to the reference time is recorded as the inherent time deviation value of the battery management node relative to the virtual time axis.

5. The method for monitoring the health status of lithium batteries based on the Internet of Things as described in claim 1, characterized in that, The steps for generating data profiles covering the same time points across the entire virtual timeline through resampling and extrapolation of missing data include: Based on the constructed virtual timeline, the time position is adjusted for the battery health value sequence reported by each battery management node; Based on the inherent time deviation value corresponding to each node, the local timestamp carried by each battery health value in the sequence is converted into the corresponding virtual moment on the virtual time axis; After the conversion is completed, a unified gridded time reference is established to address the issue of inconsistent values ​​for multiple reported sequences on the virtual time axis, thereby obtaining health alignment data for each battery management node at the same virtual time.

6. The method for monitoring the health status of lithium batteries based on the Internet of Things as described in claim 1, characterized in that, By resampling and extrapolating missing data, data sections covering the same time points across the entire virtual timeline are generated. The steps for extrapolating missing data include: For a time interval between two adjacent synchronization reference points that does not contain other synchronization reference points, if at least one battery management node in the interval is missing a battery health value, or if it is necessary to take a value at any tiny moment on the virtual time axis in the interval, then the preset mathematical degradation model is called to perform smooth deduction calculation. The smooth extrapolation calculation uses the battery health value after the node is aligned with the starting synchronization reference point of the interval as the initial value. Based on the current cycle number and historical degradation rate stored locally, the linear degradation trend is expressed by the first component representing calendar aging, and the non-linear degradation trend is expressed by the second component representing cycle charge and discharge. The two components are added together and extrapolated according to the time span from the starting reference point to the target virtual minute moment to obtain the estimated health value.

7. The method for monitoring the health status of lithium batteries based on the Internet of Things as described in claim 1, characterized in that, The steps for calculating state differences and generating a continuous sequence of difference features include: After obtaining the data cross-section of the same time node that fully covers the entire virtual time axis, the battery consistency index is calculated to form the state difference characteristic data at that specific virtual moment. The calculation process of the battery consistency index is automatically re-executed after each virtual time axis update, generating state difference feature data that is continuously distributed over time, and arranged in chronological order of virtual moments to form the difference feature sequence.

8. The method for monitoring the health status of lithium batteries based on the Internet of Things as described in claim 1, characterized in that, The steps for adjusting the battery equilibrium state based on the differential characteristic sequence include: Obtain the maximum range and standard deviation in the differential feature sequence, and compare the maximum range and standard deviation with the preset multi-level action thresholds to determine the balance type and balance intensity. When the maximum range is greater than the set trigger threshold, the two branches with the largest single-cell voltage difference corresponding to the maximum range are automatically selected and used as target equalization branches. The corresponding energy flow direction parameters and energy transfer parameters are calculated based on the difference feature sequence, and the direction parameters and energy transfer parameters are combined into an equalization control command. The equalization control command is sent directly to the corresponding battery management node, which then adjusts its internal equalization execution logic according to the equalization control command.

9. The method for monitoring the health status of lithium batteries based on the Internet of Things as described in claim 1, characterized in that, The steps for predicting the remaining battery life include: Obtain the average health value and the corresponding trend decay rate in the differential feature sequence; Extract the total length of the virtual timeline accumulated by each battery management node during the virtual time transition process; By using the average health value, the trend decay rate and the total length of the virtual time axis, the remaining available virtual time span between the current virtual moment and the battery reaching the preset health termination threshold is calculated. The remaining available virtual time span is converted into the corresponding actual calendar time span using a preset time mapping rule, and the converted actual calendar time span is determined as the remaining battery usage time and output.

10. An IoT-based lithium battery health status monitoring system, used to implement the IoT-based lithium battery health status monitoring method according to any one of claims 1-9, characterized in that, include: The module includes: event log generation module, health value calculation module, time deviation value calculation module, data profile generation module, differential feature sequence generation module, and remaining time prediction module. The event record generation module is used to monitor the current and voltage of the battery pack through the battery management node, determine the charge / discharge state transition event, and generate an event record containing the occurrence time, event type, and channel marker. The health value calculation module is used to calculate the battery health value and combine the local timestamp, the battery health value and the unused event record time into a reporting data packet for transmission. The time deviation value calculation module is used to receive the data packets reported by each battery management node, compare the event recording time and determine the synchronization reference point, and calculate the time deviation value of each battery management node relative to the virtual time axis. The data profile generation module is used to convert the local timestamp into a virtual time based on the time deviation value, and generate a data profile covering the same time node across the entire virtual time axis by resampling and extrapolating from missing data. The differential feature sequence generation module is used to extract health data at the same virtual time in the data section, calculate the state difference, and generate a continuous differential feature sequence. The remaining time prediction module is used to adjust the battery balance state according to the difference feature sequence and predict the remaining battery usage time.

Images