A power generation side energy storage module health state monitoring and production scheduling management platform

Abstract

The present application relates to the technical field of energy storage module monitoring and production scheduling, in particular to a power generation side energy storage module health state monitoring and production scheduling management platform. It includes a feature acquisition unit that obtains the charge and discharge response and thermal field distribution characteristics under the working condition fluctuation of the energy storage module and extracts the time sequence characteristic parameters; an evolution prediction unit for inputting the time sequence characteristic parameters into the model to predict the preliminary remaining time, correcting to obtain the actual retirement remaining time, and synchronously predicting the expected average impedance and expected remaining capacity of the remaining healthy modules; an impedance matching decision unit constructs a customized module production parameter matrix; a flexible production scheduling unit analyzes the matrix and automatically adjusts the process formula to produce and manufacture matched replacement modules; a closed-loop feedback unit is used to update the cluster-level account after the replacement module is connected to the grid and issue parameter correction to the system for execution. Precise feedforward compensation of logistics aging is realized, and the risk of circulating current and overcharging is avoided.

Description

Technical Field

[0001] This invention relates to the field of energy storage module monitoring and production scheduling technology, specifically to a platform for monitoring the health status and managing the production scheduling of energy storage modules on the power generation side. Background Technology

[0002] With the surge in the scale of new energy grid connection, generation-side energy storage systems, as a key new energy storage infrastructure, play a core role in "primary frequency regulation" and "peak shaving and valley filling." In real-world field operation scenarios, energy storage power stations frequently experience high-frequency step dispatch commands from the grid (high-rate charge and discharge impacts), coupled with structural heat dissipation dead zones inside the container, often leading to abnormally high polarization resistance of individual modules within the cluster. To prevent thermal runaway of this deteriorated module, the on-site battery management system (BMS) can only trigger the "weakest link effect," fully reducing the upper limit of the charge and discharge current of the entire battery cluster. To restore rated capacity, the power station operator must place an order with the manufacturing plant to custom-produce a matching replacement module, which is the common "short-board replacement" operation and maintenance scenario in engineering.

[0003] However, in existing energy storage component replacement and maintenance scenarios, serious impedance mismatch incidents frequently occur, where "new modules arrive at the site but still cannot be fully connected to the grid." From the on-site reporting of the anomaly, factory production scheduling, to long-distance trunk logistics and on-site storage, there is often a physical time span of ten days or even longer. Existing Manufacturing Execution Systems (MES) usually mechanically copy the impedance data of the old modules "at this moment" on-site to issue production instructions; however, they ignore the fact that the new modules undergo a gradual "normal temperature calendar aging" during long-distance logistics, while the old battery clusters on-site are still undergoing severe "non-linear cyclic aging" under harsh operating conditions. This misalignment of evolutionary trajectories means that when the new modules are actually delivered to the site and connected to the grid, their impedance has already deviated from that of the continuously deteriorating old battery clusters, which can easily trigger intra-cluster circulating currents after grid connection.

[0004] In summary, existing platforms mostly rely on artificial intelligence models (such as conventional time series prediction networks). Due to the lack of constraints from underlying thermodynamic mechanisms, these deep learning algorithms are prone to severe parameter drift when extrapolating long-term evolutions or encountering extreme and sudden operating conditions. They are unable to provide decision feedforwards with engineering reliability. Therefore, there is an urgent need for a power generation-side energy storage module health status monitoring and production scheduling management platform to achieve industrial internet collaboration from underlying physical perception to intelligent manufacturing scheduling. Summary of the Invention

[0005] The purpose of this invention is to provide a health status monitoring and production scheduling management platform for energy storage modules on the power generation side, so as to solve the problem mentioned in the background art that the existing platform lacks feedforward prediction and compensation for time-varying aging of logistics.

[0006] To achieve the above objectives, the present invention aims to provide a health status monitoring and production scheduling management platform for power generation-side energy storage modules, comprising:

[0007] The feature acquisition unit is used to continuously acquire the charge and discharge response data and thermal field distribution characteristics of each energy storage module in the power generation side energy storage battery cluster under operating condition fluctuations, and extract time-series feature parameters to characterize the micro-deterioration of the modules. ;

[0008] The evolution prediction unit is used to input the time-series feature parameters into a preset health evolution model to predict the initial remaining time when the target deteriorating module reaches the retirement threshold. And combining the charge / discharge response data with the thermal field distribution characteristics, the preliminary remaining time is determined. Perform inverse nonlinear reduction correction to obtain the actual remaining decommissioning time. Simultaneously, the remaining time before actual retirement is predicted. At any given time, the expected average impedance of the remaining healthy modules within the battery cluster containing the target degraded module. With expected remaining capacity ;

[0009] Impedance matching decision unit, the impedance matching decision unit being configured to match the expected mean impedance and expected remaining capacity Based on, and in conjunction with the actual remaining time of retirement. Construct a customized module production parameter matrix to avoid impedance mismatch between new and old modules within the cluster;

[0010] Flexible scheduling unit, the flexible scheduling unit is used for the remaining time of actual decommissioning Within the pre-set production lead time before arrival, the customized module production parameter matrix is ​​analyzed, the cell sorting rules and module assembly process formula are automatically adjusted, and production is scheduled to manufacture the battery clusters within the remaining time before their actual retirement. Replacement modules with performance parameters that match the current timeline;

[0011] A closed-loop feedback unit is used to address the remaining time of the replacement module's actual decommissioning. After the target degraded module is replaced and put into operation, an updated cluster-level ledger is generated and sent to the battery management system to which the target degraded module belongs to perform adaptive parameter correction.

[0012] As a further improvement to this technical solution, the feature acquisition unit includes a working condition response capture module, a physical thermal field sensing module, and a time-series feature extraction module.

[0013] The operating condition response capture module is used to capture the charging and discharging response data of the module when it is subjected to a step current of 1C or higher under the condition of sudden power change on the generation side. The charging and discharging response data specifically includes the voltage transient drop depth of the energy storage module when it is subjected to a step current under the condition of sudden power change on the generation side. and polarization internal resistance recovery time constant ;

[0014] The physical thermal field sensing module is used to acquire the thermal field distribution characteristics of the battery cell tab region and the central region inside the energy storage module. The thermal field distribution characteristics specifically include the real-time temperature difference gradient. ;

[0015] The timing feature extraction module is used to extract the voltage transient drop depth within a preset time window. With polarization internal resistance recovery time constant The first-order rate of change over time determines the dynamic growth slope of the polarization internal resistance; the real-time temperature difference gradient within the preset time window is used. Time integration over time is performed to obtain the cumulative thermal stress integral value. The dynamic growth slope of the polarization internal resistance and the cumulative thermal stress integral value are then mapped to the same time-domain coordinate system for multi-dimensional feature fusion to obtain time-series feature parameters for characterizing the micro-deterioration of the module. .

[0016] As a further improvement to this technical solution, the preset health evolution model is a multi-head self-attention neural network that incorporates irreversible physical decay boundary constraints;

[0017] The evolution prediction unit includes a mechanism feature fusion module, a deep time series extrapolation module, and an extreme condition abrupt change correction module;

[0018] The mechanism feature fusion module extracts the frequency domain features of the historical dispatch power spectrum on the generation side through the historical output curve of the energy storage converter, and combines the frequency domain features with the time-series feature parameters. Perform orthogonal concatenation mapping to construct a multidimensional joint input feature tensor;

[0019] The deep temporal extrapolation module is used to input the multi-dimensional joint input feature tensor into the self-attention temporal encoding layer of the multi-head self-attention neural network to capture the temporal feature parameters under normal operating conditions. The long- and short-term nonlinear decay characteristics that evolve over time, and the parameter drift of the multi-head self-attention neural network output constrained by the irreversible physical decay boundary, yield a preliminary remaining time. And the mean impedance trajectory curve of the target battery cluster over future time. Remaining capacity trajectory curve ;

[0020] When the extreme operating condition sudden change correction module detects a continuous high-rate step peak-shaving command triggered on the generation side, it calculates the transient electrothermal coupling stress factor based on the extracted high-current impact duration and the transient temperature rise rate of the battery cell, and then applies the transient electrothermal coupling stress factor to the preliminary remaining time. Perform inverse nonlinear reduction correction to obtain the actual remaining decommissioning time. and in the mean impedance trajectory curve With the remaining capacity trajectory curve The above refers to the actual remaining time of retirement. Synchronous addressing is performed using time index coordinates, and the values ​​at the corresponding times are extracted as the expected mean impedance. With expected remaining capacity .

[0021] As a further improvement to this technical solution, the parameter drift of the multi-head self-attention neural network output is limited based on the irreversible physical decay boundary constraint. The specific steps involved are as follows:

[0022] The multi-dimensional joint input feature tensor is input into the multi-head self-attention neural network. The attention weight matrix of the frequency domain features, the dynamic growth slope of polarization internal resistance, and the cumulative thermal stress integral value on the historical time series is calculated by multiple parallel attention heads to extract multi-scale time-series correlation features.

[0023] In the forward inference of the multi-head self-attention neural network, the current prediction time step is extracted. With the next prediction time step Network pre-output node data;

[0024] The irreversible physical decay boundary constraint is executed. The irreversible physical decay boundary constraint is quantized into a physical monotonicity partial derivative condition in the neural network. The physical monotonicity partial derivative condition is set as follows: the first-order partial derivative with respect to the time step is calculated for the network pre-output node data. It is required that the first-order rate of change of the expected mean impedance with time must remain non-negative, and the first-order rate of change of the expected remaining capacity with time must remain non-positive.

[0025] The first-order rate of change of the network pre-output node data is substituted into the physical monotonicity partial derivative condition for verification. If the network pre-output node data is found to violate the physical monotonicity partial derivative condition, an exponential penalty gradient is applied to the model loss function to correct the weights, thereby limiting the parameter drift of the network during long-period extrapolation, until the network reaches steady-state convergence that meets the irreversible physical decay boundary constraint.

[0026] The multidimensional joint input feature tensor is fed into the multi-head self-attention neural network that has completed steady-state convergence. Through forward computation, the preliminary remaining time is output. And the mean impedance trajectory curve of the target battery cluster over future time. Remaining capacity trajectory curve .

[0027] As a further improvement to this technical solution, the preliminary remaining time... Perform inverse nonlinear reduction correction to obtain the actual remaining decommissioning time. The specific steps involved are as follows:

[0028] The system monitors the charging and discharging current of the target battery cluster in real time. If the absolute value of the current continuously exceeds the preset safety rate threshold, a high-precision timer is activated to accumulate and calculate the duration of the large current surge. ;

[0029] The real-time temperature data collected by the feature acquisition unit is acquired synchronously, and the current transient temperature rise rate of the battery cell is calculated through discrete first-order difference operation. ;

[0030] The duration of the high current impact With the transient temperature rise rate of the battery cell As a joint index key, it is retrieved into the pre-calibrated and stored electrothermal coupling acute injury mapping matrix, and spatially smoothed by bilinear interpolation. The output is a dimensionless parameter greater than or equal to 1, which serves as the transient electrothermal coupling stress factor. ;

[0031] Read the preliminary remaining time output by the depth time extrapolation module Perform nonlinear reduction correction with the initial remaining time. With transient electrothermal coupling stress factor The remaining time of retirement is considered as the actual retirement time. ;

[0032] The actual remaining retirement time was calculated. Then, in the mean impedance trajectory curve Remaining capacity trajectory curve In the context of the actual remaining time before retirement, Addressing and truncation are performed at time points to extract the transient values ​​at the corresponding moments as the final expected mean impedance. With expected remaining capacity ;

[0033] and the actual remaining time of retirement Expected mean impedance With expected remaining capacity The output is sent to the impedance matching decision unit.

[0034] As a further improvement to this technical solution, the impedance matching decision unit includes a reference impedance alignment module and an in-transit aging reverse compensation module.

[0035] The reference impedance alignment module is used to read the expected mean impedance. With expected remaining capacity For the replacement modules to be produced, based on their reference internal resistance requirements and expected average impedance, and by introducing a preset passive balancing tolerance index, an initial impedance matching constraint equation is constructed. The initial impedance matching constraint equation is used to limit the allowable impedance deviation boundary when the replacement module is connected to the grid, in order to avoid parallel circulating current and series overcharging between the old and new modules.

[0036] The in-transit aging reverse compensation module is used to read the actual remaining time before decommissioning. Furthermore, considering the geographical span between the manufacturing base and the power generation-side energy storage station, the transportation logistics cycle constant and the ambient temperature sequence along the route are extracted. The attenuation matrix of the replacement module during the logistics transportation and warehousing static storage stages due to calendar aging is calculated. And the attenuation matrix during the journey at room temperature The initial impedance matching constraint equation is introduced as a negative compensation variable to generate a customized module production parameter matrix containing the target value of the physical factory internal resistance, which is then output to the flexible scheduling unit.

[0037] As a further improvement to this technical solution, the in-transit aging reverse compensation module generates the customized module production parameter matrix, involving the following steps:

[0038] The reference impedance alignment module, for multiple discrete state of charge points, constructs the initial impedance matching constraint equation based on the reference internal resistance requirement of the replacement module under ideal grid-connected conditions and the expected average impedance of the target battery cluster, and introduces a preset passive equalization current allowable deviation coefficient as a tolerance constraint index. The initial impedance matching constraint equation requires that the absolute impedance difference between the reference internal resistance requirement and the expected average impedance must not exceed the dynamic allowable deviation boundary determined by multiplying the expected average impedance by the passive equalization current allowable deviation coefficient.

[0039] The in-transit aging reverse compensation module is based on the remaining time of the actual decommissioning. Working backwards from the production and delivery dates, we extract the estimated logistics and transportation time from the logistics planning. And discrete environmental temperature sequences along the route predicted based on geographic meteorological data. ;

[0040] The discrete Arrhenius aging time-varying integral model is used to estimate the logistics transportation time. With the discrete ambient temperature sequence along the route As a boundary condition input, the passivation increment of the solid-phase interface at each discrete charge state point is integrated and accumulated to obtain the attenuation matrix at room temperature during the journey. ;

[0041] The attenuation matrix during the journey at room temperature As a negative compensation variable, the initial impedance matching constraint equation is introduced to perform reverse correction of the factory calibration, resulting in a customized module production parameter matrix containing the target value of the physical factory internal resistance.

[0042] The factory calibration logic of the customized module production parameter matrix requires that the absolute impedance difference between the target value of the physical factory internal resistance and the equivalent superposition value of the attenuation matrix at room temperature during transit, and the expected average impedance, must not exceed the dynamic allowable deviation boundary determined by multiplying the expected average impedance by the allowable deviation coefficient of the passive equalization current. This is used to guide the downstream flexible scheduling unit to pre-deduct logistics aging redundancy before leaving the factory.

[0043] As a further improvement to this technical solution, the flexible scheduling unit includes a material dynamic screening module and a process linkage control module;

[0044] The material dynamic screening module is used in the actual remaining decommissioning time. Within the preset production lead time before arrival, the customized module production parameter matrix is ​​analyzed, the physical factory internal resistance target value in the customized module production parameter matrix is ​​extracted, and the front-end cell sorting equipment is controlled to screen out cell batches whose AC internal resistance characteristics are close to the physical factory internal resistance target value.

[0045] The process linkage control module is used to automatically adjust the bus laser welding parameters in the module assembly process when the internal resistance of a single cell cannot fully match the target value of the physical internal resistance at the factory. By adjusting the laser welding penetration depth or effective welding area, a controllable contact thermal resistance is introduced to compensate for the residual between the cell body impedance and the target value of the physical internal resistance at the factory, thereby manufacturing a replacement module that meets the factory calibration.

[0046] As a further improvement to this technical solution, the process linkage control module introduces controllable contact thermal resistance by adjusting the laser welding penetration depth or effective welding area to manufacture a replacement module that meets the factory specifications. The specific steps involved are as follows:

[0047] After the material dynamic screening module completes the first round of grouping, obtain the average actual AC internal resistance of the batch of battery cells to be welded at the current workstation of the production line. ;

[0048] Extract the target value of physical internal resistance from the customized module production parameter matrix. And calculate the target value of the physical internal resistance at the factory. With the average actual AC internal resistance Impedance compensation residuals between ;

[0049] If impedance compensation residual If the value is greater than 0, then the impedance compensation residual will be... As input features, they are input into a pre-calibrated and stored process mapping model. The process mapping model contains a nonlinear mapping relationship between contact thermal resistance and laser welding parameters. The corresponding variable parameter welding command is obtained by solving and matching. Specifically, the variable parameter welding command is a target welding laser power reduction command or a target effective weld point reduction command.

[0050] The variable-parameter welding instructions are compiled into low-level executable machine-readable code and sent to the laser welding equipment on the module assembly line via the industrial control bus to perform adaptive variable-parameter welding. This accurately generates an equivalent impedance compensation residual at the interface between the cell terminal and the bus. Customized contact thermal resistance;

[0051] The average actual AC internal resistance of the batch of cells to be welded is obtained by... The customized contact thermal resistance is physically superimposed on the structure, thereby creating an overall internal resistance that is absolutely close to the target value of the physical factory internal resistance. And it meets the factory-specified replacement module requirements.

[0052] As a further improvement to this technical solution, the closed-loop feedback unit is used to monitor the remaining time of actual decommissioning after the replacement module has undergone logistics transportation and room temperature settling. After the target degraded module is replaced and put into operation at the node, the final grid-connected identity identifier of the replaced module is extracted;

[0053] and the expected mean impedance With the expected remaining capacity Assign the final grid-connected identity identifier and generate an updated cluster-level ledger;

[0054] The updated cluster-level ledger is then sent to the field battery management system to which the target degraded module belongs, driving the field battery management system to remove the derating and current-limiting charge / discharge protection mechanism set based on the original degradation state of the target degraded module, thus completing the physical closed loop from production capacity to operation and maintenance.

[0055] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0056] 1. In a power generation-side energy storage module health status monitoring and production scheduling management platform, to address the logistical cycle difference of up to tens of days between factory production scheduling and on-site grid connection, a multi-head self-attention neural network integrating irreversible physical decay boundary constraints is constructed, combined with an extreme operating condition abrupt change correction module (based on transient electrothermal coupling stress factor). ), pinpointing the actual remaining time before the target degraded module is deactivated. and the expected mean impedance within the cluster at that moment. Furthermore, a decay matrix is ​​introduced during the journey when the vehicle is stationary at room temperature. It is used as a negative compensation variable for reverse correction during factory calibration, ensuring that the physical impedance of the replacement module can be perfectly assimilated with the continuously aging old battery clusters in the field at the moment of grid connection after long-distance logistics, thus eliminating the risk of parallel circulating current and series overcharging within the cluster from the source.

[0057] 2. In a power generation-side energy storage module health status monitoring and production scheduling management platform, after the replacement module is connected to the grid, its final grid connection identity is extracted, and the expected average impedance is... With expected remaining capacity The system is forcibly deployed to the field battery management system (BMS) of the target degraded module, generating an updated cluster-level ledger. This drives the underlying system to remove the derating and current-limiting charge / discharge protection mechanisms forced upon it due to the original degraded module's degradation state, restoring current sharing and synchronization of the individual cell voltage distribution within the cluster. This completely eliminates the risk of impedance mismatch and ensures the target battery cluster's performance. It can quickly return to its full-power rated operating range at any time. Attached Figure Description

[0058] Figure 1 This is a flowchart illustrating the overall process of the present invention.

[0059] The meanings of the labels in the diagram are as follows:

[0060] 1. Feature acquisition unit; 11. Operating condition response capture module; 12. Physical thermal field sensing module; 13. Temporal feature extraction module;

[0061] 2. Evolutionary prediction unit; 21. Mechanism feature fusion module; 22. Deep time series extrapolation module; 23. Extreme condition abrupt change correction module;

[0062] 3. Impedance matching decision unit; 31. Reference impedance alignment module; 32. In-transit aging reverse compensation module;

[0063] 4. Flexible scheduling unit; 41. Dynamic material screening module; 42. Process linkage control module;

[0064] 5. Closed-loop feedback unit. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] For examples, please refer to Figure 1 As shown, a platform for monitoring the health status and managing the production scheduling of energy storage modules on the power generation side is provided. This platform includes a feature acquisition unit 1, which continuously acquires the charge and discharge response data and thermal field distribution characteristics of each energy storage module in the power generation side energy storage battery cluster under fluctuating operating conditions, and extracts time-series feature parameters to characterize the micro-deterioration of the modules. ;

[0067] In this embodiment, when the present invention is applied to a new energy power generation-side energy storage power station, its energy storage system needs to frequently respond to random fluctuations in wind power and photovoltaic output, as well as high-frequency step scheduling of primary frequency regulation, resulting in the battery modules being in a non-steady-state charge-discharge cycle for a long time. The feature acquisition unit, through the coordinated operation of its subordinate operating condition response capture module 11, physical thermal field sensing module 12, and time-series feature extraction module 13, achieves parameterized and accurate deconstruction of the microscopic aging state of each energy storage module.

[0068] Furthermore, the operating condition response capture module 11 is electrically connected to the battery management system (BMS) and high-frequency electrical parameter sampling circuit inside the on-site container or battery compartment. When a sudden change in wind and solar power output triggers a grid dispatch command switch, causing the energy storage module to withstand a large step current of 1C or higher, the operating condition response capture module captures the transient electrochemical response characteristics at the moment of the sudden change in operating condition in real time, specifically including: voltage transient drop depth. With polarization internal resistance recovery time constant .

[0069] Among them, voltage transient drop depth The characteristic voltage drop at the moment the charging / discharging current jumps, after the energy storage module's terminal voltage escapes ohmic polarization, is caused by the combined effects of electrochemical and concentration polarization; polarization internal resistance recovery time constant. This refers to the characteristic relaxation time required for the polarized capacitor to discharge during the process of the module terminal voltage recovering to the open circuit voltage in the phase of step current removal or constant current switching.

[0070] The physical thermal field sensing module 12 is communicatively connected to multiple high-precision temperature sensors distributed within the energy storage module (to meet the requirements for real-time temperature gradient under high-frequency fluctuation conditions on the power generation side). To meet the requirement of precise transient temperature capture, multiple high-precision temperature sensors in this embodiment are preferably any one or a combination of a negative temperature coefficient (NTC) thermistor, a Pt100 platinum resistance thermometer, or an integrated digital temperature sensor chip with an absolute measurement error of less than or equal to 0.1℃. These high-precision temperature sensors are fixed inside the energy storage module in a distributed array. Specifically, a representative cell within each energy storage module has the aforementioned high-precision temperature sensors attached to its tab region (as the point of highest temperature rise) and its central region (as the point of normal temperature rise). The physical thermal field sensing module 12 calculates the real-time temperature gradient within each energy storage module by synchronously reading the real-time temperature difference between the tab region and the central region at the same time point. Specifically, the real-time temperature data at the underlying physical parameter level includes: the transient absolute temperature value of the electrode tab collected by a high-precision temperature sensor fixed to the electrode tab area of ​​the battery cell. And the central transient absolute temperature value collected by a high-precision temperature sensor fixed in the center region of the battery cell. The physical thermal field sensing module 12 will collect the transient absolute temperature value of the electrode. With the aforementioned central transient absolute temperature value Reconstruction generates a sequence of absolute physical temperatures of the battery cells arranged according to the sampling time series. .

[0071] Furthermore, the physical thermal field sensing module 12 synchronously reads the difference in the real-time temperature data (i.e., the transient absolute temperature value of the electrode tab) on the same time segment. With the aforementioned central transient absolute temperature value The algebraic difference between them is used to calculate the real-time temperature gradient inside each energy storage module. .

[0072] The real-time temperature gradient The magnitude of the value directly reflects the degree of thermodynamic imbalance caused by structural thermal resistance and uneven heat dissipation flow field within the module, and is also the fundamental environmental factor that induces the accelerated degradation of non-uniformity within the cluster modules.

[0073] The time-series feature extraction module 13 pre-sets a time sliding window as a preset time window. In this embodiment, the preset time window is preferably the past 30 consecutive natural days. First, the voltage transient drop depth continuously recorded within the preset time window is retrieved. With polarization internal resistance recovery time constant The voltage transient sag depth is determined by using time series analysis with discrete first-order differential operators or least-squares linear regression. With polarization internal resistance recovery time constant The time derivatives of the two variables are used to obtain their first-order rates of change. The first-order rates of change of the two variables are normalized and weighted to obtain the dynamic growth slope of the polarization resistance. This dynamic growth slope of the polarization resistance quantifies the micro-nonlinear rate at which the polarization resistance deteriorates with the accumulation of operating conditions through a determined physical mechanism.

[0074] Simultaneously, the real-time temperature difference gradient continuously collected within the preset time window is retrieved. The discrete integration operation is performed over time to calculate the cumulative thermal stress integral value. This integral value represents the total amount of uneven thermal load that the cell has been subjected to over a long period of time, which quantifies how much "thermal damage" the battery has suffered over a period of time.

[0075] Finally, the time-series feature extraction module 13 eliminates the dimensional differences between the dynamic growth slope of the polarization internal resistance and the cumulative thermal stress integral value through a data normalization operator, and performs orthogonal fusion of multi-dimensional features under the same time domain section to reconstruct a two-dimensional degradation feature vector. (in, This represents the slope of the dynamic increase in polarization internal resistance. This represents the cumulative thermal stress integral value. This represents the temporal feature parameters generated by the reconstruction. (representing the transpose operator), and outputs it as a temporal feature parameter to characterize the micro-degradation of the module. Subsequently, it is sent as a deterministic boundary condition input to the evolution prediction unit 2.

[0076] The power generation-side energy storage module health status monitoring and production scheduling management platform also includes an evolution prediction unit 2. The evolution prediction unit 2 is used to input time-series characteristic parameters into a preset health evolution model to predict the preliminary remaining time when the target deteriorated module reaches the retirement threshold. And combining the charge / discharge response data with the thermal field distribution characteristics, the preliminary remaining time is determined. Perform inverse nonlinear reduction correction to obtain the actual remaining decommissioning time. Simultaneously, the remaining time before actual retirement is predicted. At any given time, the expected average impedance of the remaining healthy modules within the battery cluster containing the target degraded module. With expected remaining capacity ;

[0077] In this embodiment, the underlying network architecture of the evolution prediction unit 2 is a multi-head self-attention neural network built on the framework of artificial intelligence and deep learning computing, which integrates the boundary constraints of irreversible physical decay. Internally, through the cascaded collaboration of the mechanism feature fusion module 21, the deep temporal extrapolation module 22, and the extreme working condition change correction module 23, it overcomes the engineering defects of conventional artificial intelligence models (such as pure data-driven RNN or LSTM machine deep learning algorithms) in long-term extrapolation and prediction failure when dealing with sudden changes in working conditions.

[0078] In the front-end processing stage of the data stream, the mechanism feature fusion module 21 retrieves historical power output curve data of the power generation-side energy storage converter (PCS) via the communication bus, performs a Fast Fourier Transform (FFT), and extracts the frequency domain features of the power spectrum of the energy storage power station when executing peak shaving, frequency regulation, and other scheduling commands. Subsequently, in order to enable the neural network to simultaneously capture external macroscopic loads and internal fatigue stresses at the input layer, the mechanism feature fusion module 21 obtains time-series feature parameters. and combine frequency domain features with time-series feature parameters Perform orthogonalization concatenation mapping within the tensor space to construct a multidimensional joint input feature tensor. :

[0079] ;

[0080] In the formula, This represents the frequency domain characteristics of the dispatch power spectrum analyzed by the Fast Fourier Transform, used to quantitatively characterize the energy density of high-frequency random power fluctuations in the external power grid. This represents the multidimensional joint input feature tensor, which serves as the standard high-dimensional input boundary condition for subsequent multi-head self-attention neural networks in terms of data structure.

[0081] In obtaining the multidimensional joint input feature tensor Subsequently, the deep time series extrapolation module 22 performs long-range prediction under normal operating conditions; and in order to limit parameter drift during network extrapolation and prevent the output of physical paradoxes that violate objective common sense (such as abnormal decrease in internal resistance or abnormal rebound in capacity), the deep time series extrapolation module 22 performs the following steps:

[0082] The multidimensional joint input feature tensor The data is fed into the self-attention temporal encoding layer of the multi-head self-attention neural network. The network uses multiple parallel attention heads to dynamically calculate the attention weight matrix of frequency domain features, dynamic growth slope of polarization internal resistance, and cumulative thermal stress integral value at different historical time spans, thereby accurately extracting multi-scale temporal correlation features that lead to inconsistent aging.

[0083] In the forward inference loop of the network, the module extracts the current prediction time step. With the next prediction time step Network pre-output node data;

[0084] The irreversible physical decay boundary constraints are clearly defined and enforced. In the underlying computational logic, these constraints are quantified as "physical monotonicity partial derivative conditions." Specifically, for the network's pre-output node data, the first-order rate of change over time is calculated, requiring that the expected mean impedance's first-order rate of change over time must always be non-negative. Furthermore, the expected rate of change of the remaining capacity over time must always be non-positive. ),in, The expected mean impedance (refers to the average AC internal resistance of the remaining healthy modules in the target battery cluster that have not reached the retirement safety threshold (i.e., retirement threshold) when the multi-head self-attention neural network extrapolates to a certain prediction section in the future, used to quantify the sum of charge transfer resistance caused by the intensification of polarization effect inside the module and the thickening of the solid electrolyte interface (SEI) film). This represents the prediction time step or time variable (representing the future time sequence in which the multi-head self-attention neural network is performing long-term evolution prediction). Represents the partial derivative operator (partial derivative operator) With independent variable Combined to form time partial derivative operators In mathematics and underlying computational logic, this operator is used to solve for the expected mean impedance. or the expected remaining capacity The dynamic first-order rate of change over time (i.e., the evolutionary gradient) is used as a criterion for verifying the direction of the physical monotonicity of irreversible aging by its positive or negative value (non-negative or non-positive). The expected remaining capacity (refers to the benchmark value of the ampere-hour (Ah) capacity of the above-mentioned healthy module, which can be actually used for charging and discharging scheduling on the power generation side, on the same future prediction section, in order to quantify the remaining capacity of the active lithium ion stock and its reversible electrochemical reaction inside the cell).

[0085] The first-order rate of change is substituted into the physical monotonicity partial derivative condition for verification; wherein, in this embodiment, a condition including mean square error is constructed. The total loss function with exponential penalty terms :

[0086] ;

[0087] ;

[0088] In the formula, This represents the total loss function of the preset health evolution model, which is used as the optimization objective function of the network backpropagation algorithm; This represents a preset dynamic adaptive adjustment coefficient used to balance the weight strength of data fitting accuracy and physical boundary constraints during the optimization iteration process. Specifically, it ensures that when the network touches the "physical paradox" (such as a decrease in internal resistance), the resulting monotonic penalty gradient on the backpropagation path is at least two orders of magnitude stronger than the regular base error gradient. The absolute dominance of ) forces the elimination of any weight updates that deviate from the physical trajectory, therefore, The range of values ​​is preferably limited to Within the high penalty range, and based on the loss of the baseline prediction error during network training. The gradient descent convergence state dynamically and adaptively increases; This represents the exponentially monotonic penalty loss term; This represents the preset penalty sensitivity amplification factor, a hyperparameter used to amplify the absolute value of the penalty gradient when the network touches the physical red line. Specifically, in long-period extrapolation, parameter drift often starts from tiny perturbations. When the value is within this range, any tiny anomaly in the first-order rate of change that violates the laws of physics (e.g.) And when the absolute value is only 0.05), through The product amplification can instantly break through the noise threshold of the exponential function, producing a steep exponential burst, giving the network extremely high sensitivity and zero tolerance to physical boundaries. Based on this, The range is preferably set in between; It represents the time partial derivative (first-order rate of change) of the expected mean impedance in the pre-output node data of the network as a function of time. It represents the time partial derivative (first-order rate of change) of the expected remaining capacity in the pre-output node data of the network as a function of time. This represents a nonlinear activation operator used to perform one-way interception, meaning that no penalty is triggered when the first-order rate of change conforms to physical laws, but subsequent operations are activated once it violates physical laws. Represented by natural constant An exponential function operator with base is used to achieve a geometric progression amplification of the penalty amount.

[0089] Furthermore, if the network output conforms to objective laws (such as...) ),but Operator blocking computation, penalty term is 0; once an anomalous prediction violating physical monotonicity is identified (such as... (physical paradox) The function is activated instantly; the outer nested natural exponential function Errors that violate physical laws are amplified geometrically. An exponential penalty gradient with a very large value is applied to the loss function. This penalty gradient, through backpropagation, forcibly pulls the attention weight matrix that caused the paradox back to a physically safe range that conforms to the thermodynamic aging law, until the multi-head self-attention neural network completes weight updates and reaches physical steady-state convergence. It is worth noting that the above process involving partial derivative calculation, loss function penalty, and steady-state convergence belongs to the model training optimization stage of this multi-head self-attention neural network (including offline pre-training and online adaptive fine-tuning during platform idle periods), which aims to solidify irreversible physical boundaries from the underlying genes of the model.

[0090] It is worth noting that the aforementioned retirement safety threshold is a physical boundary judgment benchmark pre-programmed into the system's underlying layer. Specifically, based on the conventional operating standards of the power generation-side energy storage industry, this retirement threshold is quantified in this embodiment as any one or a combination of the following physical extreme red lines: the actual usable capacity of the target deteriorated module decays to 80% of the factory rated capacity (i.e., the State of Health (SOH) scale drops to 0.8); or, the AC polarization internal resistance of the target deteriorated module dynamically increases to within the range of 150% to 200% of the initial factory calibration benchmark value. In the data flow calculation logic, this retirement safety threshold serves as the absolute cutoff boundary for the multi-head self-attention neural network to perform evolutionary extrapolation on the time axis. When the impedance or capacity decay trajectory output by the network (i.e., the mean impedance trajectory curve and the remaining capacity trajectory curve) touches the aforementioned physical extreme red line, the algorithm engine immediately stops extrapolating and uses this to determine the preliminary remaining time. The time span ends at the anchor point.

[0091] When the platform is in the real-time operational simulation phase at the industrial site, the deep temporal extrapolation module 22 does not need to re-execute the aforementioned cumbersome backpropagation convergence calculation. Instead, the deep temporal extrapolation module 22 directly feeds the multi-dimensional joint input feature tensor captured in real time into the multi-head self-attention neural network that has already completed steady-state convergence. Through a single extremely fast unidirectional network forward calculation, it can instantly output a preliminary remaining time with absolute engineering reliability. And the mean impedance trajectory curve characterizing the future evolution trend of the target battery cluster over time. Remaining capacity trajectory curve Since the network has internalized the physical boundary constraints during training, the trajectory curves output by its forward extrapolation are naturally immune to parameter drift. At this point, since the actual decommissioning node after encountering extreme mutation conditions has not yet been determined, the deep time series extrapolation module 22 does not output the final static expected impedance and capacity scalars, but instead caches the trajectory curves containing full time axis information locally for downstream modules to call and extract.

[0092] Furthermore, when power generation faces severe weather or regional grid failures, energy storage systems often enter extreme operating conditions requiring large-rate step-shaving, leading to a precipitous drop in their lifespan; at this point, the initial remaining time... This will result in a dangerous lag; the extreme condition abrupt change correction module 23 will perform emergency takeover through the following steps:

[0093] The extreme operating condition sudden change correction module 23 monitors the charging and discharging current of the target battery cluster in real time (the underlying hardware interface listens to the high-frequency sampling bus of the battery management system). When it detects that the absolute value of the charging and discharging current continuously exceeds the preset safety rate threshold, the module triggers an internal high-precision timer to accumulate and calculate the determined duration of the large current impact. Simultaneously, the internal microprocessor is invoked to process the real-time temperature data (i.e., the cell's physical absolute temperature sequence) acquired by the feature acquisition unit (specifically through the physical thermal field sensing module 12). Perform discrete first-order difference operations:

[0094] ;

[0095] In the formula, The output of the solution represents the transient temperature rise rate of the battery cell, which characterizes the transient burst speed at which the Joule heat and chemical reaction heat inside the battery cell cannot dissipate to the outside in time under extreme high-rate discharge, resulting in a rapid accumulation of heat in the battery cell tabs and central region. Indicates the current high-frequency sampling time The latest physical absolute temperature value of the battery cell corresponding to the real-time temperature data; This indicates the time step of the high-frequency sampling period preset by the underlying field communication bus and battery management system (BMS) (preferably in the range of 100 milliseconds to 1 second in this embodiment). This step determines the time resolution of the platform in capturing the precursors of thermal runaway or acute thermal damage.

[0096] Real-time acquisition of the current transient temperature rise rate of the battery cell (in, This represents an extremely small change in temperature. (representing an extremely small time step), characterizing the acute tearing intensity of the battery structure under abrupt operating conditions;

[0097] The duration of the high current impact With the transient temperature rise rate of the battery cell Encapsulated as a joint index key, an address is initiated to the local non-volatile memory, and the electrothermal coupling acute injury mapping matrix, pre-constructed based on destructive calibration experiments, is loaded. Specifically, in order to enable the extreme condition mutation correction module to efficiently address and execute bilinear interpolation in the underlying computing architecture, the electrothermal coupling acute injury mapping matrix... Specifically, it is constructed and stored as a A two-dimensional discrete real number array;

[0098] Specifically, define a finite set of discrete independent variables. This represents the division through calibration experiments. The duration of a large current surge is a node, and it satisfies the following conditions: ;

[0099] Define a finite set of discrete independent variables This represents the division through calibration experiments. Each cell transient temperature rise rate node, and satisfies ;

[0100] The electrothermal coupling acute injury mapping matrix constitutes a set of discrete mathematical functions that map from a two-dimensional input domain to a one-dimensional physical damage domain. Any element in the matrix The mathematical definition of is:

[0101] ;in, ;

[0102] In the formula, This represents the nonlinear damage function established by the aforementioned physical destructive calibration experiment; element Representation matrix In the Column and number The scalar element stored at the row intersection position represents the element within a defined duration. With a determined rate of temperature rise The reference value of the dimensionless transient electrothermal coupling stress factor measured under combined cross-operation conditions ( ); The set of discretized values ​​representing the duration of the high-current impact constitutes the horizontal axis (X-axis) of the mapping matrix. Represents a set The first in The duration of a large current surge is used to quantify the specific time length for which the energy storage module can withstand high-rate charging and discharging (such as discrete calibration points like 30 seconds and 60 seconds). Represents a set The total number of elements, i.e., the dimension of the matrix's grid in the X-axis direction (positive integer). The set of discretized values ​​representing the transient temperature rise rate of the battery cell constitutes the vertical axis (Y-axis) of the mapping matrix. Represents a set The first in Individual cell transient temperature rise rate node values ​​are used to quantify the transient slope of temperature rise caused by heat accumulation inside the cell (such as discrete calibration points such as 0.2℃ / s, 0.5℃ / s, etc.). Represents a set The total number of elements, i.e., the dimension of the matrix's grid in the Y-axis direction (positive integer). This represents the mapping operator, indicating that the matrix completes the transformation from a cross-input domain of two physical quantities (time and temperature gradients) to a real domain of a single physical quantity (stress-damage floating-point number). Dimensionality reduction mapping.

[0103] Between the discrete grid points of the electrothermal coupling acute injury mapping matrix, a bilinear interpolation operator is invoked for spatial smoothing solution. The final output is a dimensionless floating-point number greater than or equal to 1.0, which serves as the transient electrothermal coupling stress factor quantifying the extent of acute injury propagation. ;

[0104] Since the feature parameters captured in real time are usually continuous floating-point numbers, they cannot directly hit discrete integer grid points in the matrix; therefore, the transient electrothermal coupling stress factor is calculated by performing bilinear interpolation through the built-in algorithm engine. :

[0105] The duration of the high-current impact currently being captured in real time is set to... (Right now The current transient temperature rise rate of the battery cell is (Right now );

[0106] The addressing logic of the extreme condition sudden change correction module 23 is in the set of independent variables of the matrix. and In this process, a binary search algorithm is used to quickly locate the surrounding real coordinates. The four nearest neighbor matrix points are used to define the grid points:

[0107] The adjacent lower bound nodes of the X-axis (time dimension) With upper bound node ,satisfy ;

[0108] The adjacent lower bound nodes of the Y-axis (temperature rise dimension) With upper bound node ,satisfy ;

[0109] Through the aforementioned boundary locking, the module retrieves data from the static constant table (i.e., the electrothermal coupling acute injury mapping matrix). Simultaneously extract the corresponding four physical calibration reference values:

[0110] bottom left node value The value of the bottom right node Top left node value and the value of the top right node .

[0111] First, at a fixed lower bound of the temperature rise rate. With the upper realm On two horizontal baselines, respectively based on the actual duration Using the distance ratio, perform first-order algebraic linear interpolation to calculate the stress values ​​for the two intermediate transition states:

[0112] ;

[0113] ;

[0114] After obtaining the stress values ​​for the two intermediate transition states mentioned above, then in the vertical direction, based on the actual temperature rise rate... The distance ratio is used to perform a final algebraic first-order linear interpolation mapping to obtain the final transient electrothermal coupling stress factor. :

[0115] ;

[0116] In the formula, This represents the measured duration of the continuous high-current impact captured in real time on-site, serving as the input reference parameter for the interpolation algorithm in the X-axis direction. This represents the measured value of the continuous transient temperature rise rate of the battery cell captured in real time on site, which serves as the input reference parameter for the interpolation algorithm in the Y-axis direction. and Represents from matrix set The address match found in the middle is exactly the envelope. The two adjacent discrete calibration time nodes; and Represents from matrix set The address match found in the middle is exactly the envelope. The two adjacent discrete calibration temperature rise rate nodes; , , and These represent the electrothermal coupling acute injury mapping matrices respectively. The dimensionless stress damage reference values ​​are pre-stored at these four adjacent intersecting grid points; This indicates that after the first round of X-axis linear interpolation, the calculated temperature rise rate is... The theoretical intermediate transition state stress damage value at time; This indicates that after the first round of X-axis linear interpolation, the calculated temperature rise rate is... The theoretical intermediate transition state stress damage value at time; This represents the transient electrothermal coupling stress factor that is finally output after a smooth dimensionality reduction solution in two-dimensional space.

[0117] Furthermore, the preliminary remaining time output by the depth time extrapolation module 22 is read. Perform nonlinear reduction correction with the initial remaining time. With transient electrothermal coupling stress factor The remaining time of retirement is considered as the actual retirement time. :

[0118] ;

[0119] In the formula, This indicates the actual remaining time before the target degraded module is decommissioned, after corrections based on the underlying physical mechanisms.

[0120] Specifically, due to the existence of transient thermal coupling damage (i.e., extreme step current impacts can trigger acute thermal damage and irreversible consumption of active lithium inside the cell, leading to the transient electrothermal coupling stress factor), (A nonlinear penalty constant that is always greater than 1) penalizes the optimistic remaining lifespan originally estimated based on stable operating conditions, thereby accurately quantifying the actual lifespan of the target module after being subjected to this extreme operating condition through nonlinear compression and reduction; after calculating this absolute retirement time anchor point, the extreme operating condition abrupt change correction module 23 calls the aforementioned cached mean impedance trajectory curve. Remaining capacity trajectory curve (Specifically, the trajectory curve containing full time axis information is instantiated as: a discrete time-series mapping data table with future time step sequences as index keys and corresponding impedance and capacity prediction values ​​as mapping values), and the actual remaining decommissioning time is... Used as a time index coordinate for synchronous addressing (requiring the actual remaining time of retirement). As the target matching key value, it is substituted into the aforementioned discrete time-series mapping data table to perform table lookup matching and numerical truncation, thereby accurately "cutting out" the new module on the dynamic curve. The precise physical state exhibited by the on-site aging battery clusters at the moment of grid connection is used to extract the final static expected average impedance. With expected remaining capacity ;

[0121] The actual remaining time of retirement after calculation As the updated absolute retirement time point, and the expected mean impedance With the expected remaining capacity The data is directly transmitted to the downstream impedance matching decision unit 3 and flexible production scheduling unit 4. Through the forced downward correction of this parameter, it can trigger order insertion and material preparation instructions to the manufacturing end with extreme agility when encountering extreme working conditions, and completely open up the causal closed loop between on-site physical changes and flexible production capacity scheduling. It makes up for the lag blind spot of the pure time-series extrapolation model when facing sudden unknown working conditions, and accurately transforms the unpredictable physical damage depreciation on-site into the rigid deadline node for the downstream flexible production scheduling unit 4 and the in-transit aging reverse compensation module 32 to back-calculate the production delivery date and logistics cycle, thereby fundamentally eliminating the supply gap of replacement modules and unplanned shutdown accidents of energy storage power stations caused by prediction failure.

[0122] The power generation-side energy storage module health status monitoring and production scheduling management platform also includes an impedance matching decision unit 3, which is used to match the expected average impedance. and expected remaining capacity Based on, and in conjunction with the actual remaining time of retirement. Construct a customized module production parameter matrix to avoid impedance mismatch between new and old modules within the cluster;

[0123] In this embodiment, the expected mean impedance is obtained by the impedance matching decision unit 3. Expected remaining capacity and the actual remaining time of decommissioning after eliminating errors caused by sudden changes in operating conditions. And translate it into production scheduling boundary instructions at the physical manufacturing end:

[0124] At the data flow front end, in order to completely avoid the risks of parallel loop current and series overcharging between the old and new modules at the physical level, the reference impedance alignment module 31 reads the expected average impedance. With expected remaining capacity They are responsible for developing initial rules designed to define the grid-connected allowable impedance deviation boundaries for replacement modules;

[0125] Specifically, for multiple discrete charge state points Extract the reference internal resistance requirement of the replacement module to be produced under ideal grid-connected conditions. and the expected mean impedance of the target battery cluster. Simultaneously, a preset passive balancing current allowable deviation coefficient is introduced. (In this embodiment, based on the maximum heating power limit of the equalization resistor, it is preferably set to) As a tolerance constraint index, based on the above characteristic parameters, the initial impedance matching constraint equation is constructed as follows:

[0126] ;

[0127] In the formula, For discrete state-of-charge points, representing the percentage of remaining available charge within the module, it is used as an independent evaluation dimension for this constraint equation; This indicates that the replacement modules to be produced are in a specific... The required reference internal resistance value that must be present under ideal grid connection conditions at the node level; This indicates that the health modules within the target battery cluster are in the same... Expected mean impedance at the node; The absolute value operator is used to calculate the absolute impedance difference between the above-mentioned reference internal resistance requirement and the expected average impedance, shielding the positive and negative deviation directions and only representing the deviation magnitude. The preset allowable deviation coefficient of the passive balancing current serves as a tolerance constraint indicator. The operator is less than or equal to, used to limit the calculated absolute impedance difference to not exceed the value specified on the right side of the equation. The "dynamic allowable deviation boundary" determined by multiplication;

[0128] It is worth noting that in the above initial impedance matching constraint equation: the left side of the equation solves for the absolute impedance difference between the reference internal resistance requirement and the expected average impedance, while the right side of the equation calculates the dynamic allowable deviation boundary determined by multiplying the expected average impedance by the deviation coefficient; only when the absolute impedance difference is tightly enveloped within this dynamic allowable deviation boundary can subsequent production scheduling actions be triggered.

[0129] Furthermore, after establishing the initial tolerance boundary, the data stream enters the in-transit aging reverse compensation module 32; due to the geographical distance between the energy storage module manufacturing base and the power generation site, in-transit calendar aging will cause a positive drift of internal resistance.

[0130] Therefore, the in-transit aging reverse compensation module 32 reads the actual remaining retirement time issued in the previous step. First, perform a reverse timeline deduction: based on the actual remaining retirement time... Using the absolute grid connection cutoff node as a reference point, production and delivery dates are determined by working backwards, and the estimated logistics and transportation time is accurately extracted by calling the API of an external logistics planning system. (e.g., a total transportation and warehousing time of 15 days); simultaneously, by integrating geographic and meteorological data interfaces, estimated logistics transportation time along the route is extracted. Discrete ambient temperature series fluctuating over time within a period ;

[0131] After obtaining the logistics time and temperature boundary conditions, the discrete Arrhenius aging time-varying integral model was called for solution: Since the reaction rate of passivation increment at the solid-phase interface inside the cell is greatly affected by the temperature-induced exponential effect (such as the nonlinear decay caused by diurnal temperature difference), the time-slice discrete accumulation method was used for integration accumulation:

[0132] ;

[0133] In the formula, This means that the calculated increase in internal resistance of the replacement module due to aging of the logistics calendar at a specific state of charge point, and the set of increments of each SOC node constitute the attenuation matrix during transit at room temperature. ; This represents the apparent activation energy for the growth of the solid electrolyte interfacial film in this chemical system. Represents the ideal gas constant (approximately) ); Indicates the input time step Discrete ambient temperature (Kelvin) at time points; This indicates the time slice step size when performing discrete numerical integration (preferably set to 1 hour). This represents a pre-calibrated, SOC-related pre-exponential aging factor, used to characterize the collision frequency of fundamental chemical reactions (i.e., deposition of reaction byproducts during solid electrolyte interfacial film growth) in a specific battery material system without temperature excitation. Furthermore, for the lithium iron phosphate (LiFePO4) cell system commonly used in current power generation-side energy storage, this factor is applied at a specific time step. The value is taken in hours (h), and the impedance is taken in milliohms (h). Under the dimensional system of ), this pre-exponential factor The range of values ​​is preferably limited to Within the order of magnitude range; it is worth noting that the pre-exponential factor Specific calibration values ​​and A non-linear positive correlation mapping relationship: that is, the target module The higher the value (indicating higher lithium intercalation abundance at the negative electrode, more active electrochemical potential, and a greater probability of side reaction collisions), the higher the corresponding pre-exponential factor. The value approaches the upper limit of that interval; the above differences Discrete nodes The numerical set is pre-programmed into the non-volatile read-only memory (ROM) at the bottom layer of the module.

[0134] After obtaining the absolutely quantized attenuation matrix, the final control logic is reverse-engineered: the attenuation matrix obtained during the journey at room temperature is then used as the final control logic. As a negative compensation variable, it is forcibly introduced into the initial impedance matching constraint equation for reverse correction of factory calibration:

[0135] ;

[0136] In the formula, This represents the target value of the physical internal resistance generated by reverse deduction, which is the absolute manufacturing control command that the factory's flexible production line must directly aim at and approximate when screening and assembling battery cells. As an equivalent superposition operator, the logic here is limited to: requiring the physical factory internal resistance target value to be assigned to the production line. This must be combined with the inevitable increase in aging redundancy that will occur during the module's transit time. Perform algebraic merging;

[0137] This modified equation strictly adheres to the aforementioned factory calibration logic constraints at the control level: it requires that the final physical factory internal resistance target value be assigned to the production line. Rather than the inevitable increase during its journey. The resulting equivalent superposition value must strictly meet the aforementioned constraints of the dynamic allowable deviation boundary (i.e., it must not exceed the dynamic allowable deviation boundary determined by multiplying the expected mean impedance by the passive equalization current allowable deviation coefficient). The customized module production parameter matrix is ​​ultimately output as an instruction stream to the downstream flexible scheduling unit 4, which guides the production line to pre-deduct the material aging redundancy before assembly and delivery, so as to achieve precise impedance engagement across time and space from source to network end.

[0138] It is worth noting that, through the above algebraic reconstruction, a target value including the physical factory internal resistance was generated. The customized module production parameter matrix is ​​specifically a structured discrete real number mapping array, designed to enable seamless parsing and execution of downstream flexible scheduling units by machine vision and industrial control programs. ;

[0139] By pre-dividing the battery module's full charge and discharge range at equal intervals... A set of discrete charged state points .

[0140] The customized module production parameter matrix is ​​essentially a one-dimensional column vector (or key-value pair mapping table), and any element within the matrix... Specifically:

[0141] ;

[0142] in, ;

[0143] In the formula, Indicates the corresponding number The absolute physical internal resistance target value calculated by reconstructing the inverse physical correction under a discrete state of charge node; Indicates the effective operating range of the battery module (e.g.) The total number of discrete state-of-charge test nodes (positive integers, e.g., in equal intervals of available capacity segments) When dividing by step size, ); Indicates by A finite set of independent variables consisting of discrete test nodes constitutes a one-dimensional independent reference coordinate axis for production parameter calibration. Represents a set The first in The specific discrete charge state node values ​​(such as...) , (equal discrete percentage of electricity); The terminating boundary node of the equally spaced set represents the last discrete charged state extreme node in the set (i.e., the upper limit of full charge in the target calibration interval). Nodes or test endpoints, corresponding to (state) Represents array The first position located by the pointer traversal One scalar element; This represents the index cursor variable (with a value of 1 to 1) used by the underlying program to perform memory addressing and matching. (positive integers between)

[0144] Specifically, the customized module production parameter matrix completely reduces the multidimensional, time-varying physical aging degradation to a "static internal resistance calibration curve that only varies with SOC"; when this matrix When the data is sent to the flexible scheduling unit 4, the cell sorting machine at the front end of the production line can directly traverse the matrix. Each element serves as a multi-point screening threshold; this matrix can also be directly retrieved at the mid-production line. Used to compare the internal resistance of the semi-finished product at the current workstation with that of the target element. The residuals are used to dynamically compensate for the laser welding penetration depth in a closed loop, thus achieving zero-error conversion of complex prediction data into actual production line actions.

[0145] The power generation-side energy storage module health status monitoring and production scheduling management platform also includes a flexible scheduling unit 4, which is used to schedule production during the remaining time before actual decommissioning. Within the pre-set production lead time before arrival, the customized module production parameter matrix is ​​analyzed, the cell sorting rules and module assembly process formula are automatically adjusted, and production is scheduled to manufacture the battery clusters within the remaining time before their actual retirement. Replacement modules with performance parameters that match the current timeline;

[0146] In this embodiment, the customized module production parameter matrix containing the absolute physical factory internal resistance target value is obtained from the impedance matching decision unit 3 (instantiated as a discrete real number mapping array on the underlying data structure). ), and the actual remaining decommissioning time output by evolution prediction unit 2. Then; the flexible scheduling unit 4 will use the actual remaining time of decommissioning. As the absolute delivery deadline, the production schedule is triggered in reverse order after deducting the predetermined logistics and production time. The material dynamic screening module 41 analyzes the discrete real number mapping array. The corresponding physical internal resistance target value is extracted as the core screening threshold of the sorting machine. By calling the automated testing system (ATE) in the factory, the batch of cells whose internal resistance benchmark best matches the target curve is accurately captured from the cell warehouse for the first round of grouping.

[0147] Furthermore, due to the inherent discreteness of the cell manufacturing process, it is extremely difficult for a single cell to be absolutely identical to the target physical internal resistance value calculated backward from the logistics (especially when the target factory impedance is required to be deliberately adjusted to reserve aging redundancy, the internal resistance of the sorted standard cells is often too small). At this time, the process linkage control module 42 intervenes and takes over, performing extremely precise closed-loop fine-tuning with a combination of hardware and software at the structural assembly end. The specific data flow and control logic are as follows:

[0148] The process linkage control module 42 first uses the ATE tester at the front end of the production line to obtain the average actual AC internal resistance of the batch of battery cells to be welded that have completed the first round of assembly at the current production line station. Subsequently, from the discrete real number mapping array issued in the preceding sequence... Extract the corresponding physical internal resistance target value from the factory. The difference calculation is performed in the underlying logic unit to obtain the impedance compensation residual that must be manually compensated by the assembly process. (Right now When identified (When this indicates that the current cell is relatively healthy and has low internal resistance, requiring the introduction of additional internal resistance to match the old cluster), the process linkage control module 42 does not adjust blindly, but rather compensates for the impedance residual. As a quantitative input feature, it initiates table lookup or interpolation operations to the system's built-in process mapping model.

[0149] It is worth noting that the process mapping model is a priori mathematical model built by the system based on DOE (Design of Experiments) physical calibration. Specifically, before the system is put into operation, a stepped welding destructive calibration test is performed in a laboratory environment for the specific cell electrode material (such as aluminum electrode) and bus material (such as aluminum bus or copper-aluminum composite bus). By sending discrete gradient laser power to the laser welding equipment (such as decreasing in 50W increments from 2000W) or changing the number of effective weld points, and by directly measuring the contact thermal resistance of the welded joint surface after forming using a four-wire high-precision micro-ohmmeter, a large number of measured sample pairs are collected. This allows for the fitting and generation of a nonlinear response characteristic curve or discrete lookup table characterizing the relationship between "laser welding process parameters" and "generated contact thermal resistance micro-ohm value". This nonlinear response characteristic curve and lookup table constitute the process mapping model embedded in non-volatile memory.

[0150] Specifically, the pre-calibrated and stored process mapping model includes a nonlinear mapping relationship between contact thermal resistance and laser welding parameters. This nonlinear mapping relationship typically exhibits a typical inverse proportional nonlinear monotonically decreasing trend: that is, as the laser power decreases stepwise, the effective weld depth and weld pool volume of the electrode and busbar during fusion welding decrease nonlinearly; due to the sharp decrease in the effective conductive cross-sectional area of ​​the bonding interface, the contact internal resistance connected in series at this physical node increases significantly in a quadratic or exponential manner. This process mapping model utilizes this nonlinearly amplified physical law, enabling the process linkage control module 42 to "extract" the required impedance compensation residual with extreme sensitivity and precision by slightly reducing the laser power. This does not require any destructive modifications to the overall mechanical fastening structure of the module.

[0151] Furthermore, a precise mathematical matching solution is initiated for this process mapping model (specifically, based on the input impedance compensation residual). In the underlying logic unit, a binary search algorithm is called to traverse the discrete lookup table, or a one-dimensional linear interpolation algorithm is called to perform reverse root-finding on the nonlinear response characteristic curve, to derive the corresponding variable parameter welding command (specifically, the variable parameter welding command is specifically classified in the underlying unit as a target welding laser power reduction command or a target effective weld point reduction command).

[0152] Finally, the variable-parameter welding instructions are compiled in real time into executable machine-readable code at the bottom layer of a programmable logic controller (PLC) or robotic arm. This code is then directly sent to the laser welding equipment on the module assembly line via the factory's PROFINET or EtherCAT industrial control bus to perform adaptive variable-parameter welding. At the underlying physical execution level, the controller of the laser welding equipment responds to the received machine-readable code, automatically adjusting the output power curve of the laser generator or driving the galvanometer to change the scanning trajectory of the laser beam. This precisely reduces the molten pool depth or effective melting area at the interface between the electrode and the busbar. Without compromising the mechanical strength of the module structure, an absolute equivalent impedance compensation residual is artificially and controllably generated at the connection point at the physical level. Customized contact thermal resistance.

[0153] This invention utilizes the series equivalence principle in electrical engineering to determine the average actual AC internal resistance of a batch of battery cells to be welded. The customized contact thermal resistance is superimposed in series on the module's physical structure (i.e. This achieves precise compensation for the internal resistance residual, thereby manufacturing an overall internal resistance that is absolutely close to the physical factory-set internal resistance target value. And it meets the factory-specified replacement module requirements.

[0154] The power generation-side energy storage module health status monitoring and production scheduling management platform also includes a closed-loop feedback unit 5, which is used to monitor the replacement module during its remaining actual decommissioning time. After the target degraded module is replaced and put into operation, an updated cluster-level ledger is generated and sent to the battery management system to which the target degraded module belongs to perform adaptive parameter correction.

[0155] In this embodiment, after the replacement module is manufactured and shipped, the platform's data control is transferred to the closed-loop feedback unit 5 at the tail end; the replacement module enters the trunk logistics, during which predictable solid-phase interface calendar aging occurs internally. When the timeline precisely reaches the remaining time of the actual decommissioning... At that time, on-site maintenance personnel will install the replacement module into the target battery cluster.

[0156] Due to the attenuation matrix accumulated by the new module during its on-the-road operation and static setting at room temperature It precisely fills the internal resistance difference that was intentionally reserved at the factory (i.e. Its current actual grid-connected physical impedance and the expected average impedance of the surrounding older modules within the cluster. Achieving a perfect balance eliminates the intra-cluster circulation and the barrel effect of charging and discharging from a physical level.

[0157] Furthermore, after completing the physical wiring, the closed-loop feedback unit 5 detects and extracts the grid-connected identity identifier of the new module (such as the MAC address or serial number of the BMS slave board) through the communication gateway of the field energy storage station. Since the new module has been perfectly assimilated with the current battery cluster in terms of physical performance, the closed-loop feedback unit 5 directly uses the previously calculated expected average impedance. With the expected remaining capacity Assign a grid-connected identity identifier, generate an updated cluster-level ledger, and then send an update command to the battery management system (BMS) main control board on site.

[0158] After receiving the instruction, the on-site battery management system will overwrite and update the old ledger data for the degraded modules in the cluster with this new cluster-level ledger. Based on the impedance and capacity benchmarks that have been agreed upon in the cluster, it will automatically remove the "derating / current limiting charge and discharge protection mechanism" that was forced to be set due to the performance shortcomings of the old modules.

[0159] Specifically, before physical replacement, the target degraded module, due to its drastically increased internal resistance and capacity decay, would prematurely reach the overcharge or over-discharge cutoff voltage threshold when responding to high-rate charge / discharge scheduling by the power grid. To prevent thermal runaway caused by the degraded module exceeding the voltage limit, the on-site battery management system was forced to fully reduce the upper limit of the charge / discharge current of the entire battery cluster (i.e., current limiting protection) based on the "weakest link" principle, directly resulting in a severe reduction in the overall available output of the energy storage system.

[0160] And at present After seamless component replacement and updating of the logbook, the newly added replacement module has its physical impedance completely aligned with the expected average impedance of the other healthy modules. During the charging and discharging process, the voltage distribution of individual cells in the entire target battery cluster will regain a high degree of synchronization and current sharing. After monitoring that the voltage difference has converged back to within the safety zone, the on-site battery management system automatically releases the previous current constraint threshold. At this point, the risk of impedance mismatch in the entire target battery cluster is completely eliminated, and the system returns to a fully healthy scheduling state within the rated power range, realizing a closed-loop scheduling of the entire life cycle of the industrial internet across spatial dimensions.

[0161] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A platform for monitoring the health status and managing the production scheduling of energy storage modules on the power generation side, characterized in that, include: feature The acquisition unit (1) is used to continuously acquire the charge and discharge response data and thermal field distribution characteristics of each energy storage module in the power generation side energy storage battery cluster under operating condition fluctuations, and extract time-series feature parameters. ; Evolution prediction unit (2), the evolution prediction unit (2) is used to input the time series feature parameters into a preset health evolution model to predict the initial remaining time when the target deterioration module reaches the retirement threshold. And combining the charge / discharge response data with the thermal field distribution characteristics, the preliminary remaining time is determined. Perform inverse nonlinear reduction correction to obtain the actual remaining decommissioning time. Simultaneously, the remaining time before actual retirement is predicted. At any given time, the expected average impedance of the remaining healthy modules within the battery cluster containing the target degraded module. With expected remaining capacity ; Impedance matching decision unit (3), the impedance matching decision unit (3) is used to match the expected mean impedance. and expected remaining capacity Based on, and in conjunction with the actual remaining time of retirement. Construct a customized module production parameter matrix; Flexible scheduling unit (4), the flexible scheduling unit (4) is used for the remaining time of actual decommissioning Within the pre-set production lead time before arrival, the customized module production parameter matrix is ​​analyzed, the cell sorting rules and module assembly process formula are automatically adjusted, and production is scheduled to manufacture the battery clusters within the remaining time before their actual retirement. Replacement modules with performance parameters that match the current time; Closed-loop feedback unit (5), the closed-loop feedback unit (5) is used to address the remaining time of actual decommissioning of the replacement module. After the target degraded module is replaced and put into operation, an updated cluster-level ledger is generated and sent to the battery management system to which the target degraded module belongs to perform adaptive parameter correction.

2. The power generation-side energy storage module health status monitoring and production scheduling management platform according to claim 1, characterized in that, The feature acquisition unit (1) includes a working condition response capture module (11), a physical thermal field sensing module (12), and a time-series feature extraction module (13). The operating condition response capture module (11) is used to capture the charging and discharging response data of the module when it is subjected to a step current of 1C or more under the power change condition on the power generation side. The charging and discharging response data specifically includes the voltage transient drop depth of the energy storage module when it is subjected to a step current under the power change condition on the power generation side. and polarization internal resistance recovery time constant ; The physical thermal field sensing module (12) is used to acquire the thermal field distribution characteristics of the battery cell tab region and the central region inside the energy storage module. The thermal field distribution characteristics specifically include the real-time temperature difference gradient. ; The time-series feature extraction module (13) is used to extract the voltage transient drop depth within a preset time window. With polarization internal resistance recovery time constant The first-order rate of change over time determines the dynamic growth slope of the polarization internal resistance; the real-time temperature difference gradient within the preset time window is used. Time-integral calculations are performed over time to obtain the cumulative thermal stress integral value. The dynamic growth slope of the polarization internal resistance and the cumulative thermal stress integral value are then mapped to the same time-domain coordinate system for multi-dimensional feature fusion to obtain time-series feature parameters. .

3. The power generation-side energy storage module health status monitoring and production scheduling management platform according to claim 1, characterized in that, The preset health evolution model is a multi-head self-attention neural network that incorporates irreversible physical decay boundary constraints; The evolution prediction unit (2) includes a mechanism feature fusion module (21), a deep temporal extrapolation module (22), and an extreme condition abrupt change correction module (23). Among them, the mechanism feature fusion module (21) extracts the frequency domain features of the historical dispatch power spectrum on the generation side through the historical output curve of the energy storage converter, and combines the frequency domain features with the time series feature parameters. Perform orthogonal concatenation mapping to construct a multidimensional joint input feature tensor; The deep temporal extrapolation module (22) is used to input the multi-dimensional joint input feature tensor into the self-attention temporal encoding layer of the multi-head self-attention neural network to capture the temporal feature parameters under normal operating conditions. The long- and short-term nonlinear decay characteristics that evolve over time, and the parameter drift of the multi-head self-attention neural network output constrained by the irreversible physical decay boundary, yield a preliminary remaining time. And the mean impedance trajectory curve of the target battery cluster over future time. With remaining capacity trajectory curve ; When the extreme operating condition sudden change correction module (23) detects that the generator side has triggered a continuous high-rate step peak shaving command, it calculates the transient electrothermal coupling stress factor based on the extracted high current impact duration and the transient temperature rise rate of the battery cell, and calculates the preliminary remaining time based on the transient electrothermal coupling stress factor. Perform inverse nonlinear reduction correction to obtain the actual remaining decommissioning time. and in the mean impedance trajectory curve With the remaining capacity trajectory curve The above refers to the actual remaining time of retirement. Synchronous addressing is performed using time index coordinates, and the values ​​at the corresponding times are extracted as the expected mean impedance. With expected remaining capacity .

4. The power generation-side energy storage module health status monitoring and production scheduling management platform according to claim 3, characterized in that, The specific steps involved in limiting the parameter drift of the multi-head self-attention neural network output based on the irreversible physical decay boundary constraint are as follows: The multi-dimensional joint input feature tensor is input into the multi-head self-attention neural network. The attention weight matrix of the frequency domain features, the dynamic growth slope of polarization internal resistance, and the cumulative thermal stress integral value on the historical time series is calculated by multiple parallel attention heads to extract multi-scale time-series correlation features. In the forward inference of the multi-head self-attention neural network, the current prediction time step is extracted. With the next prediction time step Network pre-output node data; The irreversible physical decay boundary constraint is executed. The irreversible physical decay boundary constraint is quantized into a physical monotonicity partial derivative condition in the neural network. The physical monotonicity partial derivative condition is set as follows: the first-order partial derivative with respect to the time step is calculated for the pre-output node data of the network. It is required that the first-order rate of change of the expected mean impedance with time must remain non-negative, and the first-order rate of change of the expected remaining capacity with time must remain non-positive. The first-order rate of change of the network pre-output node data is substituted into the physical monotonicity partial derivative condition for verification. If the network pre-output node data is found to violate the physical monotonicity partial derivative condition, an exponential penalty gradient is applied to the model loss function to correct the weights, thereby limiting the parameter drift of the network during long-period extrapolation, until the network reaches steady-state convergence that meets the irreversible physical decay boundary constraint. The multidimensional joint input feature tensor is fed into the multi-head self-attention neural network that has completed steady-state convergence. Through forward computation, the preliminary remaining time is output. And the mean impedance trajectory curve of the target battery cluster over future time. With remaining capacity trajectory curve .

5. The power generation-side energy storage module health status monitoring and production scheduling management platform according to claim 4, characterized in that, Regarding the preliminary remaining time Perform inverse nonlinear reduction correction to obtain the actual remaining decommissioning time. The specific steps involved are as follows: The system monitors the charging and discharging current of the target battery cluster in real time. If the absolute value of the current continuously exceeds the preset safety rate threshold, a high-precision timer is activated to accumulate and calculate the duration of the large current surge. ; The real-time temperature data collected by the feature acquisition unit is acquired synchronously, and the current transient temperature rise rate of the battery cell is calculated through discrete first-order difference operation. ; The duration of the high current impact With the transient temperature rise rate of the battery cell As a joint index key, it is retrieved into the pre-calibrated and stored electrothermal coupling acute injury mapping matrix, and spatially smoothed by bilinear interpolation. The output is a dimensionless parameter greater than or equal to 1, which serves as the transient electrothermal coupling stress factor. ; Read the preliminary remaining time output by the depth time extrapolation module (22) Perform nonlinear reduction correction with the initial remaining time. With transient electrothermal coupling stress factor The remaining time of retirement is considered as the actual retirement time. ; The actual remaining retirement time was calculated. Then, in the mean impedance trajectory curve With remaining capacity trajectory curve In the context of the actual remaining time before retirement, Addressing and truncation are performed at each time point to extract the transient values ​​at the corresponding moments as the final expected mean impedance. With expected remaining capacity ; and the actual remaining time of retirement Expected mean impedance With expected remaining capacity The output is sent to the impedance matching decision unit (3).

6. The power generation-side energy storage module health status monitoring and production scheduling management platform according to claim 1, characterized in that, The impedance matching decision unit (3) includes a reference impedance alignment module (31) and an in-transit aging reverse compensation module (32). The reference impedance alignment module (31) is used to read the expected mean impedance. With expected remaining capacity Furthermore, for the replacement modules to be produced, based on their reference internal resistance requirements and expected average impedance, and by introducing a preset passive balancing tolerance index, an initial impedance matching constraint equation is constructed. The initial impedance matching constraint equation is used to limit the allowable impedance deviation boundary when the replacement module is connected to the grid. The in-transit aging reverse compensation module (32) is used to read the actual remaining time of retirement. Furthermore, considering the geographical span between the manufacturing base and the power generation-side energy storage station, the transportation logistics cycle constant and the ambient temperature sequence along the route are extracted. The attenuation matrix of the replacement module during the logistics transportation and warehousing static storage stages due to calendar aging is calculated. And the attenuation matrix during the journey at room temperature The initial impedance matching constraint equation is introduced as a negative compensation variable to generate a customized module production parameter matrix containing the target value of the physical factory internal resistance, and output to the flexible scheduling unit (4).

7. The power generation-side energy storage module health status monitoring and production scheduling management platform according to claim 6, characterized in that, The in-transit aging reverse compensation module (32) generates the customized module production parameter matrix, involving the following steps: The reference impedance alignment module (31) constructs the initial impedance matching constraint equation for multiple discrete state of charge points, based on the reference internal resistance requirement of the replacement module under ideal grid-connected conditions and the expected average impedance of the target battery cluster, and introduces a preset passive equalization current allowable deviation coefficient as a tolerance constraint index; wherein, the initial impedance matching constraint equation requires that the absolute impedance difference between the reference internal resistance requirement and the expected average impedance shall not exceed the dynamic allowable deviation boundary determined by multiplying the expected average impedance by the passive equalization current allowable deviation coefficient. The in-transit aging reverse compensation module (32) is based on the remaining time of actual retirement. Working backwards from the production and delivery dates, we extract the estimated logistics and transportation time from the logistics planning. And discrete environmental temperature sequences along the route predicted based on geographic meteorological data. ; The discrete Arrhenius aging time-varying integral model is used to predict logistics transportation time. With the discrete ambient temperature sequence along the route As a boundary condition input, the passivation increment of the solid-phase interface at each discrete charge state point is integrated and accumulated to obtain the attenuation matrix at room temperature during the journey. ; The attenuation matrix during the journey at room temperature The initial impedance matching constraint equation is introduced as a negative compensation variable to perform reverse correction of the factory calibration, resulting in a customized module production parameter matrix that includes the target value of the physical factory internal resistance. The factory calibration logic of the customized module production parameter matrix is ​​limited to the following: the absolute impedance difference between the target value of the physical factory internal resistance and the equivalent superposition value of the attenuation matrix at room temperature during transit, and the expected average impedance must not exceed the dynamic allowable deviation boundary determined by multiplying the expected average impedance by the allowable deviation coefficient of the passive equalization current.

8. The power generation-side energy storage module health status monitoring and production scheduling management platform according to claim 1, characterized in that, The flexible scheduling unit (4) includes a material dynamic screening module (41) and a process linkage control module (42). The material dynamic screening module (41) is used in the actual remaining decommissioning time. Within the preset production lead time before arrival, the customized module production parameter matrix is ​​analyzed, the physical factory internal resistance target value in the customized module production parameter matrix is ​​extracted, and the front-end cell sorting equipment is controlled to screen out cell batches whose AC internal resistance characteristics are close to the physical factory internal resistance target value. The process linkage control module (42) is used to automatically adjust the bus laser welding parameters in the module assembly process when the internal resistance of a single cell cannot fully match the target value of the physical factory internal resistance. By adjusting the laser welding penetration or effective welding area, controllable contact thermal resistance is introduced to compensate for the residual between the cell body impedance and the target value of the physical factory internal resistance, so as to manufacture a replacement module that meets the factory calibration.

9. The power generation-side energy storage module health status monitoring and production scheduling management platform according to claim 8, characterized in that, The process linkage control module (42) introduces controllable contact thermal resistance by adjusting the laser welding penetration depth or effective welding area to manufacture a replacement module that meets the factory specifications. The specific steps involved are as follows: After the material dynamic screening module completes the first round of grouping, obtain the average actual AC internal resistance of the batch of battery cells to be welded at the current workstation of the production line. ; Extract the target value of physical internal resistance from the customized module production parameter matrix. And calculate the target value of the physical internal resistance at the factory. With the average actual AC internal resistance Impedance compensation residuals between ; If impedance compensation residual If the value is greater than 0, then the impedance compensation residual will be... As input features, they are input into a pre-calibrated and stored process mapping model, and the corresponding variable parameter welding instructions are obtained by solving and matching. Specifically, the variable parameter welding instructions are either the target welding laser power reduction instruction or the target effective weld point reduction instruction. The variable-parameter welding instructions are compiled into low-level executable machine-readable code and sent to the laser welding equipment on the module assembly line via the industrial control bus to perform adaptive variable-parameter welding. This accurately generates an equivalent impedance compensation residual at the interface between the cell terminal and the bus. Customized contact thermal resistance; The average actual AC internal resistance of the batch of cells to be welded is obtained by... The customized contact thermal resistance is physically superimposed on the structure, thereby creating an overall internal resistance that is absolutely close to the target value of the physical factory internal resistance. And it meets the factory-specified replacement module requirements.

10. The power generation-side energy storage module health status monitoring and production scheduling management platform according to claim 1, characterized in that, The closed-loop feedback unit (5) is used to monitor the remaining time of actual decommissioning after the replacement module has undergone logistics transportation and room temperature storage. After the target degraded module is replaced and put into operation at the node, the final grid-connected identity identifier of the replaced module is extracted; and the expected mean impedance With the expected remaining capacity Assign the final grid-connected identity identifier and generate an updated cluster-level ledger; The updated cluster-level ledger is then sent to the field battery management system to which the target degraded module belongs, driving the field battery management system to remove the derating and current-limiting charge / discharge protection mechanism set based on the original degradation state of the target degraded module, thus completing the physical closed loop from production capacity to operation and maintenance.

Images