A lithium battery energy storage platform thermal energy management optimization system

Abstract

This invention relates to the field of lithium battery energy storage and thermal management technology, specifically to a thermal energy management optimization system for a lithium battery energy storage platform. The system includes a baseline model construction step: constructing an ideal electrothermal coupling baseline model based on static parameters and dynamic time-series data, and calculating the baseline simulation temperature field; a fault mode simulation step: injecting abnormal physical factors in parallel to generate multiple sets of theoretical abnormal temperature curves; a result comparison and analysis step: calculating the first and second residual vectors between the measured data, theoretical curves, and the baseline field; and a verification and strategy generation step: classifying faults based on the waveform similarity between the residual vectors and outputting a verification report. This invention effectively eliminates the interference of normal temperature rise caused by load fluctuations by introducing an intermediate reference system, achieving early warning of thermal runaway risks.

Description

Technical Field

[0001] This invention relates to the field of lithium battery energy storage and thermal management technology, specifically to a thermal energy management optimization system for a lithium battery energy storage platform. Background Technology

[0002] In existing lithium battery energy storage platforms, the system typically integrates battery management modules and thermal management components. By collecting real-time data such as voltage, current and temperature of individual battery cells, the system monitors and protects the operating status of the energy storage power station.

[0003] The current mainstream thermal fault early warning method mainly adopts the threshold monitoring method, that is, a fixed temperature alarm upper limit or temperature rise rate threshold is preset in the system. Once the real-time temperature data collected by the sensor exceeds the preset range, the system determines that there is a risk of thermal runaway and triggers the corresponding alarm mechanism.

[0004] However, when the energy storage platform is operating under high-frequency frequency regulation or severe load fluctuations, the large current surge will cause a significant normal temperature rise in the battery. This physical temperature rise often masks the weak thermal characteristics caused by early faults such as micro-short circuits and insulation aging. Traditional threshold monitoring methods cannot physically separate the common effects of current fluctuations on temperature, making it difficult to distinguish between normal load temperature rise and abnormal fault temperature rise. As a result, the system has low accuracy in identifying hidden thermal runaway risks and is prone to missed or false alarms. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a thermal energy management optimization system for a lithium battery energy storage platform. Specifically, the technical solution of this invention includes:

[0006] The model data interface module is used to obtain static physical parameters and dynamic time-series data of the lithium battery energy storage platform from an external battery management system or to be configured by the user. The dynamic time-series data includes real-time current, cell voltage and sensor real-time temperature.

[0007] The benchmark simulation model construction module is used to construct an ideal electrothermal coupling benchmark model based on the static physical parameters, and drive the model based on the dynamic time series data to simulate and calculate the benchmark simulation temperature field of the lithium battery energy storage platform under fault-free operating conditions.

[0008] The fault mode simulation module is used to inject predefined abnormal physical factor perturbations corresponding to multiple fault types into the ideal electrothermal coupling reference model in parallel and perform simulations to generate multiple sets of theoretical abnormal temperature curves with different fault characteristics.

[0009] The simulation result comparison and analysis module is used to calculate the first residual vector between the sequence of real-time temperature data from the sensor and the reference simulated temperature field, and the second residual vector between each set of theoretical abnormal temperature curves and the reference simulated temperature field.

[0010] The model verification and strategy generation module is used to calculate the waveform similarity between the first residual vector and each group of second residual vectors, and classify the shape of the first residual vector according to the waveform similarity, and finally output a simulation verification report describing the suspected fault type and the corresponding strategy.

[0011] Preferably, the model data interface module includes:

[0012] The static parameter configuration unit is used to receive or extract the three-dimensional geometric topology data, material thermal property parameters and internal flow channel model data of the simulated object, and configure the static physical parameters as the initial boundary conditions of the ideal electrothermal coupling reference model.

[0013] The dynamic data excitation unit is used to receive external real-time synchronous data streams, which include load current, environmental parameters and cooling medium flow rate, and load the dynamic time-series data as a time-varying excitation source to the corresponding input port of the ideal electrothermal coupling reference model.

[0014] Preferably, the benchmark simulation model construction module includes:

[0015] The multiphysics modeling unit is used to construct the ideal electrothermal coupling benchmark model in the simulation environment based on the electrochemical student heat equation and the lumped parameter thermal network theory. The model is configured to only respond to the simulation of normal temperature rise caused by standard load fluctuations.

[0016] The pure field simulation calculation unit is used to input the dynamic time series data into the ideal electrothermal coupling reference model for numerical calculation, and output the theoretical temperature field sequence stripped of preset aging factors and internal short circuit factors as the reference simulation temperature field.

[0017] Preferably, the fault mode simulation module includes:

[0018] The fault factor library unit is used to store a variety of predefined abnormal physical factors, including aging disturbance factors for simulating component performance degradation, micro short-circuit disturbance factors for simulating abnormal leakage current paths, and convection attenuation disturbance factors for simulating deterioration of heat dissipation conditions.

[0019] The parallel disturbance simulation unit is used to superimpose the abnormal physical factors as independent disturbance terms onto the corresponding parameters of the ideal electrothermal coupling reference model, perform parallel calculations within the same simulation time window, and output multiple sets of theoretical abnormal temperature curves containing specific fault morphology characteristics.

[0020] Preferably, the simulation result comparison and analysis module includes:

[0021] The measured data residual generation unit is used to perform point-by-point difference calculation between the real-time temperature data sequence of the sensor and the reference simulated temperature field to generate the first residual vector that eliminates the influence of the reference operating condition fluctuation. The first residual vector represents the mixed feature morphology of unknown anomalies and external interference.

[0022] The theoretical feature residual extraction unit is used to perform point-by-point difference calculation between each set of theoretical abnormal temperature curves and the benchmark simulation temperature field to generate multiple sets of second residual vectors that contain only specific fault morphology features. Each set of second residual vectors represents a pure preset fault thermal feature template.

[0023] Preferably, the model validation and policy generation module includes:

[0024] The feature matching degree calculation unit is used to calculate the waveform matching degree between the first residual vector and each group of second residual vectors within a preset simulation analysis time window using a dynamic time warping algorithm or a Pearson correlation coefficient algorithm, so as to obtain the waveform similarity.

[0025] The fault mode classification unit is used to sort the waveforms according to their similarity and mark the preset fault type corresponding to the second residual vector with the highest matching degree as the inferred fault type for the first residual vector.

[0026] Preferably, the model validation and policy generation module further includes:

[0027] The simulation confidence decision unit, used to make a confidence decision on the presumed fault type, is configured as follows:

[0028] If the waveform similarity is greater than or equal to a preset confidence threshold, it is determined that the first residual vector is highly consistent with the theoretical characteristics of the presumed fault type, and a high-confidence fault verification conclusion is generated.

[0029] If the waveform similarity is less than the confidence threshold, it is determined that the first residual vector cannot be effectively matched with any preset fault theory features, and a verification report prompting the inspection of model or data noise is generated.

[0030] Preferably, the model validation and policy generation module further includes:

[0031] The simulation optimization strategy generation unit, configured to generate an optimization suggestion report for the simulation model or control strategy based on the fault verification conclusions, is as follows:

[0032] If the fault type is presumed to be a micro short circuit, the optimization suggestion report includes a suggestion to introduce the corresponding fault disturbance factor in subsequent simulations, as well as a simulation scenario description for testing the effectiveness of the external current regulation strategy.

[0033] If the fault type is presumed to be heat dissipation deterioration, the optimization recommendation report includes a recommendation to calibrate the local convective heat transfer coefficient, as well as a set of simulation boundary condition parameters for reconstructing and testing the control model of the external cooling system.

[0034] Compared with existing technologies, this system effectively eliminates the interference of normal temperature rise caused by current in high-frequency frequency regulation or severe load fluctuation scenarios in energy storage power stations by introducing an ideal electrothermal coupling reference model as an intermediate reference system and performing dual-track differential calculation. By calculating the residuals between the sensor measured data and the reference field, as well as the residuals between the theoretical fault curve and the reference field, the system eliminates the common influence of current fluctuations on temperature at the physical level. This enables the accurate identification of weak fault signals such as micro-short circuits or insulation aging that are masked by normal temperature rise, solving the technical problem that traditional threshold alarms cannot distinguish between high current high temperature and fault high temperature, and realizing early warning of thermal runaway risk. Attached Figure Description

[0035] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0036] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0038] Example 1:

[0039] Please see Figure 1 A thermal energy management optimization system for a lithium battery energy storage platform, which serves as an upper-level simulation and verification platform for the lithium battery energy storage platform, includes:

[0040] The model data interface module is used to obtain static physical parameters and dynamic time-series data of the lithium battery energy storage platform from an external battery management system or by user configuration. The dynamic time-series data includes real-time current, cell voltage and sensor real-time temperature.

[0041] The benchmark simulation model building module is used to build an ideal electrothermal coupling benchmark model based on static physical parameters, and drive the model based on dynamic time series data to simulate and calculate the benchmark simulation temperature field of the lithium battery energy storage platform under fault-free operating conditions.

[0042] The fault mode simulation module is used to inject predefined abnormal physical factor perturbations corresponding to various fault types into an ideal electrothermal coupling reference model in parallel and perform simulations to generate multiple sets of theoretical abnormal temperature curves with different fault characteristics.

[0043] The simulation result comparison and analysis module is used to calculate the first residual vector between the sequence of real-time temperature data from the sensor and the reference simulated temperature field, and the second residual vector between each set of theoretical abnormal temperature curves and the reference simulated temperature field.

[0044] The model verification and strategy generation module is used to calculate the waveform similarity between the first residual vector and each group of second residual vectors, and classify the shape of the first residual vector according to the waveform similarity. Finally, it outputs a simulation verification report describing the suspected fault type and the corresponding response strategy.

[0045] This embodiment details the core architecture and synthetic analysis method operation logic of the system. The system is configured as a verification environment based on digital twins. Its core logic is not to directly detect faults, but to infer the abnormal characteristics of real data by constructing the difference between the ideal fault-free state and the theoretical fault state in real time. The system operation process is as follows: The system uses the benchmark simulation model construction module to build a completely healthy ideal model. This model only responds to changes in current load and ambient temperature and outputs the benchmark simulation temperature field.

[0046] Meanwhile, the fault mode simulation module, based on the ideal model, injects multiple hypothetical fault factors in parallel and calculates multiple sets of theoretical abnormal temperature curves; the simulation result comparison and analysis module subtracts the actual temperature collected by the sensor from the reference temperature to obtain the first residual vector representing the actual deviation, and subtracts each set of theoretical abnormal temperatures from the reference temperature to obtain multiple sets of second residual vectors representing theoretical fault feature templates; the model verification and strategy generation module determines what kind of fault has occurred by comparing the waveform similarity between the actual deviation and the theoretical fault feature templates.

[0047] This embodiment introduces an ideal electrothermal coupling benchmark model as an intermediate reference system, effectively eliminating the interference of normal temperature rise caused by drastic load fluctuations in the high-frequency regulation scenario of energy storage power stations. Traditional threshold alarms cannot distinguish between high temperature caused by large current and high temperature caused by micro short circuits. However, this system eliminates the common impact of current fluctuations on temperature from a physical level by calculating residuals, thereby accurately identifying weak fault signals masked by normal temperature rises and realizing early warning of thermal runaway risks.

[0048] Example 2:

[0049] The model data interface module includes:

[0050] The static parameter configuration unit is used to receive or extract the three-dimensional geometric topology data, material thermal property parameters and internal flow channel model data of the simulated object, and configure the static physical parameters as the initial boundary conditions of the ideal electrothermal coupling reference model.

[0051] The dynamic data excitation unit is used to receive external real-time synchronous data streams, including load current, environmental parameters and cooling medium flow rate, and load the dynamic time-series data as a time-varying excitation source to the corresponding input port of the ideal electrothermal coupling reference model.

[0052] This embodiment provides a specific configuration for the model data interface module, aiming to bridge the boundary between the physical world and the digital world. The static parameter configuration unit is configured to receive the physical properties of the simulated object. In this embodiment, the static physical parameters include, but are not limited to, the three-dimensional geometric dimensions of the battery module, the specific heat capacity and thermal conductivity of the cell material, and the topology of the coolant flow channel. These parameters are usually derived from the CAD design files or factory calibration data of the battery system, and they define the skeleton of the simulation model.

[0053] The dynamic data excitation unit synchronously acquires external inputs via a CAN bus or Ethernet interface at a preset sampling frequency; the dynamic timing data specifically includes the real-time charging and discharging current of the energy storage system, the ambient temperature outside the battery pack, and the temperature of the cooling medium at the inlet of the liquid cooling system. And feedback on the speed or flow rate of the liquid cooling system pump; it should be noted here that, in order to strictly correspond to the cooling medium flow rate defined in the embodiment and eliminate the defects of black box mapping, the system has an explicit fluid dynamics conversion logic at the input end. Considering that the energy storage platform usually contains multiple parallel battery clusters or modules, this logic is configured to execute the following conversion formula including fluid distribution coefficient and unit conversion factor:

[0054]

[0055] in, : The flow rate of the cooling medium in the flow channel of the target simulation module at any given time ( ); The volume conversion constant based on the International System of Units (SI) is physically based on... Lift( )equal cubic meter( ), used to change the displacement unit from liters ( Converted to cubic meters ) dimensional conversion factor; Real-time collected total pipeline pump speed, i.e., rev / min; The pump's displacement per revolution (L / rev) is derived from static parameters. Pump volumetric efficiency coefficient, dimensionless, with a value range of 0.8-0.95; Effective cross-sectional area of ​​cooling channels in a single module ( (), derived from flow channel model data; The total number of parallel flow path branches in the system; Time unit conversion factor, unit is This makes the final calculation result The dimensions strictly satisfy ;

[0056] This coefficient is used to determine the rotational speed. The time base is converted from minutes to seconds, i.e. This ensures that the flow units in the molecule are correctly converted. This makes the final calculation result The dimensions strictly satisfy This eliminates the ambiguity of dimension mismatch; the formula ensures a deterministic mapping from the sensor's raw readings to the model's physical input, effectively solving the problems of velocity distribution calculation and dimension unification under multi-branch flow channels;

[0057] This embodiment distinguishes between static parameters and dynamic excitations, ensuring the rigidity of the simulation model in physical structure and its flexibility in operating conditions. Especially in the scenario of dynamic adjustment of liquid cooling system, the real-time flow rate is incorporated into the dynamic excitation, enabling the benchmark model to dynamically follow the adjustment of cooling system, preventing false alarms caused by changes in cooling strategy, and ensuring a high degree of synchronization between digital twin and physical entity in variable environments.

[0058] Example 3:

[0059] The benchmark simulation model building module includes:

[0060] The multiphysics modeling unit is used to construct an ideal electrothermal coupling benchmark model in the simulation environment based on the electrochemical student heat equation and the lumped parameter thermal network theory. The model is configured to only respond to the simulation of normal temperature rise caused by standard load fluctuations.

[0061] The pure field simulation solution unit is used to input dynamic time series data into an ideal electrothermal coupling reference model for numerical calculation, and outputs a theoretical temperature field sequence stripped of preset aging factors and internal short-circuit factors as a reference simulation temperature field.

[0062] This embodiment details the implementation principle of the benchmark simulation model construction module; the multiphysics modeling unit uses the lumped parameter method to construct the thermal network model. To accurately calculate the heat generation rate of the battery cell and solve the algebraic loop problem in electrothermal coupling, this embodiment discretizes the Bernardi heat generation equation with clear timing; the heat generation rate calculation formula is as follows:

[0063]

[0064] in, Current simulation step Instantaneous heat production rate; Real-time load current is derived from the dynamic data excitation unit; The ohmic internal resistance of the battery cell comes from the calibration value in the static parameter configuration unit; Current simulation step The known temperature state at the initial time, where the subscript... Representing the battery body, this variable characterizes the average lumped temperature inside the cell. This value is derived from the iterative calculation output of the previous time step. The time point serves as the initial condition; here, it is explicitly agreed that a forward explicit computation logic is used, that is, utilizing the known state. The current heat production rate is calculated, thereby decoupling the electrothermal coupling equation in numerical calculation and avoiding the algebraic loop problem; Entropy coefficient, derived from the thermal properties of materials, is measured in volts per Kelvin. Characterizing the rate of change of battery electromotive force with temperature

[0065] The pure field simulation unit uses the aforementioned heat generation rate, combined with the heat transfer equation, to calculate the baseline simulation temperature field; the heat balance equation is as follows:

[0066]

[0067] in, The equivalent thermal mass of the battery module ( ), The average specific heat capacity of the battery material ( ), The effective heat exchange area between the battery module and the cooling medium ( ); The temperature of the cooling medium or ambient boundary, indicated by the subscript. This represents the cooling environment. To ensure the executability of this differential equation in a digital system and to define the computation timing, this embodiment uses the explicit Euler method to discretize it, specifying a serial execution flow of reading the state -> calculating heat generation -> updating the temperature. The specific code-level implementation formula is as follows:

[0068]

[0069] in, The baseline simulation temperature at the next moment; The current baseline simulation temperature; The simulation step size is configured to be 0.1s in this embodiment, which satisfies the thermal inertial stability constraint CFL condition. These represent equivalent mass, specific heat capacity, and heat transfer area, respectively. The real-time convective heat transfer coefficient, expanded in the discretized formula, is as follows: ;in, The basic natural convection heat transfer coefficient, in units of The values ​​are derived from the system's thermal calibration test under static cooling medium conditions. For the flow velocity gain coefficient, in order to eliminate ambiguity in the unit representation and ensure dimensional consistency, its unit is explicitly characterized as... This composite unit is derived from the unit of convective heat transfer coefficient. Divide by the unit of flow velocity The coefficient, derived from the Dittus-Boelter correlation and calculated based on the hydraulic diameter of the flow channel, is used to quantify the nonlinear enhancement relationship of the forced convection effect with flow velocity.

[0070] The value of this coefficient is derived from the fluid dynamics experimental calibration of a specific flow channel structure, that is, by changing the flow velocity. And measure the actual heat transfer coefficient Using the least squares method to analyze the correlation equation The results obtained through parameter fitting demonstrate the physical property that the fluid boundary layer thickness decreases with increasing flow velocity.

[0071] Regarding the boundary driving temperature in the formula In order to solve the problem of liquid cooling pump shutdown, i.e. To address the resulting discontinuities in physical states and numerical oscillations in simulations, this embodiment constructs a stagnant fluid thermal inertia sub-model to replace the simple conditional jump logic; the specific execution logic is as follows: A state variable is set. Characterizes the real-time temperature of the cooling medium within the flow channel; if ,but That is, the fluid is rapidly updated to the inlet temperature; if Then, perform a natural cooling iteration:

[0072]

[0073] in, Let be the thermal time constant of the quiescent fluid within the flow channel. To ensure the physical calculability of this parameter, the calculation formula is defined in this embodiment as follows:

[0074]

[0075] in, For the density of the cooling medium ( ), Specific heat capacity of the cooling medium ( Both are derived from the static physical parameter configuration of Example 2; The fluid volume of the flow channel within a single module ( ), For the effective heat transfer area of ​​the flow channel ( ); The physical meaning of is the thermal inertia time constant of the static fluid in the flow channel. It characterizes the speed at which the temperature of the residual fluid in the pipe relaxes with the change of ambient temperature when the coolant stops flowing. This parameter is verified by experimentally measuring the time constant of the fluid static cooling curve.

[0076] Finally, take the... This processing ensures that the temperature field curve of the simulation model maintains physical smoothness under the condition of frequent start-up and shutdown of the liquid cooling system, avoiding spurious residual signals caused by boundary condition steps.

[0077] Example 4:

[0078] The fault mode simulation module includes:

[0079] The fault factor library unit is used to store a variety of predefined abnormal physical factors, including aging disturbance factors for simulating component performance degradation, micro short-circuit disturbance factors for simulating abnormal leakage current paths, and convection attenuation disturbance factors for simulating deterioration of heat dissipation conditions.

[0080] The parallel disturbance simulation unit is used to superimpose abnormal physical factors as independent disturbance terms onto the corresponding parameters of the ideal electrothermal coupling reference model, perform parallel calculations within the same simulation time window, and output multiple sets of theoretical abnormal temperature curves containing specific fault morphology characteristics.

[0081] This embodiment features a refined design for the fault mode simulation module, aiming to generate a hypothesis space for negative samples. The fault factor library unit stores mapping rules that transform physical faults into mathematical parameters, including an aging disturbance factor to simulate the increase in internal resistance caused by battery aging, a micro-short circuit disturbance factor to simulate the parallel leakage current path formed by internal micro-short circuits, and a convection attenuation disturbance factor to simulate the decrease in heat dissipation capacity caused by flow channel blockage or fan dust accumulation. The parallel disturbance simulation unit injects the above factors into the baseline model to generate various theoretical abnormal temperature curves. For aging faults, the resistance term in the heat generation formula is corrected, and the corrected resistance is defined. for:

[0082]

[0083] To clarify the source of parameters and overcome the shortcomings of functional description, this embodiment defines... That is, the aging ratio and the battery The non-linear mapping function between health states is as follows:

[0084]

[0085] in, The preset impedance growth factor has a value of [value missing]. This value is based on the same model of battery. Accelerated aging cycle test data under charge / discharge rates, fitting the internal resistance growth curve and An empirical constant determined by the correlation of the decay trajectory; this formula quantifies the effect of... As the resistance decreases from 1.0, the internal resistance exhibits a non-linear increasing physical law, thus determining the specific... Numerical value;

[0086] For micro-short circuit faults, an additional short-circuit heat generation term is introduced, and the corrected heat generation rate is defined. for:

[0087]

[0088] in, For individual unit voltage; for That is, the short-circuit equivalent resistance. In order to transform the abstract dendrite growth model into executable simulation parameters, this embodiment establishes a model based on the severity level of dendrite growth. Logarithmic mapping rules from 0 to 3:

[0089]

[0090] Right now correspond For extremely low leakage current, correspond Corresponding to significant micro-short circuits, this allows for precise definition of internal short circuit heat sources of varying severity;

[0091] To address the issue of deteriorating heat dissipation, the heat transfer coefficient in the heat dissipation formula is modified, and the modified heat transfer coefficient is defined. for:

[0092]

[0093] against That is, the convection attenuation ratio. In this embodiment, a discretized lookup table is constructed based on the measured thermal resistance change rate recorded in the historical operation and maintenance database and the failure mode and effects analysis report of the heat dissipation system. After wind tunnel test determination, the filter dust accumulation condition is set to correspond to... Single fan failure condition Cooling pump shutdown condition The system directly calls the discrete values ​​from the lookup table mentioned above during simulation.

[0094] This embodiment enables the system to proactively imagine the temperature behavior when various faults occur through parameterized injection. This mechanism does not require historical fault data, but generates fault characteristics based on physical mechanisms, thereby solving the pain point of lacking negative sample data of thermal runaway in the energy storage industry, and enabling the system to have diagnostic capabilities immediately in newly built power plants that have never experienced faults.

[0095] Example 5:

[0096] The simulation result comparison and analysis module includes:

[0097] The measured data residual generation unit is used to perform point-by-point difference calculation between the real-time temperature data sequence of the sensor and the benchmark simulated temperature field to generate the first residual vector that eliminates the influence of the benchmark operating condition fluctuation. The first residual vector represents the mixed feature morphology of unknown anomalies and external disturbances.

[0098] The theoretical feature residual extraction unit is used to perform point-by-point difference calculation between each set of theoretical abnormal temperature curves and the benchmark simulation temperature field to generate multiple sets of second residual vectors that contain only specific fault morphology features. Each set of second residual vectors represents a pure preset fault thermal feature template.

[0099] This embodiment details the calculation logic of the simulation result comparison and analysis module, which performs dual-track difference calculation; the measured data residual generation unit performs the first-path difference operation to calculate the first residual vector, as shown in the following formula:

[0100]

[0101] in, The first residual vector is at The numerical value of a moment represents the deviation of the real world from the ideal world; The temperature sequence acquired in real time by the sensor mentioned in the aforementioned model data interface module is uniformly represented by the symbol in this formula. Indicates subscript Represents Sensor, derived from dynamic data interface; The baseline simulation temperature field sequence is derived from the baseline model.

[0102] The theoretical feature residual extraction unit performs the second-path difference operation to calculate multiple sets of second residual vectors, as shown in the following formula:

[0103]

[0104] in, :No. The second residual vector corresponding to each type of fault; Injection The theoretical abnormal temperature curve calculated after considering various fault factors

[0105] This embodiment transforms complex temperature analysis into a residual morphology analysis with significant characteristics through dual-track differential analysis. The first residual vector eliminates the large temperature rise caused by current fluctuations, retaining only the abnormal quantities, while the second residual vector extracts the pure thermal feature fingerprints of various faults. For example, the residual caused by micro-short circuits usually shows a continuous linear slow increase, while the residual caused by heat dissipation deterioration shows a high positive correlation with the current amplitude. This processing method greatly reduces the difficulty of subsequent pattern recognition.

[0106] Example 6:

[0107] The model validation and policy generation module includes:

[0108] The feature matching degree calculation unit is used to calculate the waveform matching degree between the first residual vector and each group of second residual vectors within a preset simulation analysis time window using the dynamic time warping algorithm or the Pearson correlation coefficient algorithm, so as to obtain the waveform similarity.

[0109] The fault mode classification unit is used to sort the waveforms according to their similarity and mark the preset fault type corresponding to the second residual vector with the highest matching degree as the inferred fault type for the first residual vector.

[0110] This embodiment describes the matching algorithm in the model validation and policy generation module; the feature matching degree calculation unit is configured to perform Z-Score standardization on the first residual vector and each set of second residual vectors before performing similarity calculation, as shown in the formula:

[0111]

[0112] in, To prevent the numerical stability constant from having a denominator of zero, this embodiment takes a value of [value missing]. ;

[0113] When using the Dynamic Time Warping (DTW) algorithm, the core steps are as follows: Construct a distance matrix and calculate the squared Euclidean distance between each point in the first residual vector and each point in the second residual vector; use dynamic programming to search for the optimal warping path. Specifically, Sakoe-Chiba band constraints are used to limit the deviation range of the search path, and the recursive step size mode is defined as follows. To ensure the monotonicity and continuity of the path, calculate the minimum cumulative distance. Convert the minimum cumulative distance into waveform similarity. Special dimension correction is needed here: Since the input vector has been Z-score standardized, i.e., the mean is 0 and the variance is 1, its distance metric has deviated from the original physical unit, i.e., degrees Celsius. Therefore, the original variance based on physical noise is no longer valid. The normalization formula will fail because in the normalization space... This no longer represents the level of physical noise, which can lead to extremely low or high similarity calculation results. This embodiment uses a normalized kernel function adapted to the dimensionless feature space.

[0114]

[0115] in, Cumulative path distance; The length of the regular path is the number of steps. The preset morphological tolerance factor is dimensionless and its value range is set to [value range missing]. to This range was determined through receiver operating characteristic (ROC) curve analysis of historical failure samples, where... The high sensitivity setting is used for detecting minor faults. For high specificity settings used to suppress false alarms, the recommended default value in this embodiment is [default value]. This parameter defines the average single-point Z-distance threshold allowed when two waveforms are judged to be similar within the normalized space.

[0116] This correction ensures that the algorithm focuses on the topological shape comparison of the waveform rather than the absolute amplitude, and solves the problem of algorithm failure caused by the mismatch of the normalization parameter dimensions after Z-Score standardization.

[0117] However, to address the logical flaws that Z-Score standardization may lead to, such as high impedance aging and the indistinguishability of micro-short circuits due to their similar shapes, this embodiment introduces a residual energy ratio criterion in the feature matching degree calculation unit. The specific logic is as follows:

[0118] Before performing Z-Score normalization, the system computes the first residual vector. With the second residual vector root mean square amplitude ratio :

[0119]

[0120] The fault mode classification unit is configured to perform dual verification: only if waveform similarity Greater than the preset threshold, and the amplitude ratio Within the preset physical reasonable range Only when the waveforms are highly similar within a short period of time is it determined to be this type of fault; if the waveforms are highly similar but... If so, it is determined to be a global internal resistance drift caused by aging, rather than a local micro-short circuit;

[0121] The fault mode classification unit sorts the calculated similarities and identifies the presumed fault type;

[0122] The model validation and policy generation module also includes:

[0123] The simulation confidence decision unit, used to make confidence decisions on the presumed fault type, is configured as follows:

[0124] If the waveform similarity is greater than or equal to the preset confidence threshold, it is determined that the first residual vector is highly consistent with the theoretical characteristics of the inferred fault type, and a high-confidence fault verification conclusion is generated.

[0125] If the waveform similarity is less than the confidence threshold, it is determined that the first residual vector cannot be effectively matched with any preset fault theory features, and a verification report prompting the inspection of model or data noise is generated.

[0126] This embodiment introduces a simulated confidence decision unit to construct a defensive false alarm filtering mechanism; this unit sets a confidence threshold. The threshold value typically ranges from 0.85 to 0.95; in this embodiment, 0.90 is selected as a typical value. Regarding waveform similarity... It is a dimensionless number and the sensor noise variance Dimensional physical quantities To address the logical flaws inherent in direct comparisons, this embodiment clarifies the mathematical conversion and calibration relationship between the two: numerical values Instead of directly assigning values ​​based on the noise variance, the system is calibrated using statistical benchmarks based on the noise variance of the on-site temperature sensors. The specific calibration process involves the system pre-collecting data for a period of time not less than [a certain duration]. Calculate the physical variance of a measured noise sequence that is 1 minute long and under static or constant current conditions. ;

[0127] The system constructs a virtual channel that generates pure Gaussian white noise, which has the same characteristics as... With the same physical intensity, the system inputs the pure noise sequence into the aforementioned Z-Score normalization and similarity calculation module to calculate the spurious similarity distribution between the pure noise sequence and each fault template; finally, the distribution is taken as the spurious similarity distribution. quantile values, that is, in At a confidence level, the upper limit of random similarity fluctuations caused by pure noise is used as the noise baseline, and... The value is set as a value with a certain safety margin, such as 1.2 times, above this baseline. Through this computational link of noise injection-normalization-similarity inversion, the system successfully established the noise variance in the physical space. The mapping relationship between the confidence threshold and the dimensionless feature space ensures... The value can dynamically adapt to the actual accuracy level of the sensor on site;

[0128] The decision logic is as follows: The system finds the maximum value among all similarities; in response to the maximum value being greater than or equal to... If the system determines the diagnosis as confirmed, it outputs a high-confidence report; otherwise, it responds when the maximum value is less than [a certain value]. The system determined it to be a mismatch;

[0129] The decision mechanism in this embodiment effectively filters out sensor noise or undefined external interference. If the actual temperature deviation is caused by sensor drift or random electromagnetic interference, its residual waveform is usually irregular noise and will not be highly correlated with any theoretical fault residual generated based on physical mechanisms. Therefore, the system will automatically ignore these non-fault anomalies, avoiding frequent false alarms in industrial sites with complex electromagnetic environments.

[0130] The model validation and policy generation module also includes:

[0131] The simulation optimization strategy generation unit, used to generate optimization suggestion reports for simulation models or control strategies based on fault verification conclusions, is configured as follows:

[0132] If the fault type is presumed to be a micro short circuit, the optimization recommendation report includes a suggestion to introduce the corresponding fault disturbance factor in subsequent simulations, as well as a simulation scenario description for testing the effectiveness of the external current regulation strategy.

[0133] If the fault type is presumed to be heat dissipation deterioration, the optimization recommendation report includes a recommendation to calibrate the local convective heat transfer coefficient, as well as a set of simulation boundary condition parameters for reconstructing and testing the control model of the external cooling system.

[0134] This embodiment describes the closed-loop feedback mechanism of the simulation optimization strategy generation unit; in response to the presumed fault type being a micro-short circuit, the system performs micro-short circuit parameter identification; in order to overcome the direct optimization of resistance in the original physical model... Potential numerical instability, i.e., when The hourly heat term tends to infinity, and The fact that the variable is in the denominator position results in extremely strong gradient nonlinearity. In this embodiment, at the algorithm implementation level, the optimization variable is transformed into short-circuit conductance. This transformation simultaneously converts the nonlinear optimization problem concerning resistance into a quasi-linear problem concerning conductance, ensuring the smoothness of the objective function gradient; the system is constructed based on... Least square objective function for variables :

[0135]

[0136] in, To inject into the baseline model The simulated temperature sequence generated after the perturbation term; the calculation of the Jacobian matrix is ​​also adjusted accordingly to be sensitive to conductivity:

[0137]

[0138] By solving the optimal The system can reliably identify even the slightest leakage current. From extremely small to severe short circuits The large range of fault conditions prevents the optimization process from diverging; in response to the presumed fault type of heat dissipation degradation, the system targets the proportion of convection attenuation. The model is inverted and the calibrated parameters are written into the static configuration unit to achieve adaptive evolution of the model.

[0139] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A thermal energy management optimization system for a lithium battery energy storage platform, characterized in that, The system serves as the upper-level simulation and verification platform for the lithium battery energy storage platform, and includes: The model data interface module is used to obtain static physical parameters and dynamic time-series data of the lithium battery energy storage platform from an external battery management system or to be configured by the user. The dynamic time-series data includes real-time current, cell voltage and sensor real-time temperature. The benchmark simulation model construction module is used to construct an ideal electrothermal coupling benchmark model based on the static physical parameters, and drive the model based on the dynamic time series data to simulate and calculate the benchmark simulation temperature field of the lithium battery energy storage platform under fault-free operating conditions. The fault mode simulation module is used to inject predefined abnormal physical factor perturbations corresponding to multiple fault types into the ideal electrothermal coupling reference model in parallel and perform simulations to generate multiple sets of theoretical abnormal temperature curves with different fault characteristics. The simulation result comparison and analysis module is used to calculate the first residual vector between the sequence of real-time temperature data from the sensor and the reference simulated temperature field, and the second residual vector between each set of theoretical abnormal temperature curves and the reference simulated temperature field. The model verification and strategy generation module is used to calculate the waveform similarity between the first residual vector and each group of second residual vectors, and classify the shape of the first residual vector according to the waveform similarity, and finally output a simulation verification report describing the suspected fault type and the corresponding strategy.

2. The thermal energy management optimization system for a lithium battery energy storage platform according to claim 1, characterized in that, The model data interface module includes: The static parameter configuration unit is used to receive or extract the three-dimensional geometric topology data, material thermal property parameters and internal flow channel model data of the simulated object, and configure the static physical parameters as the initial boundary conditions of the ideal electrothermal coupling reference model. The dynamic data excitation unit is used to receive external real-time synchronous data streams, which include load current, environmental parameters and cooling medium flow rate, and load the dynamic time-series data as a time-varying excitation source to the corresponding input port of the ideal electrothermal coupling reference model.

3. The thermal energy management optimization system for a lithium battery energy storage platform according to claim 1, characterized in that, The benchmark simulation model construction module includes: The multiphysics modeling unit is used to construct the ideal electrothermal coupling benchmark model in the simulation environment based on the electrochemical student heat equation and the lumped parameter thermal network theory. The model is configured to only respond to the simulation of normal temperature rise caused by standard load fluctuations. The pure field simulation calculation unit is used to input the dynamic time series data into the ideal electrothermal coupling reference model for numerical calculation, and output the theoretical temperature field sequence stripped of preset aging factors and internal short circuit factors as the reference simulation temperature field.

4. The lithium battery energy storage platform thermal energy management optimization system according to claim 1, characterized in that, The fault mode simulation module includes: The fault factor library unit is used to store a variety of predefined abnormal physical factors, including aging disturbance factors for simulating component performance degradation, micro short-circuit disturbance factors for simulating abnormal leakage current paths, and convection attenuation disturbance factors for simulating deterioration of heat dissipation conditions. The parallel disturbance simulation unit is used to superimpose the abnormal physical factors as independent disturbance terms onto the corresponding parameters of the ideal electrothermal coupling reference model, perform parallel calculations within the same simulation time window, and output multiple sets of theoretical abnormal temperature curves containing specific fault morphology characteristics.

5. The thermal energy management optimization system for a lithium battery energy storage platform according to claim 1, characterized in that, The simulation result comparison and analysis module includes: The measured data residual generation unit is used to perform point-by-point difference calculation between the real-time temperature data sequence of the sensor and the reference simulated temperature field to generate the first residual vector that eliminates the influence of the reference operating condition fluctuation. The first residual vector represents the mixed feature morphology of unknown anomalies and external interference. The theoretical feature residual extraction unit is used to perform point-by-point difference calculation between each set of theoretical abnormal temperature curves and the benchmark simulation temperature field to generate multiple sets of second residual vectors that contain only specific fault morphology features. Each set of second residual vectors represents a pure preset fault thermal feature template.

6. The thermal energy management optimization system for a lithium battery energy storage platform according to claim 1, characterized in that, The model validation and policy generation module includes: The feature matching degree calculation unit is used to calculate the waveform matching degree between the first residual vector and each group of second residual vectors within a preset simulation analysis time window using a dynamic time warping algorithm or a Pearson correlation coefficient algorithm, so as to obtain the waveform similarity. The fault mode classification unit is used to sort the waveforms according to their similarity and mark the preset fault type corresponding to the second residual vector with the highest matching degree as the inferred fault type for the first residual vector.

7. The lithium battery energy storage platform thermal energy management optimization system according to claim 6, characterized in that, The model validation and policy generation module also includes: The simulation confidence decision unit, used to make a confidence decision on the presumed fault type, is configured as follows: If the waveform similarity is greater than or equal to a preset confidence threshold, it is determined that the first residual vector is highly consistent with the theoretical characteristics of the presumed fault type, and a high-confidence fault verification conclusion is generated. If the waveform similarity is less than the confidence threshold, it is determined that the first residual vector cannot be effectively matched with any preset fault theory features, and a verification report prompting the inspection of model or data noise is generated.

8. The thermal energy management optimization system for a lithium battery energy storage platform according to claim 7, characterized in that, The model validation and policy generation module also includes: The simulation optimization strategy generation unit, configured to generate an optimization suggestion report for the simulation model or control strategy based on the fault verification conclusions, is as follows: If the fault type is presumed to be a micro short circuit, the optimization suggestion report includes a suggestion to introduce the corresponding fault disturbance factor in subsequent simulations, as well as a simulation scenario description for testing the effectiveness of the external current regulation strategy. If the fault type is presumed to be heat dissipation deterioration, the optimization recommendation report includes a recommendation to calibrate the local convective heat transfer coefficient, as well as a set of simulation boundary condition parameters for reconstructing and testing the control model of the external cooling system.

Images