A method and system for multi-point wireless temperature measurement and liquid cooling control of energy storage battery surface
By setting wireless temperature measurement tags and temperature prediction models on the surface of energy storage batteries, the problems of uneven hotspot distribution and lag in liquid cooling regulation in energy storage systems are solved. This enables accurate detection of battery surface temperature and efficient regulation of the liquid cooling system, improving the predictability and energy efficiency of thermal management.
Patent Information
- Application Number
- CN202611402639.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-10
- Publication Date
- 2026-11-17
AI Technical Summary
Existing energy storage systems suffer from uneven distribution of hot spots on the battery surface and difficulty in accurate detection when storing high-power lithium batteries. Liquid cooling regulation is also lagging, resulting in low thermal management efficiency and a large workload for space utilization and maintenance.
By using wireless temperature measurement tags for multi-point temperature detection, combined with a tag location mapping table and a temperature prediction model, accurate monitoring of battery surface temperature and dynamic control of the liquid cooling system can be achieved. Liquid cooling control commands are generated through thermal risk ranking and rolling optimization, and adjustments are made in advance to predict and respond to the formation of hot spots.
It enables precise detection of battery surface temperature and efficient allocation of liquid cooling resources, reduces assembly complexity and maintenance workload, avoids pre-heating adjustments, and improves the predictability and energy efficiency of thermal management.
Smart Images

Figure CN123120948A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of thermal management and safety control of energy storage systems, and in particular to a method and system for multi-point wireless temperature measurement and liquid cooling combined control of energy storage battery surface. Background Technology
[0002] When high-power lithium-ion battery energy storage systems are used for primary frequency regulation, secondary frequency regulation, and rapid power support, the charging and discharging currents change drastically. Internal resistance heating, contact heating of the tabs and connectors, and differences in heat transfer from the liquid cooling plate all act simultaneously on the battery surface, resulting in randomly distributed hotspots in varying locations. Existing energy storage systems primarily use wired NTC temperature measurement. If the temperature measurement points are expanded to multiple surface locations of each battery, taking a 22-cell cluster with 400 cells per cluster and 4 temperature measurement points per cell as an example, 35,200 temperature measurement points would be required. The corresponding sensor leads, acquisition channels, and connectors would severely occupy the internal space of the battery rack and significantly increase the workload of assembly, insulation, and maintenance.
[0003] At the control level, current liquid cooling regulation mainly relies on parameters reflecting the average thermal state of the region, such as the main pipe return water temperature, cluster outlet water temperature, or the temperature of a small number of representative batteries. This makes it impossible to directly determine the location of the hot spot on the battery surface and its corresponding liquid cooling branch. Furthermore, the refrigeration unit load rate and the branch regulating valves have inherent action delays in response to battery temperature; threshold control based solely on the current measured temperature often triggers the regulation command only after the hot spot has formed. Therefore, establishing a precise mapping between hot spot locations and liquid cooling actuators, and realizing executable predictive joint regulation based on this mapping, is a key technical problem that needs to be solved in the field of high-power energy storage battery thermal management. Summary of the Invention
[0004] In view of this, this application provides a method and system for multi-point wireless temperature measurement and liquid cooling combined control of the surface of an energy storage battery, which solves the problems mentioned in the background art.
[0005] In a first aspect, this application discloses a method for combined multi-point wireless temperature measurement and liquid cooling control of the surface of an energy storage battery, including:
[0006] Wireless temperature measurement tags are set at multiple preset locations on the surface of each battery, and a tag location mapping table is established. The mapping table includes at least: tag code, battery information, and corresponding liquid cooling branch information. The temperature data of each wireless temperature measurement tag is read in a polling manner through a reader, the temperature data is time-aligned, and each tag is associated with the corresponding liquid cooling branch according to the location mapping table. The operating parameters of the battery system are obtained, the battery surface temperature is predicted through a temperature prediction model, and a temperature safety upper bound is obtained based on the temperature measurement uncertainty, tag data age, and prediction residual. The temperature margin of each battery is calculated based on the temperature safety upper bound, and the battery clusters are ranked by thermal risk. For battery clusters with high thermal risk, control commands for the liquid cooling system are generated in a rolling manner under the condition of meeting the temperature safety upper bound, and the flow rate of the corresponding liquid cooling branch is adjusted. If the predicted temperature is still higher than the temperature safety upper bound after adjustment, a power limiting request is output to the energy management system.
[0007] Secondly, this application discloses a multi-point wireless temperature measurement and liquid cooling combined control system for the surface of an energy storage battery, comprising:
[0008] Wireless temperature measurement tags, each with a unique code, are placed at multiple preset locations on the surface of each battery to measure the temperature at their location and wirelessly transmit the temperature data. A reader and antenna are used to poll and read the temperature data from tags within each reading area. An edge temperature measurement gateway receives tag data uploaded by each reader, performs time alignment, data age determination, and quality marking. A thermal management controller establishes a tag location mapping table, predicts temperature using a temperature prediction model to obtain a safe upper limit for temperature, prioritizes thermal risks based on the remaining temperature margin, and generates control commands for the liquid cooling system. The liquid cooling system includes at least a chiller unit, a circulating pump, main supply and return water pipes, a liquid cooling branch for the battery cluster, and regulating valves, and receives and executes the control commands. A communication interface is used to output power limiting requests to the energy management system.
[0009] Compared with existing technologies, the multi-point wireless temperature measurement and liquid cooling combined control method and system for the surface of energy storage batteries disclosed in this application have the following advantages:
[0010] 1. This application uses wireless temperature measurement tags to replace traditional wired NTC sensors. The unique code of the tag enables precise association between temperature data and the location of battery clusters, modules, cells, and surfaces. The reader can read the temperature data of tens of thousands of tags in batches by polling the partitions. Therefore, there is no need to lay out independent signal lines, acquisition channels and connectors for each supplementary measurement point. This effectively avoids the occupation of internal space of the battery rack by tens of thousands of sensor leads, and significantly reduces assembly complexity, insulation risk and subsequent maintenance workload.
[0011] 2. This application uses a tag location mapping table to associate the unique code of the tag with the battery location and the corresponding liquid cooling branch and regulating valve, and establishes an association matrix to limit the temperature prediction range to nodes with physical or liquid cooling association. Therefore, the detected hot spot location can be directly converted into a specific liquid cooling branch regulation object, realizing the on-demand allocation of cooling resources and avoiding the energy waste of traditional solutions where some areas are overcooled and hot spots still do not receive enough cooling.
[0012] 3. This application constructs a temperature prediction model that integrates physical mechanisms and data-driven approaches. It uses historical data to predict short-term temperature changes and incorporates data age, temperature measurement uncertainty, and prediction residuals into the upper limit of temperature safety. Therefore, it can overcome the time lag of temperature response and achieve early adjustment before hot spots form. This avoids the passive situation of traditional threshold control that only takes effect after the temperature exceeds the limit. When the liquid cooling regulation capacity is insufficient, it further outputs the available power boundary or power reduction request to the EMS, realizing system-level thermal safety coordination to reduce heat generation from the source. Attached Figure Description
[0013] Figure 1 This is a schematic diagram of the overall structure of the battery surface temperature measurement and liquid cooling combined control system.
[0014] Figure 2 A schematic diagram showing the arrangement of wireless temperature measurement tags on the surface of a single battery;
[0015] Figure 3 Flowcharts are constructed for wireless tag time alignment, data quality processing, and temperature safety upper bounds.
[0016] Figure 4 This is a diagram of the temperature prediction model structure.
[0017] Figure 5 Here is a flowchart of the combined control of liquid cooling and rolling.
[0018] Figure 6 This is a flowchart of the project implementation steps.
[0019] Explanation of reference numerals in the attached diagram: 20, UHF RFID reader and antenna; 201, UHF RFID temperature measurement tag; 21, edge temperature measurement gateway; 22, thermal management controller; 3, liquid cooling system; 31, refrigeration unit; 32, circulating pump; 33, regulating valve. Detailed Implementation
[0020] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.
[0021] This application uses a 5MW / 2.5MWh high-power lithium-ion battery energy storage system as an example to illustrate the implementation process. The system includes 22 battery clusters, each with 400 lithium iron phosphate batteries connected in series. The cell model, single-cell capacity, rated rate, DC internal resistance range, chiller model, pump and valve parameters are recorded in the prototype equipment configuration table and serve as the basis for model parameter identification and control boundary setting. Firstly, this application provides a method for multi-point wireless temperature measurement and liquid cooling combined control of the energy storage battery surface, which includes at least:
[0022] Step S1: Tag Arrangement and Position Mapping. Wireless temperature measurement tags (UHF RFID tags) are placed at multiple preset locations on the surface of each battery. At least two UHF RFID tags are placed on each battery surface, and each tag has a unique code. In this embodiment, four UHF RFID tags are placed on the surface of each battery, with preset locations near the positive electrode tab, near the negative electrode tab, in the center of the battery surface, and at locations with weak heat exchange from the liquid cooling plate. This creates 35,200 tag measurement points. The model, protocol, passive or semi-passive type, temperature range, nominal accuracy, temperature transition time, anti-metal structure, insulating layer, and adhesive material of the selected UHF RFID tags are recorded in the tag configuration table.
[0023] Before installation, at least two-point calibration is performed using a patch NTC or thermocouple, and the steady-state error and dynamic response time are recorded in the tag calibration table. The tag adopts a structure suitable for metal surfaces, and the isolation layer, adhesive material, and installation direction between the tag and the battery casing are determined according to the tag's technical requirements. During installation, a tag position mapping table is established to convert hotspot locations into liquid-cooled branch adjustment objects. The mapping table includes at least: tag code, battery information, and corresponding liquid-cooled branch information. Battery information includes at least: battery cluster number, battery module number, battery number, and surface position number. Liquid-cooled branch information includes at least: liquid-cooled branch number, cluster-level regulating valve number, and module-level regulating valve number.
[0024] Step S2: Reading Partitions, Time Alignment, and Data Quality Processing. Each battery cluster is divided into several reading areas based on the battery rack structure, with at least one antenna configured in each area. The number of reading areas, the number of tags per area, and the polling cycle are determined based on the tag temperature transition time, the number of reader channels, and the effective read rate under full tag load conditions. The number of antennas, reader channels, and switching methods for each area are determined through a full-load reading test. The test is conducted after the batteries, busbars, liquid cooling plates, and the metal structure of the enclosure are installed. The number of tags per area, the complete scan time, the effective read rate, the continuous missed read rate, and the worst-case location are recorded. If a certain area fails to achieve the effective read rate specified in the project, an antenna is added, the reading area is reduced, or the tag installation direction is adjusted.
[0025] The reader polls and reads the temperature data from each wireless temperature measurement tag. Since the actual reading time of each tag differs within a single scan, the temperature data needs to be time-aligned. This involves: first, obtaining the start time of the current control cycle as a unified time reference; then, calculating the time difference between the reading time of each wireless temperature measurement tag and the unified time reference; and finally, correcting the reading temperature of each tag to the equivalent temperature value under the unified time reference using the following formula: ,in, For the i-th label at control time The equivalent temperature value, For the i-th tag at the reading time Reading temperature The temperature rise rate is estimated for the most recent valid reading of the i-th tag. When the time difference exceeds the set allowable value, the tag is marked as missing, and the corresponding temperature data is not used in subsequent control.
[0026] After time alignment of the temperature data, the method further includes quality assessment of the tag temperature data, specifically, obtaining the communication quality of each tag. The communication quality is determined based on the data age, the number of successful reads in the current period, the quality of the read signal, the number of consecutive missed read periods, and the dispersion of multiple reads.
[0027] Obtain the temperature measurement uncertainty of each label The temperature measurement uncertainty is determined based on tag calibration error, installation bonding thermal resistance, and dynamic response error. It should be noted that communication quality and temperature measurement uncertainty are not combined into the same confidence level. The temperature measurement uncertainty is obtained through the following steps: at at least two calibration points within the allowable temperature range of the selected tag, the tag temperature is compared with the reference temperature of the patch NTC or thermocouple to obtain the steady-state error and response time under step temperature changes. The specific calibration points, allowable errors, and response times are determined by the selected tag model and project technical requirements. After time alignment and quality assessment, the temperature data of each tag, together with the corresponding missing mask, constitutes the battery surface temperature matrix, which is used as input for subsequent temperature prediction models.
[0028] After completing the data quality assessment, an association matrix A is established between temperature measurement nodes based on the tag location mapping table. When two tags are located on the same battery surface, adjacent batteries, upstream or downstream of the same liquid cooling branch, or affected by the same regulating valve, the corresponding matrix elements... ,otherwise The purpose of this correlation matrix is to limit the range of nodes participating in the weighting process during temperature prediction. After identifying hotspots, the controller prioritizes using the tag states that have in-plane heat conduction, spatial adjacency, or liquid cooling association with the hotspots, and determines the corresponding liquid cooling branches and regulating valves through a mapping table, without performing indiscriminate fully connected calculations on all tags.
[0029] In this embodiment, if the prototype is not equipped with a module-level valve or a branch flow sensor, the corresponding control quantity is removed from the candidate set. The branch flow can be estimated based on valve characteristics, differential pressure, and pump curve, and the estimation error is recorded.
[0030] Step S3: Temperature Prediction. First, the operating parameters of the battery system are obtained. The surface temperature of the battery is predicted using a temperature prediction model, and a safe upper bound for temperature is obtained based on the measurement uncertainty, tag data age, and prediction residuals. The operating parameters include at least: the charging and discharging current of the battery cluster, the battery cluster voltage, the ambient temperature, the outlet water temperature of the liquid cooler, the actual valve opening and branch flow rate, as well as SOC, SOH, and the circulation pump speed. The temperature prediction model includes: a physical temperature rise calculation layer, a cluster-level local storage pool, a system-level global storage pool, a topology-constrained attention layer, and a shared readout layer.
[0031] The physical temperature rise calculation layer calculates the base temperature rise using battery current, internal resistance, coolant temperature, and branch flow rate. The base temperature at the i-th tag location is: ,in, The temperature at the current moment. For time step, This represents the equivalent heat capacity at that location. This represents the amount of heat allocated to this location at the current moment. For the flow rate of the branch, This is the equivalent heat transfer coefficient. Battery heat generation includes at least irreversible heat: ;in, The DC internal resistance varies with SOC, SOH, and temperature. Additional heat generated by the tabs, connecting pieces, and contact resistance. This is a coefficient for distributing the total heat generation to the label location. If the cell data and test conditions permit, a reversible entropy term can also be added. , Contact thermal resistance and heat source distribution coefficient are identified through thermal balance tests and small-scale excitation of valves, flow rates, and power within a safe range.
[0032] The reservoir calculation layer includes a local reservoir for each battery cluster and a system-level global reservoir. The local reservoir describes the dynamic relationship between cluster temperature, actual valve position, branch flow rate, and cluster current. The global reservoir describes the dynamic relationship between the chiller, circulating pump, outdoor temperature, and inter-cluster liquid flow coupling. The reservoir uses leakage integrals for state updates, as detailed below:
[0033] For a local reservoir, candidate states are first calculated based on the current input vector and the reservoir state from the previous time step: ;in, This represents a candidate state for a local reserve pool. The local input vector includes module-level temperature statistics of the c-th cluster, tag temperature rise rate, valve command and actual position, branch flow, cluster current, SOC and SOH; This reflects the state of the reserve pool at the previous moment. and These are the input weight matrix and the internal connection weight matrix, respectively. This is the bias term. Then, the candidate state is weighted and fused with the previous state using the leakage integral to obtain the updated state for the current time step: ;in, The leakage rate of the local reserve pool is used to control the retention rate of historical information.
[0034] Similarly, the candidate states of the global reserve pool are: ;in, This represents a candidate state for the global reserve pool. The global input vector includes the highest temperature and temperature difference of each cluster, the outlet and return water temperatures of the chiller unit, the circulation pump speed, the supply and return water pressure difference, the outdoor temperature, and the total power. This refers to the state of the global reserve pool at the previous moment. and These are the input weight matrix and the internal connection weight matrix, respectively. This is a bias term. The global reserve pool is also updated using leaked integrals: ;in, This represents the leakage rate of the global reserve pool.
[0035] The internal connection matrix of the reservoir is a sparse random matrix, scaled according to the target spectral radius. Input weights and internal connection weights remain unchanged after initialization; topology constraint attention parameters and shared readout layer parameters are adjusted during field updates. The reservoir sparsity, spectral radius, number of nodes, and leakage rate are determined by the validation error of continuously running data, without pre-setting unverified fixed values.
[0036] A topology-constrained attention layer calculates weights for associated nodes and several recent historical states. Within the node range defined by the association matrix, based on the spatial location in the label location mapping table and the liquid cooling flow direction, the attention weights of each associated node's state at multiple historical moments to the target label are calculated, first constructing the label node state. The state of the i-th label node is formed by concatenating the label's local reservoir state, global reservoir state, and the label's own state: ;in, Let be the state vector of the i-th label node at time k. This refers to the local reserve pool state of the cluster described by the label. This is the global reserve pool status. This is the label's own state vector. Further includes: ;in, The temperature value of the i-th label after time alignment. To estimate the rate of temperature change, Encode the label position. For communication quality, For temperature measurement uncertainty, This is the missing mask. Based on this, the attention weights of the associated nodes to the target label are calculated. The weight of node j to target label i at historical time kl is denoted as:
[0037] ;in, This is the intermediate relevance vector. , , The weight matrix is a learnable matrix. Let kl be the state of node j at a historical moment. The node relationship characteristics include at least one of the following: distance between nodes, whether they belong to the same battery, upstream and downstream direction of liquid cooling, branch number, and common valve number.
[0038] When there is a relationship between two tag nodes, that is... At that time, the intermediate relevance vector is compressed into a scalar attention score through activation function and linear transformation: ,in, This is a learnable attention weight vector. It is the hyperbolic tangent activation function.
[0039] When there is no association between two tag nodes, that is At that time, set its attention score to negative infinity: Then, the attention scores of all associated node-historical time-pairs are normalized to obtain the attention weights: ;in, The normalized attention weights represent the importance of the state of neighbor node j at historical time k to the target label i. Finally, the states of each associated node at each historical time are weighted and summed according to the attention weights to form the context vector of the target label: ;in, This is the context vector of the i-th label node, which gathers valid information from all associated neighbor nodes at multiple historical moments for further processing by the shared readout layer.
[0040] The shared readout layer, to address the mapping issue between module-level features and tag-by-tag output, receives the current temperature, location code, communication quality, temperature measurement uncertainty, missing mask, local reservoir state, global reservoir state, and context vector for each tag. The readout features are defined as follows: ; mapped to the prediction residual through a readout function shared by all labels. The predicted residuals are then superimposed onto the base temperature output by the physical temperature rise calculation layer, and the final predicted temperature is output label by label. ;in, Encode the label position. For missing masks, all tags share the same read function. This structure uses the same set of readout parameters to perform batch calculations on the tags, outputting the future temperature for each tag, without configuring an independent readout matrix for each tag.
[0041] A reservoir with fixed internal connection weights retains the historical state of the thermal process, and adjusts the attention and shared readout layer parameters during field updates; topology-constrained attention only aggregates the state of nodes that have a location or liquid cooling association with the target label.
[0042] In this embodiment, continuous verification data are grouped according to operating conditions and communication quality, and the set quantile of the under-prediction residual for each group of temperature predictions is calculated as the basic prediction safety margin. The quantile can be the 95th quantile or other quantiles determined by project safety requirements. When a tag experiences consecutive missed reads, the missed read compensation is increased based on the maximum temperature rise rate calibrated on-site and the data age. The upper bound for the temperature safety of the i-th tag at a future time is:
[0043] ;in, To predict temperature, To predict residual margin, For temperature measurement uncertainty, The maximum temperature rise rate at the i-th tag position. For data age, data age This is the time difference between the last successful read of the tag and the current time.
[0044] In this embodiment, temperature control employs a three-level threshold system: the upper limit of control is denoted as T. ctl The warning threshold is denoted as T. warn The BMS shutdown threshold is denoted as T. stop All three conditions must be met: T ctl <T warn <T stopThe control upper limit is a target constraint used by the thermal management controller for rolling regulation; the early warning threshold is used to trigger a power limiting request to the EMS; and the BMS shutdown threshold is independently determined and executed by the BMS's original independent hard-wired protection channel. The UHF RFID temperature measurement and thermal management controller is only used for early warning and fine regulation; the final safety shutdown protection is executed by the BMS according to its independent safety logic.
[0045] Step S4: Calculate the temperature margin of each battery based on the temperature safety upper bound and sort the battery clusters by thermal risk.
[0046] The temperature margin for the i-th label is defined as the difference between the upper limit of the temperature safety range and the current temperature. ;in, To control the upper limit, Let i be the upper bound of the temperature safety for the i-th label at future moments. A smaller value indicates the tag is closer to the control limit, and a higher thermal risk. Since each battery surface has multiple temperature measurement tags, the system has tens of thousands of tag measurement points. The tag-level temperature margin needs to be aggregated into a cluster-level indicator. The temperature margin of a battery cluster is calculated from all tags in that cluster. The minimum value is used to reflect the worst local thermal state within the cluster, avoiding the obscuring of local hot spots by the average temperature.
[0047] The battery clusters are sorted in ascending order of their temperature margin. The smaller the temperature margin, the higher the thermal risk of the battery cluster. The battery clusters ranked higher are given priority for control in subsequent rolling control.
[0048] Step S5: Rolling Liquid Cooling Joint Control. In each control cycle, the battery clusters and modules with the highest thermal risk are selected based on the remaining temperature margin. Only the liquid cooling control valves for these objects generate candidate actions, while the other control valves maintain their current opening. An upper limit is set for the number of hot spot clusters. The parameters are determined by the controller's computing power test and stored as control parameters.
[0049] The control quantity u(k) refers to the current commands for the valves, circulating pumps, and refrigeration units; candidate control sequences. Refers to a set of control variables in the prediction time domain: .
[0050] Under the condition of satisfying the upper bound constraint on temperature safety, a rolling time-domain optimization function is constructed. The upper bound constraint on temperature safety is as follows: .
[0051] The optimization function is obtained by weighted summation of safety terms, temperature uniformity terms, auxiliary power consumption, and actuator action quantities.
[0052] The formula for calculating the safety item is as follows: The term is used to characterize the risk of predicted temperatures exceeding the upper limit of the temperature safety threshold. The formula for calculating the mean temperature term is: It is used to characterize the temperature uniformity within the battery cluster, where The intra-plane temperature difference is the difference between the highest and lowest temperatures on the label of the same battery surface, used to determine local unevenness within a single cell. This is the maximum temperature difference between clusters, which is the maximum difference between the highest labeled temperatures of each battery cluster, used to determine unevenness between clusters. These are the weighting coefficients for each item. Auxiliary power consumption. The actuator motion is used to characterize the energy consumption of the liquid cooling system. This is a penalty term for the actuator's motion quantity.
[0053] Each actuator's candidate actions are selected from a finite set of actions consisting of maintaining the current opening, increasing by one step, and decreasing by one step. In the prediction time domain, the optimal combination of valve opening and pump speed is searched from the finite set of actions to achieve the optimal function. The candidate sequence is expanded layer by layer using the finite set of actions, with each layer retaining only the B best-performing sequences. This process is repeated until the maximum number of candidates, N, is reached. max Or maximum computation time T max Stop the search and execute the currently evaluated optimal sequence. B, N max and T max The time to inference and the control cycle are determined based on the controller hardware and recorded in the control parameter table.
[0054] The model input includes both control commands and actual actuator feedback, and performs multi-step predictions based on the valve action delay, flow transmission delay, and chiller unit response delay identified on-site. For candidate sequence s, the... The local and global reserve pools are updated using candidate control variables and execution latency:
[0055] ;
[0056] ;
[0057] Recalculate the shared readout characteristics:
[0058] ;
[0059] in, For the execution delay vector of valves, pumps and refrigeration units, the future current is determined according to the power command or power limit value, and the future flow rate is calculated according to the candidate valve position, pump speed and branch pressure difference.
[0060] Only the first control variable of the optimal candidate control sequence is executed in each control cycle. Before executing the optimal control command, an action check is set to limit the single-cycle variation of valve opening, pump speed, chiller unit outlet water temperature setpoint, and minimum hold time. After the check passes, the optimal control command for the current moment is executed, and the next cycle is recalculated based on the new measurement results. Different thresholds are used for entering and exiting the warning state, forming hysteresis control to prevent frequent switching between the control valve and the chiller unit in adjacent cycles.
[0061] Step S6: Handling insufficient cooling capacity. If the predicted temperature is still higher than the upper limit of the temperature safety range after adjustment, a power limiting request is sent to the energy management system.
[0062] The refrigeration unit can establish a mapping between outdoor temperature, return water temperature and refrigeration capacity based on the manufacturer's performance table or on-site calibration data, without using an uncalibrated fixed linear derating factor.
[0063] If the predicted temperature remains above the safety upper limit after rolling adjustment, the following steps are executed sequentially: cancel the energy-saving operation mode of the liquid cooling system, increase the circulation pump speed and the valve opening of the branch corresponding to the hot spot, output the available power boundary of the high-risk battery cluster to the energy management system, and if the temperature safety upper limit constraint is still not met, output a system-wide power reduction request. When the BMS protection condition is reached, the BMS will execute cluster switching or shutdown.
[0064] For systems configured with independent bidirectional DC / DC converters, independent PCS channels, or other power conversion branches with closed-loop adjustable cluster current, the EMS performs inter-cluster power distribution based on the temperature margin, SOC, SOH, and cluster current limits of each cluster. For systems where battery clusters are directly connected in parallel to the same DC bus and the cluster current cannot be independently adjusted, the thermal management controller outputs the cluster available power boundary, requests to exit high-risk clusters, or requests for overall system power reduction.
[0065] It should be noted that in this invention, UHF RFID temperature measurement is only used to increase the coverage of battery surface temperatures and provide early warning information; it does not replace the original hardwired temperature, voltage, current, and insulation protection channels of the BMS. When the UHF RFID detects an abnormal temperature, the thermal management controller issues an early warning, liquid cooling adjustment, or power limiting request. The final protection action, such as cluster switching or shutdown, is executed by the BMS according to its independent safety logic. The BMS continues to perform current limiting, cluster switching, or shutdown protection according to its own safety logic, and its hardwired protection channels are independent of the thermal management control system of this invention.
[0066] In this embodiment, the above-mentioned thermal management control method undergoes the following deployment and training steps before being put into online operation:
[0067] Determine the number and location of UHF RFID tags based on the selected battery structure, and complete tag temperature calibration, anti-metal installation, and location mapping documentation.
[0068] Each battery cluster was divided into several reading areas, and readers and antennas were arranged. Under the condition that full tags, busbars and liquid cooling plates were installed, the area scanning time, effective reading rate, continuous missed reading rate and worst reading position were tested, and the polling cycle was determined based on the full load reading test results.
[0069] The system integrates the actual valve opening, circulating pump speed, branch flow rate, supply and return water temperature, pressure difference, cluster current, SOC, SOH, and outdoor temperature, and establishes a unified timestamp.
[0070] Within a safe range, small excitations are applied to valve opening, pump speed, chiller setpoint, and power to identify action delay, thermal time constant, internal resistance heating parameters, and heat transfer parameters. Closed-loop normal operation data and independent excitation data are used together for model training.
[0071] During system commissioning, with battery temperature and current both within safe ranges, the setpoints for one liquid-cooled branch valve, circulating pump speed, or chiller outlet water temperature were changed, and the execution commands, actual feedback, and tag temperature responses were recorded. Power excitation was implemented according to frequency modulation test conditions. Normal closed-loop operation data and the aforementioned independent excitation data were used together to identify execution delay and thermal parameters.
[0072] The training set, validation set, and test set are divided according to the complete continuous operating interval; the parameters of the reservoir, attention, and shared readout layer are determined; and the safety margin is predicted based on the grouping and statistical analysis of the validation residuals.
[0073] Test the single inference time, number of candidate sequences, and complete solution time on the controller hardware, and limit the number of hotspot clusters and candidate actions according to the control cycle.
[0074] After completing the above deployment and training, the high-power containerized energy storage battery UHF RFID surface multi-point temperature measurement and liquid cooling joint control can be performed online according to steps S1 to S6.
[0075] This invention also provides a multi-point wireless temperature measurement and liquid cooling combined control system for the surface of an energy storage battery, used to perform the above-described method. The system includes a UHF RFID temperature measurement tag 201, a reader and antenna 20, an edge temperature measurement gateway 21, a thermal management controller 22, a liquid cooling system 3, a BMS, and an EMS. The liquid cooling system 3 includes at least a refrigeration unit 31, a circulating pump 32, main supply and return water pipes, and liquid cooling branches for the battery clusters; depending on the equipment configuration, it may also include cluster-level regulating valves 33, module-level regulating valves 33, branch flow sensors, and supply and return water differential pressure sensors.
[0076] The functions of each component correspond to the above method steps, and for specific implementation methods, please refer to the aforementioned method embodiments.
[0077] It should be noted that the above content merely illustrates the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. For those skilled in the art, various improvements and modifications can be made without departing from the principle of the present invention, and all such improvements and modifications fall within the scope of protection of the claims of the present invention.
Claims
1. A method for combined control of multi-point wireless temperature measurement and liquid cooling on the surface of an energy storage battery, characterized in that, include: Wireless temperature measurement tags are set at multiple preset positions on the surface of each battery, and a tag position mapping table is established. The mapping table includes at least: tag code, battery information and corresponding liquid cooling branch information. The temperature data of each wireless temperature measurement tag is read by polling the reader, the temperature data is time-aligned, and each tag is associated with the corresponding liquid cooling branch according to the location mapping table. The operating parameters of the battery system are obtained, the surface temperature of the battery is predicted by the temperature prediction model, and the upper limit of temperature safety is obtained based on the temperature measurement uncertainty, the age of the tag data and the prediction residual. The temperature margin of each battery is calculated based on the temperature safety upper bound, and the battery clusters are ranked by thermal risk. For battery clusters with high thermal risk, control commands for the liquid cooling system are generated on a rolling basis to adjust the flow rate of the corresponding liquid cooling branch, provided that the upper limit of temperature safety is met. If the predicted temperature after adjustment is still higher than the upper limit of the temperature safety range, output a power limiting request to the energy management system.
2. The method for multi-point wireless temperature measurement and liquid cooling combined control of the surface of an energy storage battery according to claim 1, characterized in that, The temperature data is time-aligned, specifically including: Obtain the start time of the current control cycle as a unified time reference; Calculate the time difference between the reading time of each wireless temperature measurement tag and the unified time reference; When the time difference does not exceed the set allowable value, the reading temperature of each tag is corrected to the equivalent temperature value under a unified time reference using the following calculation formula: ,in, The equivalent temperature of the i-th tag at the control time. value, For the i-th tag at the reading time Reading temperature Let be the estimated rate of temperature change for the i-th label; When the time difference exceeds the set allowable value, the label is marked as missing, and the corresponding temperature data is not included in subsequent control.
3. The method for multi-point wireless temperature measurement and liquid cooling combined control of the surface of an energy storage battery according to claim 1, characterized in that, This also includes quality assessment of the label temperature data, specifically including: The communication quality of each tag is obtained, and the communication quality is determined based on the data age, the number of successful reads in the current period, the read signal quality, the number of consecutive missed read periods, and the dispersion of multiple reads; The temperature measurement uncertainty of each label is obtained, and the temperature measurement uncertainty is determined based on the label calibration error, installation and bonding thermal resistance, and dynamic response error.
4. The method for multi-point wireless temperature measurement and liquid cooling combined control of the surface of an energy storage battery according to claim 1, characterized in that, Associating each tag with its corresponding liquid cooling branch according to the location mapping table includes: establishing an association matrix between temperature measurement nodes according to the location mapping table. When two tags are located on the same battery surface, adjacent batteries, upstream or downstream of the same liquid cooling branch, or affected by the same regulating valve, the corresponding element in the association matrix is 1; otherwise, it is 0.
5. The method for multi-point wireless temperature measurement and liquid cooling combined control of the surface of an energy storage battery according to claim 1, characterized in that, The temperature prediction model includes: The physical temperature rise calculation layer calculates the base temperature rise using battery current, internal resistance, coolant temperature, and branch flow rate. The storage pool calculation layer includes a local storage pool for each battery cluster and a system-level global storage pool. The local storage pool is used to describe the dynamic relationship between the cluster temperature, actual valve position, branch flow rate and cluster current. The global storage pool is used to describe the dynamic relationship between the chiller, circulating pump, outdoor temperature and inter-cluster liquid flow coupling. The topology-constrained attention layer calculates the attention weights of each associated node to the target label at multiple historical moments based on the spatial position in the label position mapping table and the liquid cooling flow direction within the node range defined by the association matrix, forming the context vector of the target label. The attention weights are determined based on at least one of the following relationship features: distance between nodes, whether they belong to the same battery, upstream and downstream direction of liquid cooling, liquid cooling branch number, and common regulating valve number. The shared readout layer receives the current temperature, location code, communication quality, temperature measurement uncertainty, missing mask, local reservoir status, global reservoir status, and context vector of each tag. It maps these data to prediction residuals using a readout function shared by all tags, and then superimposes the prediction residuals onto the base temperature output by the physical temperature rise calculation layer. Finally, it outputs the predicted temperature tag by tag.
6. The method for multi-point wireless temperature measurement and liquid cooling combined control of the surface of an energy storage battery according to claim 3, characterized in that, The upper limit of the safe temperature is obtained in the following way: ; in, To predict temperature, To predict the residual margin, it is determined by the quantile statistics of the historical forecast under-prediction error; For temperature measurement uncertainty, The maximum temperature rise rate at the i-th tag location is determined by field calibration data; The data age is denoted by , and the time difference between the last successful read of the tag and the current time.
7. The method for multi-point wireless temperature measurement and liquid cooling combined control of the surface of an energy storage battery according to claim 1, characterized in that, The temperature margin is the difference between the upper limit of the temperature safety range and the current temperature. The temperature margin of the battery cluster is the minimum value of the temperature margin of all tags in the cluster. The minimum value is used for thermal risk ranking.
8. The method for multi-point wireless temperature measurement and liquid cooling combined control of the surface of an energy storage battery according to claim 1, characterized in that, The control commands for the rolling liquid cooling system specifically include: Under the condition of satisfying the upper limit of temperature safety, a rolling time-domain optimization function is constructed. The evaluation index of the optimization function includes a safety term, a temperature uniformity term, auxiliary power consumption, and actuator action quantity. The safety term is used to represent the risk of the predicted temperature exceeding the upper limit of temperature safety, the temperature uniformity term is used to represent the temperature uniformity within the battery cluster, and the auxiliary power consumption and actuator action quantity are used to represent the energy consumption of the liquid cooling system. Each actuator's candidate actions are selected from a finite set of actions consisting of maintaining the current opening, increasing by one step, and decreasing by one step; Within the prediction time domain, search from a finite set of actions for the optimal combination of valve opening and pump speed that makes the optimization function optimal; Only the optimal control command at the current moment is executed, and rolling optimization is performed again in the next control cycle.
9. The method for multi-point wireless temperature measurement and liquid cooling combined control of the surface of an energy storage battery according to claim 1, characterized in that, When the predicted temperature is still higher than the safety upper limit after rolling adjustment, the following steps are executed in sequence: cancel the energy-saving operation mode of the liquid cooling system, increase the speed of the circulating pump and the opening of the valves of the hot spot corresponding to the branch, output the available power boundary of the high-risk battery cluster to the energy management system, and if the temperature safety upper limit constraint is still not met, output the overall power reduction request of the system.
10. A multi-point wireless temperature measurement and liquid cooling combined control system for the surface of an energy storage battery, characterized in that, include: Wireless temperature measurement tags are placed at multiple preset locations on the surface of each battery. Each tag has a unique code and is used to measure the temperature at its location and wirelessly transmit the temperature data. The reader and antenna are used to poll and read the temperature data of the tags in each reading area; The edge temperature measurement gateway is used to receive tag data uploaded by each reader and perform time alignment, data age determination, and quality marking. The thermal management controller is used to establish a tag location mapping table, predict the temperature through a temperature prediction model and obtain the upper limit of temperature safety, sort the thermal risks based on the temperature margin, and generate control commands for the liquid cooling system in a rolling manner. The liquid cooling system includes at least a refrigeration unit, a circulating pump, a main supply and return water pipe, a liquid cooling branch for the battery cluster, and a regulating valve, for receiving and executing the control commands; The communication interface is used to send power limiting requests to the energy management system.
