Method for dynamic balancing and thermal runaway early warning of megawatt energy storage battery management system

Abstract

The application belongs to the technical field of electrochemical energy storage, and particularly relates to a dynamic balancing and thermal runaway early warning method of a megawatt-level energy storage battery management system, which is applied to an energy storage system with a thermal imaging component and a partitioned temperature control component. The method comprises the following steps: collecting voltage data of each battery branch, calculating a voltage imbalance rate and a change rate thereof; obtaining battery surface temperature distribution data through the thermal imaging component, marking a heat source area and extracting a temperature extreme value and a change rate thereof; generating a cooperative balancing strategy based on voltage and heat source information, performing bidirectional energy transfer between battery branches, and limiting a balancing current flowing to the heat source area; inputting voltage, heat source and historical data into a thermal runaway prediction model based on a federal learning framework, and outputting a thermal runaway risk level; and driving the partitioned temperature control component to accurately dissipate heat for the corresponding heat source area according to the risk level. The application realizes electro-thermal cooperative management, and improves balancing efficiency and thermal safety early warning capability of the system.

Description

Technical Field

[0001] This invention belongs to the field of electrochemical energy storage technology, specifically relating to a dynamic balancing and thermal runaway early warning method for megawatt-level energy storage battery management systems. Background Technology

[0002] With the rapid development of electrochemical energy storage technology, megawatt-level energy storage systems have been widely used in grid peak shaving and renewable energy consumption. These systems are typically composed of a large number of batteries connected in series and parallel. During operation, due to differences in initial capacity, internal resistance, and aging degree among individual batteries, voltage imbalance between battery clusters can easily occur, leading to circulating current phenomena. This can cause some batteries to be overcharged or over-discharged, seriously affecting the overall performance and lifespan of the system.

[0003] To address the aforementioned issues, existing technologies generally employ battery equalization management systems, adjusting energy distribution among batteries through active or passive equalization methods. For example, active equalization circuits based on Cuk and Buck-Boost topologies have been extensively studied, achieving controllable energy transfer between batteries through a switching matrix and a DC-DC converter. In recent years, parallel equalization systems based on bidirectional Cuk converters, such as the one proposed in CN119944896A, have also emerged, supporting various equalization modes including many-to-one, one-to-many, and many-to-many, thus improving equalization flexibility and efficiency.

[0004] In terms of thermal safety management, traditional battery management systems (BMS) largely rely on discretely arranged temperature sensors for localized temperature monitoring and trigger alarms or heat dissipation measures based on preset temperature thresholds. Some systems combine multi-source data such as voltage and current, employing empirical models or simple statistical methods to assess the risk of thermal runaway. However, these methods are limited by the spatial coverage of the sensors, making it difficult to comprehensively capture the temperature distribution characteristics of the battery surface, easily missing local hotspots, and leading to delayed warnings. Furthermore, existing thermal management strategies are usually independent of equalization control, failing to achieve coordinated regulation of heat generation and dissipation processes, thus limiting the system's safe operation capability in scenarios with non-uniform heat distribution. Summary of the Invention

[0005] The purpose of this invention is to provide a dynamic balancing and thermal runaway early warning method for megawatt-level energy storage battery management systems, so as to solve the problems of low balancing efficiency and delayed thermal runaway early warning caused by electrothermal coupling imbalance in megawatt-level energy storage battery management systems.

[0006] The present invention achieves the above objectives through the following technical solutions: This invention proposes a dynamic balancing and thermal runaway early warning method for a megawatt-level energy storage battery management system, applicable to energy storage systems equipped with thermal imaging components and zoned temperature control components. The method includes the following steps: Collect voltage data of each battery branch in the energy storage module; determine the voltage imbalance rate and change rate based on the voltage data; obtain temperature distribution data of the battery surface through the thermal imaging component, mark the heat source area, and extract heat source area information including regional temperature extreme values ​​and temperature change rate; Based on the voltage imbalance rate and change rate, as well as the heat source area information, a collaborative balancing strategy is generated, and bidirectional energy transfer is performed between the corresponding battery branches based on the collaborative balancing strategy. The voltage data, the heat source area information, and historical operating data are input into the thermal runaway prediction model built based on the federated learning framework, and the model outputs the thermal runaway risk level for the heat source area. Based on the thermal runaway risk level, the zone temperature control component is driven to dissipate heat from the corresponding heat source area; Specifically, under the collaborative balancing strategy, the balancing current flowing to the battery branch marked as a heat source area is limited to a level lower than the balancing current flowing to the battery branch in the non-heat source area.

[0007] Furthermore, the step of determining the voltage imbalance rate based on the voltage data... and rate of change This can be achieved using the following formula: ; in, This is the highest voltage value among all battery branches; This is the lowest voltage value among all battery branches; This represents the average voltage value across all battery branches. ; in, The voltage imbalance rate at the current moment; This represents the voltage imbalance rate at the previous sampling time. The sampling period.

[0008] Furthermore, the collaborative equilibrium strategy is generated through the following steps: Compare the voltages of each battery branch, identify one or more branches with the highest voltage as energy sources, and identify one or more branches with the lowest voltage as energy targets; Identify whether the energy target is marked as a heat source region; If the energy target is marked as a heat source region, a current limiting command is generated for that branch, limiting the maximum permissible equalization current to a level lower than that for branches flowing to non-heat source regions.

[0009] Furthermore, the bidirectional energy transfer between corresponding battery branches based on the cooperative balancing strategy includes: When the voltage imbalance rate exceeds a first preset threshold, dynamic balancing is initiated. If the energy target is marked as a heat source region, one of the following strategies is executed: (1) Remove the energy target from the charging objects in the current equalization cycle and charge the branch with the second lowest voltage and no heat source instead; (2) Charge the energy target with the reduced equalization current limit according to the current limit instruction.

[0010] Furthermore, the bidirectional energy transfer is performed through a pre-constructed active balancing system, which includes: The Cuk equalization circuit includes a power inductor, a coupling capacitor, a main switch transistor, and a diode. An input switch matrix is ​​provided, with its input terminals connected to the positive terminals of each battery branch and its output terminals connected together to the positive input terminal of the Cuk equalization circuit. An output switch matrix, whose input terminal is connected to the positive output terminal of the Cuk equalization circuit, and whose output terminal is connected to the positive terminal of each battery branch. The controller is used to generate the cooperative balancing strategy, output a gating signal to dynamically configure the gating state of the input switch matrix and the output switch matrix, and output a PWM drive signal to control the on and off of the main switch.

[0011] Furthermore, the controller is configured to execute a fuzzy PID control algorithm to achieve dynamic adjustment of the equalization current, including: The fuzzy PID control algorithm uses the voltage imbalance rate and its rate of change as input variables, and the duty cycle of the PWM drive signal of the main switch in the Cuk equalization circuit as output variables. The controller constrains the upper limit of the output duty cycle of the fuzzy PID control algorithm according to the current limiting instruction in the collaborative balancing strategy, so as to limit the balancing current flowing to the heat source area to a preset level.

[0012] Furthermore, the controller constrains the upper limit of the output duty cycle of the fuzzy PID control algorithm according to the current limiting command in the cooperative balancing strategy, including: A maximum allowable equalization current value I1 is set for the battery branch that will be marked as a heat source area; Based on the output voltage U0 of the Cuk equalization circuit and the circuit parameters, the corresponding maximum safe duty cycle D is calculated according to the following formula. max : ; in, The voltage value of the energy source branch selected by the input switch matrix; The voltage value of the energy target branch selected by the output switch matrix; The equivalent series resistance of the Cuk equalization circuit; This is the estimated duty cycle value, and its initial value is given by the formula. The results are calculated and iteratively optimized during the control process; The reference duty cycle D calculated by the fuzzy PID control algorithm pid With the maximum safe duty cycle D max Compare and output the final duty cycle D. final : .

[0013] Furthermore, a thermal runaway prediction model is constructed based on the federated learning framework, including: An initial model for predicting thermal runaway was constructed based on a long short-term memory network and deployed at various energy storage nodes. At each energy storage node, the initial model for thermal runaway prediction is trained using its own voltage data, heat source area information, and historical operating data to generate local model parameter update values. The local model parameter update values ​​of each energy storage node are uploaded to a central server. The central server updates the global model parameters based on the multiple local model parameter update values ​​using a federated aggregation algorithm. The aggregated and updated global model parameters are distributed to each energy storage node to update its local thermal runaway prediction model.

[0014] Furthermore, driving the partition temperature control component to dissipate heat from the corresponding heat source area includes: Based on the physical location of the heat source area determined by the thermal imaging component, it is mapped to one or more corresponding zone fans; The rotational speed of the partitioned fans is linearly adjusted according to the thermal runaway risk level, wherein the higher the risk level, the higher the fan speed.

[0015] Furthermore, the extraction of regional temperature extremes includes: For a confirmed heat source region, its regional temperature extremes are determined according to the following rules: (1) If there are discrete temperature sensors in the region, the maximum value of these sensor readings shall be taken as the extreme value of the region temperature. (2) If there is no discrete temperature sensor in the region, the highest temperature value in the two-dimensional temperature distribution data of the region shall be taken as the extreme temperature value of the region.

[0016] The beneficial effects of this invention are as follows: 1. In terms of battery equalization management, by integrating the voltage imbalance state with the two-dimensional temperature field information obtained by thermal imaging technology, dynamic equalization control of electric and thermal synergy is realized. After identifying the heat source area, the system can actively limit the current intensity flowing to the area in the equalization strategy, or temporarily exclude it from the current equalization cycle, effectively avoiding the risk of aggravating local overheating due to equalization energy injection.

[0017] 2. In terms of thermal safety early warning and control, the thermal runaway prediction model built using the federated learning framework can improve the accuracy of risk identification by comprehensively utilizing distributed data to train the model while protecting the data privacy of each node. Combined with zoned temperature control technology, the system can accurately and on-demand heat dissipation for specific heat source areas according to the predicted risk level, which changes the traditional method of global unified heat dissipation based on fixed thresholds. Attached Figure Description

[0018] Figure 1 A flowchart of a dynamic equilibrium and thermal runaway early warning method provided in one embodiment of the present invention; Figure 2 This is a flowchart illustrating the execution of an electrothermal synergistic dynamic equilibrium strategy proposed in one embodiment of the present invention. Figure 3 This is a topology diagram of a centralized equalization circuit based on a Cuk converter in the prior art, as shown in one embodiment of the present invention. Figure 4 This is a system block diagram of an active balancing system proposed in one embodiment of the present invention. Detailed Implementation

[0019] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0020] Example 1 A specific embodiment of the present invention proposes a dynamic balancing and thermal runaway early warning method for a megawatt-level energy storage battery management system, which is applied to an energy storage system equipped with thermal imaging components and zoned temperature control components.

[0021] In the application scenario of this invention, the thermal imaging component of the energy storage system periodically acquires two-dimensional temperature distribution data of the battery surface through an infrared thermal imager and its image processing unit deployed in the battery compartment, so as to identify and locate the heat source area; the zoned temperature control component consists of multiple independently controlled fan units. These fans are divided into different zones according to the physical space inside the battery compartment. Each zone can establish a mapping relationship with the heat source area identified by the thermal imaging component, and can independently adjust its fan speed based on commands to achieve precise heat dissipation for specific areas.

[0022] Existing technologies, which generally rely on discrete temperature sensors, have spatial monitoring blind spots, making it difficult to comprehensively and timely capture localized abnormal heat points on the battery surface, resulting in delayed thermal runaway warnings. In contrast, this invention acquires two-dimensional temperature distribution data of the battery surface using an infrared thermal imager, enabling the complete identification of local hotspots and their precise locations. This provides a target for subsequent precise zoned heat dissipation. More importantly, it allows the system to incorporate heat source area information as a key decision variable, feeding it back into the dynamic balancing strategy and mitigating the risk of exacerbating localized overheating due to equalization charging from the outset.

[0023] Please see Figures 1-2 Specifically, the dynamic equilibrium and thermal runaway early warning method includes the following steps: S1. Collect real-time voltage data of each battery branch within the energy storage module. Based on the collected voltage data, calculate key indicators reflecting the consistency of the battery cluster, including the voltage imbalance rate and its rate of change. The voltage imbalance rate is defined as the ratio of the difference between the highest and lowest voltages in all battery branches to the average voltage, while its rate of change characterizes the dynamic evolution trend of the imbalance state. Simultaneously, periodically acquire two-dimensional temperature distribution data of the battery surface using deployed thermal imaging components. Based on this temperature distribution data, identify and mark heat source areas with significantly higher temperatures than the surrounding areas, and further extract key heat source area information from these marked areas, mainly including the temperature extremes (i.e., the highest temperature points) within the area and the temperature change rate of the area.

[0024] S2. Based on the voltage imbalance rate and its rate of change obtained in step S1, as well as the heat source area information, a collaborative equilibrium strategy is generated. The core of this strategy lies in combining electrical state equilibrium with thermal safety management.

[0025] Specifically, when determining which battery branch requires energy transfer—that is, transferring energy from the branch with the highest voltage to the branch with the lowest voltage—the system simultaneously checks whether the "energy target" branch, acting as the energy receiver, has been marked as a heat source region. If so, when that branch is selected as the charging target, a specific current limiting command is generated to ensure that the equalization current flowing to the battery branch marked as a heat source region is limited to a pre-set level lower than the equalization current flowing to battery branches in non-heat source regions. Subsequently, based on this collaborative equalization strategy, the system performs bidirectional energy transfer between the corresponding battery branches through an active equalization system.

[0026] S3. The real-time voltage data collected in step S1, the extracted heat source region information, and the historical operating data stored in the system are used as input features and input into a thermal runaway prediction model built based on a federated learning framework. The historical operating data may include historical temperature, current, voltage curves, etc. The model finally outputs a quantified thermal runaway risk level for the identified heat source region.

[0027] S4. Based on the thermal runaway risk level obtained in step S3, generate corresponding control commands to drive the zone temperature control components to perform targeted heat dissipation on the corresponding heat source areas.

[0028] Specifically, based on the physical location information determined by the thermal imaging component, logical heat source areas are mapped onto one or more actual zone fans or cooling units. Then, according to the risk level, the speed of these corresponding zone fans is adjusted linearly or in stages. The higher the risk level, the higher the fan speed, thereby achieving enhanced heat dissipation in high-risk areas.

[0029] Preferably, the voltage imbalance rate is determined based on voltage data. and rate of change This can be achieved using the following formula: ; in, This is the highest voltage value among all battery branches; This is the lowest voltage value among all battery branches; This represents the average voltage value across all battery branches. ; in, The voltage imbalance rate at the current moment; This represents the voltage imbalance rate at the previous sampling time. The sampling period.

[0030] In one implementation, the collaborative equilibrium strategy is generated through the following steps: First, the system controller continuously compares the real-time voltage values ​​of each battery branch, identifying one or more branches with the highest voltage as the energy source for this equalization process, and one or more branches with the lowest voltage as the energy target. This step aims to identify the branch pairs in the system with the most significant energy differences and the greatest need for equalization.

[0031] Subsequently, the identified energy target is compared with the heat source area information marked by the thermal imaging component to determine whether the energy target branch has been marked as a heat source area.

[0032] If the determination result is yes, meaning the energy target is overheating or at risk of overheating, a specific current limiting instruction will be generated. The core of this instruction is to limit the maximum permissible equalization current flowing to the energy target branch marked as a heat source area to a pre-set level lower than the equalization current flowing to the battery branch in the non-heat source area.

[0033] For example, the upper limit of the equalization current flowing to the non-heat source branch can be set to I. n Instead, the upper limit of the current flowing to the heat source branch is set to a lower value I. s (I) s n This mechanism aims to prevent high-current charging of already heated battery cells, thereby avoiding the risk of thermal runaway caused by the superposition of ohmic heat and electrochemical reactions exacerbating the temperature rise.

[0034] Specifically, bidirectional energy transfer between corresponding battery branches based on a collaborative balancing strategy includes the following processes: When the voltage imbalance rate exceeds the first preset threshold, it is determined that the battery cluster inconsistency has reached a level requiring intervention, and dynamic balancing is initiated; if the energy target is marked as a heat source area, one of the following strategies is executed: (1) Remove the energy target from the charging targets in the current equalization cycle and charge the branch with the second lowest voltage and no heat source instead; this strategy completely avoids injecting any additional equalization energy into the high-risk heat source area.

[0035] (2) In accordance with the current limit instruction, the energy target is to charge the device with the reduced upper limit of the equalization current, which reduces the heat generated during the charging process.

[0036] Please see Figure 4 In one implementation, bidirectional energy transfer is performed via a pre-built active balancing system, which includes a Cuk balancing circuit, an input switch matrix, an output switch matrix, and a controller.

[0037] ​The Cuk equalization circuit includes a power inductor, coupling capacitors, a main switch (such as a MOSFET), and diodes. The input switch matrix is ​​connected to the positive terminal of each battery branch, and its output is connected to the positive input terminal of the Cuk equalization circuit. The output switch matrix is ​​connected to the positive output terminal of the Cuk equalization circuit, and its output is connected to the positive terminal of each battery branch.

[0038] In this invention, the Cuk equalization circuit adopts a centralized equalization circuit based on a Cuk converter in the prior art, and its topology is as follows: Figure 3 As shown. This circuit uses power inductor L1, coupling capacitor C1, main switch Q1, and freewheeling diode D as core components to form a bidirectional DC-DC converter architecture. Input voltage U E The output voltage U is selected from the high-voltage battery branch by a switch matrix. O This is then connected to the low-voltage branch. The controller precisely controls the on / off state of the main switch Q1 via a PWM signal: when Q1 is on, energy is stored from the input side in L1 and C1; when Q1 is off, the stored energy is transferred to the output side through diode D, and after being filtered by C2, it charges the target branch. By adjusting the PWM duty cycle in real time, the magnitude of the balancing current can be precisely controlled. A sampling resistor R is set in the circuit to monitor the actual balancing current in real time, providing feedback for current closed-loop control.

[0039] Specifically, the input switch matrix and the output switch matrix are key to the system's ability to transfer energy between arbitrary branches. These two switch matrices can be composed of multiple controllable semiconductor switches (such as MOSFETs or relays).

[0040] The controller is used to generate a collaborative balancing strategy, output gating signals to dynamically configure the gating states of the input and output switch matrices, and output PWM drive signals to control the on and off of the main switch transistor.

[0041] In practical implementation, based on the collaborative balancing strategy, the controller outputs two sets of key control signals: The first is a gating signal, used to dynamically configure the on / off states of each switch in the input and output switch matrices. In this way, the controller can arbitrarily connect one or more high-voltage branches to the input of the Cuk balancing circuit through the input switch matrix, and simultaneously connect one or more low-voltage branches to the output of the Cuk balancing circuit through the output switch matrix, thus establishing a complete energy transfer path. The second is a PWM drive signal, used to control the duty cycle of the main switch in the Cuk balancing circuit. By adjusting the duty cycle of this PWM signal, the controller can precisely control the rate of energy transfer, i.e., the magnitude of the balancing current.

[0042] In one implementation, the controller is configured to execute a fuzzy PID control algorithm to achieve dynamic adjustment of the equalization current, including: The fuzzy PID control algorithm uses the voltage imbalance rate and its rate of change as input variables and the duty cycle of the PWM drive signal of the main switch in the Cuk equalization circuit as output variables. The controller constrains the upper limit of the output duty cycle of the fuzzy PID control algorithm according to the current limiting command in the cooperative equalization strategy, so as to limit the equalization current flowing to the heat source area to a preset level.

[0043] Through fuzzy inference, the algorithm can nonlinearly and adaptively adjust the PWM duty cycle according to the complex changes in the voltage imbalance state (e.g., the imbalance rate is large and is rapidly expanding), thereby changing the balancing current and enabling the system to respond to the imbalance state more quickly. Compared with traditional PID control, it has better dynamic performance.

[0044] In an optional implementation, the controller is configured in terms of control timing as follows: When initiating energy transfer, the gating signal is first output to establish a complete current path, and then the PWM drive signal is output to start energy transfer; when stopping energy transfer, the PWM drive signal is first stopped to interrupt energy transmission, and then the gating signal is deactivated to disconnect the current path.

[0045] Specifically, the controller constrains the upper limit of the output duty cycle of the fuzzy PID control algorithm based on the current limiting command in the cooperative balancing strategy, including: A maximum allowable equalization current value I1 is set for the battery branch that is marked as a heat source area, so as to specify that it is lower than the upper limit of the current flowing to the non-heat source branch.

[0046] Based on the output voltage U0 of the Cuk equalization circuit and the circuit parameters, the corresponding maximum safe duty cycle D is calculated according to the following formula. max : ; in, The voltage value of the energy source branch selected by the input switch matrix; The voltage value of the energy target branch selected by the output switch matrix; and All values ​​are collected in real time; The equivalent series resistance of the Cuk equalization circuit is determined by the parasitic parameters of the circuit elements and is a known fixed value or obtained through prior calibration. This is the estimated duty cycle value, and its initial value is given by the formula. The calculations are obtained and iterative optimizations are performed during the control process, such as iterative optimization based on the actual feedback of the circuit.

[0047] As an example, at the initial moment, From the formula The calculation is as follows. Subsequently, in each control cycle (or after several PWM cycles), the controller performs an iterative optimization. The actual feedback upon which the optimization is based is the deviation between the actual equalization current value Ia measured by the sampling resistor and the target safe current threshold I1. . If there is a deviation The absolute value is greater than the set tolerance. I, then according to Update the estimated value, where the iteration step size coefficient These are small positive numbers pre-calibrated through circuit experiments, used to control convergence speed and stability. Simultaneously, the system... Physical boundary protection is applied, constraining it within a reasonable range of [0.1, 0.9] to prevent computational overflow. This iterative process continues to run, ensuring... It can adapt to changes in the circuit's operating point, thereby allowing The dynamic calculation ensures that the current is strictly limited to Furthermore, it avoids significantly reducing equilibrium efficiency due to overly conservative boundaries.

[0048] The reference duty cycle D calculated by the fuzzy PID control algorithm pid With the maximum safe duty cycle D max Compare and output the final duty cycle D. final : .

[0049] Understandably, the aforementioned step of constraining the upper limit of the duty cycle ensures that no matter how large the duty cycle is calculated by the fuzzy PID algorithm for the purpose of fast balancing, the actual duty cycle ultimately applied to the main switch is the same. None of them will exceed the limits determined by thermal safety conditions. This strictly limits the equalization current flowing to the heat source area to a preset safety level I1.

[0050] In one implementation, a thermal runaway prediction model is constructed based on a federated learning framework, including: An initial thermal runaway prediction model is constructed based on a Long Short-Term Memory (LSTM) network and deployed to each energy storage node. At each energy storage node, its own voltage data, heat source region information, and historical operating data are used to train the initial thermal runaway prediction model, generating updated local model parameters. These updated local model parameters are then uploaded to a central server. The central server updates the global model parameters based on these multiple local updated values ​​using a federated aggregation algorithm. Finally, the aggregated and updated global model parameters are distributed to each energy storage node to update its local thermal runaway prediction model.

[0051] In practice, the initial LSTM model deployed in the battery management unit of each energy storage station continuously receives local real-time monitoring data, including battery branch voltage, extreme surface temperatures of specific cells identified by thermal imagers, and their changing trends. This data forms the training time series locally, used for periodic model fine-tuning and generating parameter updates. These parameter increments are uploaded to the central server in the regional cloud. The server aggregates updates from multiple sites and merges them into an optimized global model using a federated averaging algorithm. The updated model parameters are then distributed to each site, replacing their old models. Thus, each site gains the ability to predict local overheating risks earlier and more accurately, based on extensive data training. For example, slight fluctuations in output voltage accompanied by abnormal temperature rise rates in specific areas can trigger early warnings and current-limiting cooling measures.

[0052] Specifically, the drive zone temperature control component dissipates heat from the corresponding heat source areas, including: Based on the physical location of the heat source area determined by the thermal imaging component, it is mapped to one or more corresponding zone fans; the speed of the zone fans is linearly adjusted according to the thermal runaway risk level, where the higher the risk level, the higher the fan speed.

[0053] Understandably, there's a mapping between different physical areas within the battery compartment and the specific fan units responsible for cooling those areas. When the thermal imaging component identifies a heat source and reports its coordinates or number, the controller queries this mapping table to determine the target fan zone that needs to be activated. For example, if the heat source is located on the third layer of battery cluster A, the mapping will point to the fan group specifically serving the third layer of cluster A.

[0054] In one implementation, extracting regional temperature extremes includes: For a confirmed heat source region, its regional temperature extremes are determined according to the following rules: (1) If there are discrete temperature sensors in the area, the system will first read the real-time readings of one or more discrete temperature sensors (such as NTC thermistors) deployed in the physical area. Then, the maximum value among these sensor readings will be taken as the temperature extreme value of the area.

[0055] (2) If there is no discrete temperature sensor in the area, the system directly takes the highest temperature pixel value in the two-dimensional temperature distribution data provided by the thermal imaging component as the temperature extreme value of the area. This rule ensures that even in areas where discrete sensors are not deployed, the system can still obtain key temperature extreme value information based on thermal imaging data.

[0056] According to the above embodiments, this invention addresses the circulating current and thermal runaway risks in parallel operation of battery clusters in megawatt-level energy storage systems by collaboratively managing them through the fusion of electrothermal data. It collects voltage data from each branch to calculate the imbalance rate, and simultaneously uses a thermal imaging component to acquire a two-dimensional temperature distribution and identify heat source regions. Based on this, a collaborative balancing strategy is generated. When charging is required in the branch of the heat source region, the balancing current is limited to a preset safety value (e.g., a maximum of 20A). A restricted energy transfer is performed through an active balancing system. Multi-source data is input into a thermal runaway prediction model constructed through federated learning, which outputs a risk level. Finally, based on the risk level, a zoned temperature control component is driven to precisely dissipate heat from the heat source region.

[0057] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0058] In addition, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0059] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A dynamic balancing and thermal runaway early warning method for a megawatt-level energy storage battery management system, applied to an energy storage system equipped with thermal imaging components and zoned temperature control components, characterized in that... The method includes the following steps: Collect voltage data of each battery branch in the energy storage module; determine the voltage imbalance rate and change rate based on the voltage data; obtain temperature distribution data of the battery surface through the thermal imaging component, mark the heat source area, and extract heat source area information including regional temperature extreme values ​​and temperature change rate; Based on the voltage imbalance rate and change rate, as well as the heat source area information, a collaborative balancing strategy is generated, and bidirectional energy transfer is performed between the corresponding battery branches based on the collaborative balancing strategy. The voltage data, the heat source area information, and historical operating data are input into the thermal runaway prediction model built based on the federated learning framework, and the model outputs the thermal runaway risk level for the heat source area. Based on the thermal runaway risk level, the zone temperature control component is driven to dissipate heat from the corresponding heat source area; In the aforementioned collaborative balancing strategy, the balancing current flowing to the battery branch marked as a heat source area is limited to a level lower than the balancing current flowing to the battery branch in the non-heat source area. The bidirectional energy transfer is performed through a pre-constructed active balancing system, which includes: The Cuk equalization circuit includes a power inductor, a coupling capacitor, a main switch transistor, and a diode. An input switch matrix is ​​provided, with its input terminals connected to the positive terminals of each battery branch and its output terminals connected together to the positive input terminal of the Cuk equalization circuit. An output switch matrix, whose input terminal is connected to the positive output terminal of the Cuk equalization circuit, and whose output terminal is connected to the positive terminal of each battery branch. The controller is used to generate the collaborative balancing strategy, output a gating signal to dynamically configure the gating state of the input switch matrix and the output switch matrix, and output a PWM drive signal to control the on and off of the main switch. The controller is configured to execute a fuzzy PID control algorithm to achieve dynamic adjustment of the equalization current, including: The fuzzy PID control algorithm uses the voltage imbalance rate and its rate of change as input variables, and the duty cycle of the PWM drive signal of the main switch in the Cuk equalization circuit as output variables. The controller constrains the upper limit of the output duty cycle of the fuzzy PID control algorithm according to the current limiting instruction in the collaborative balancing strategy, so as to limit the balancing current flowing to the heat source area to a preset level. The controller constrains the upper limit of the output duty cycle of the fuzzy PID control algorithm according to the current limiting command in the cooperative balancing strategy, including: A maximum allowable equalization current value I1 is set for the battery branch that will be marked as a heat source area; Based on the output voltage U0 of the Cuk equalization circuit and the circuit parameters, the corresponding maximum safe duty cycle D is calculated according to the following formula. max : ； in, The voltage value of the energy source branch selected by the input switch matrix; The voltage value of the energy target branch selected by the output switch matrix; The equivalent series resistance of the Cuk equalization circuit; The duty cycle is an estimated value, and its initial value is given by the formula. The results are calculated and iteratively optimized during the control process; The reference duty cycle D calculated by the fuzzy PID control algorithm pid With the maximum safe duty cycle D max Compare and output the final duty cycle D. final : 。 2. The dynamic balancing and thermal runaway early warning method for a megawatt-level energy storage battery management system according to claim 1, characterized in that, The voltage imbalance rate is determined based on the voltage data. and rate of change This can be achieved using the following formula: ； in, This is the highest voltage value among all battery branches; This is the lowest voltage value among all battery branches; This represents the average voltage value across all battery branches. ； in, The voltage imbalance rate at the current moment; This represents the voltage imbalance rate at the previous sampling time. The sampling period.

3. The dynamic balancing and thermal runaway early warning method for a megawatt-level energy storage battery management system according to claim 1, characterized in that, The collaborative equilibrium strategy is generated through the following steps: Compare the voltages of each battery branch, identify one or more branches with the highest voltage as energy sources, and identify one or more branches with the lowest voltage as energy targets; Identify whether the energy target is marked as a heat source region; If the energy target is marked as a heat source region, a current limiting command is generated for that branch, limiting the maximum permissible equalization current to a level lower than that for branches flowing to non-heat source regions.

4. The dynamic balancing and thermal runaway early warning method for a megawatt-level energy storage battery management system according to claim 3, characterized in that, The bidirectional energy transfer between corresponding battery branches based on the cooperative balancing strategy includes: When the voltage imbalance rate exceeds a first preset threshold, dynamic balancing is initiated. If the energy target is marked as a heat source region, one of the following strategies is executed: (1) Remove the energy target from the charging objects in the current equalization cycle and charge the branch with the second lowest voltage and no heat source instead; (2) Charge the energy target with the reduced equalization current limit according to the current limit instruction.

5. The dynamic balancing and thermal runaway early warning method for a megawatt-level energy storage battery management system according to claim 1, characterized in that, A thermal runaway prediction model is constructed based on the federated learning framework, including: An initial model for predicting thermal runaway was constructed based on a long short-term memory network and deployed at various energy storage nodes. At each energy storage node, the initial model for thermal runaway prediction is trained using its own voltage data, heat source area information, and historical operating data to generate local model parameter update values. The local model parameter update values ​​of each energy storage node are uploaded to a central server. The central server updates the global model parameters based on the multiple local model parameter update values ​​using a federated aggregation algorithm. The aggregated and updated global model parameters are distributed to each energy storage node to update its local thermal runaway prediction model.

6. The dynamic balancing and thermal runaway early warning method for a megawatt-level energy storage battery management system according to claim 1, characterized in that, The step of driving the zoned temperature control component to dissipate heat from the corresponding heat source area includes: Based on the physical location of the heat source area determined by the thermal imaging component, it is mapped to one or more corresponding zone fans; The rotational speed of the partitioned fans is linearly adjusted according to the thermal runaway risk level, wherein the higher the risk level, the higher the fan speed.

7. The dynamic balancing and thermal runaway early warning method for a megawatt-level energy storage battery management system according to claim 1, characterized in that, The extracted regional temperature extremes include: For a confirmed heat source region, its regional temperature extremes are determined according to the following rules: (1) If there are discrete temperature sensors in the region, the maximum value of these sensor readings shall be taken as the extreme value of the region temperature. (2) If there is no discrete temperature sensor in the region, the highest temperature value in the temperature distribution data of the region shall be taken as the extreme temperature value of the region.

Images