Lithium ion battery thermal runaway propagation path prediction and intelligent blocking system

Abstract

The invention discloses a lithium ion battery thermal runaway propagation path prediction and intelligent blocking system. The system comprises a lithium ion battery pack, a high-precision sensor group, an intelligent blocking module, a central controller, an algorithm regulation and control module, a data processing module and a monitoring and feedback module. The intelligent blocking module dynamically adjusts the battery state according to the prediction result, and thermal runaway diffusion is avoided; the algorithm regulation and control module analyzes sensor data by using a dynamic neural radiation field NeRF model and generates an optimal blocking scheme; according to the method, the occurrence and propagation path of thermal runaway can be accurately predicted by adopting a high-precision sensor and an intelligent algorithm, so that the temperature and current are intelligently regulated and controlled for blocking, the safety and reliability of a battery system are effectively improved, and meanwhile, potential risks caused by thermal runaway can also be reduced.

Description

Technical Field

[0001] The present invention relates to a lithium-ion battery thermal runaway propagation path prediction and intelligent blocking system, and in particular to a lithium-ion battery thermal runaway propagation path prediction and intelligent blocking system based on a dynamic neural radiation field (NeRF) model and an intelligent optimization algorithm. Background Art

[0002] Today, lithium-ion batteries, thanks to their numerous advantages, are widely used in key areas such as electric vehicles and energy storage systems. However, this widespread use has also led to increasing concern about the safety of lithium-ion batteries. Among the many safety hazards, thermal runaway is particularly prominent. It is often triggered by factors such as internal short circuits, external overcharging, or exposure to high temperatures. Once thermal runaway occurs, the internal temperature of the battery rises rapidly within a very short period of time, accompanied by the release of large amounts of gas. In severe cases, it can even cause an explosion, posing a significant threat to life and property. Traditional battery safety protection technologies rely primarily on insulating materials to slow heat transfer or employ simple monitoring methods to identify potential safety hazards. However, these traditional methods have significant limitations. First, their ability to predict thermal runaway is extremely limited, making it impossible to accurately determine its occurrence and propagation path in advance. Second, their mitigation measures are relatively simple and difficult to flexibly adjust to complex operating conditions, making them ineffective in responding to various emergencies. Furthermore, traditional warning technologies have slow response times and unsatisfactory accuracy, making them inadequate for modern, complex and ever-changing applications. To overcome these shortcomings, researchers have begun exploring new approaches in recent years. For example, a single sensor is used to monitor a single key battery parameter to detect potential thermal runaway risks. However, this approach suffers from a fatal flaw: it cannot integrate multidimensional data. A battery's safety status is determined by multiple factors, and relying solely on single sensor data struggles to fully and accurately reflect the battery's true condition. Furthermore, some physical modeling-based approaches, while theoretically sound, struggle to meet real-time requirements in practice. Battery operating conditions are dynamic, and the calculation and adjustment of physical models often takes a long time, making them incapable of timely predicting and preventing thermal runaway. Meanwhile, the application of machine learning technology in the field of battery safety has also made some progress. Machine learning algorithms can learn from and analyze large amounts of historical data, uncovering underlying patterns within the data and thus predicting battery status. However, machine learning methods are not a panacea. It is highly data-dependent, requiring a large amount of high-quality data for model training. Insufficient data or low-quality data significantly reduces the model's predictive performance. Furthermore, machine learning models also have shortcomings in dynamic optimization. Battery operating conditions are constantly changing, and once machine learning models are trained, their parameters are often difficult to quickly adjust based on real-time data. This makes it difficult to promptly deliver optimal blocking strategies in the face of dynamically changing thermal runaway situations. Summary of the Invention

[0003] Purpose of the invention: The purpose of the present invention is to provide a thermal runaway propagation path prediction and intelligent blocking system based on a dynamic neural radiation field model, which can effectively prevent the spread of thermal runaway and improve battery safety and reliability.

[0004] Technical solution: The present invention provides a lithium-ion battery thermal runaway propagation path prediction and intelligent blocking system, comprising:

[0005] High-precision sensor group: including temperature sensors, pressure sensors and gas sensors, used to monitor the temperature, pressure and gas concentration changes of the battery in real time and transmit the data to the data processing module;

[0006] Thermal runaway propagation path prediction module: This module uses a dynamic neural radiation field model and combines sensor data to simulate the thermal and pressure field changes of batteries under high temperature conditions, and predicts the propagation path and speed of thermal runaway.

[0007] Intelligent blocking module: used to adjust battery temperature, current, and isolation status according to the analysis results of the thermal runaway propagation path prediction module through the instructions of the central controller;

[0008] Algorithm control module: Used to optimize the real-time data collected by sensors using the dynamic neural radiation field model, 4D-TCN algorithm, Kalman filter algorithm and CADFM risk assessment model, predict the probability of thermal runaway, and send instructions to the central controller to achieve dynamic control of battery temperature, current and isolation status in the intelligent blocking module.

[0009] Furthermore, the system also includes a data processing module for collecting sensor data and performing noise removal processing; a monitoring and feedback module for tracking intelligent blocking results. If an abnormality occurs after blocking, the feedback signal is sent to the algorithm control module for optimization.

[0010] Furthermore, the algorithm control module includes:

[0011] (1) Construct a dynamic neural radiation field model and set the initial objective function, X = (x1,…,x d ) as the candidate solution represented by the d-dimensional vector, set the maximum number of iterations of the algorithm, and define the range of the search space according to the number of iterations and the battery operating conditions;

[0012] (2) Evaluate the battery performance indicators based on the initial objective function, use the 4D-TCN algorithm to process the temporal-spatial characteristics of the sensor data, input the position information, output the dynamic neural radiation field model prediction results, and define the comprehensive objective function in combination with the risk assessment of the CADFM model;

[0013] (3) The model position with the lowest risk is set as the initial optimal position, the sensor data and spectral array are input into the comprehensive objective function, and the performance value of each model is calculated;

[0014] (4) The current state of the battery is processed through the dynamic neural radiation field model to predict the state at the future time t+1. best (t) is used as input for dynamic prediction, and the fitness value F is calculated based on the difference between the state predicted by the dynamic neural radiation field model and each safety threshold. i ;

[0015] (5) Use the dynamic neural radiation field model to predict the future heat loss path based on the current state. The dynamic neural radiation field model generates the predicted path at the future moment by inputting the battery state temperature, voltage and external control current and power; use Kalman filtering to dynamically optimize the path prediction, update the path prediction based on the error covariance, and evaluate the accuracy of each path through the fitness function;

[0016] (6) Predicting the future thermal runaway propagation path cloud map and the state estimation of the CADFM model based on the pre-training results of the dynamic neural radiation field model. Optimizing the blocking strategy based on multi-source sensor data to achieve risk minimization and energy consumption optimization within the thermal runaway critical point.

[0017] (7) Taking the battery state parameters temperature, pressure, and gas concentration as input and the blocking mode selection as output, it is judged whether the stop condition is met; outputting the efficient blocking mode, considering the state improvement, implementation cost, and uniformity of temperature distribution, and finally screening out the optimal blocking mode through the optimization algorithm. By analyzing the battery state parameters and optimizing the objective function, the optimal blocking mode is output.

[0018] Furthermore, the comprehensive objective function in step (2) is as follows:

[0019]

[0020] Among them, F(x i ) is the state variable T of the battery considered comprehensively j ,q j Prediction of changing and dynamic neural radiation field models; T j ,q j is the actual measured temperature field and heat flow; is the predicted value of temperature field and heat flux by the dynamic neural radiation field model, is the risk value of the CADFM model, is the physical constraint term of the equation, describing the electrolyte flow and gas diffusion, and α and β are weight factors.

[0021] Furthermore, the calculation formula of step (3) is as follows:

[0022]

[0023] in, is the estimated value of the battery temperature, voltage, current, power and gas concentration state of the i-th system at time t, A is the state transfer matrix, s t+1,i is the measured value of the battery temperature, pressure, and gas concentration of the i-th system at time t+1, is the estimated value of the battery temperature, voltage, current, power and gas concentration of the i-th system at time t+1, K t is the Kalman gain, which controls the weighting between prediction and observation, y t The real-time temperature, voltage, current, power and gas concentration measured by the observation data sensor, C is the observation matrix, which maps the system state variables battery temperature, voltage, current, power and gas concentration to the actual measurement values.

[0024] Furthermore, the calculation formula of step (4) is as follows:

[0025]

[0026] in, is the value of battery state temperature, voltage, current, power and gas concentration predicted at time t+1, N is the dynamic neural radiation field model used to predict the future battery state based on the current battery state, X best (t) is the optimal battery state at the current time t, W k is the weight matrix, It is a nonlinear function used to process the input sensor data temperature, voltage, current, power and gas concentration. i To optimize the prediction results of battery status, λ j is the weighting coefficient, X safe (j) is the safety threshold of the jth path, which indicates the expected value of the path within the safety range and is used to measure the difference between the predicted value and the safety threshold. m is the weighted coefficient of the gradient term, which is used to control the influence of the gradient term on the fitness value calculation. M is the control variable that needs to be optimized in the system, which adjusts the control strategy of temperature, voltage, current, power and gas concentration.

[0027] Furthermore, the state prediction formula in step (5) is as follows:

[0028]

[0029] K t =P t CT (CP t C T +R) -1

[0030] in, It represents the prediction of the system state at time t+1, B is the control matrix, which represents the impact of changes in external control input process temperature, voltage, gas concentration, etc. on the path, u t For external control input battery charging current and heat source input, y t The real-time temperature and pressure measured by the observation data sensor, C T is the transpose of the observation matrix, R is the measurement noise covariance matrix, which represents the noise or uncertainty generated during the measurement process, and P t is the state covariance matrix.

[0031] Furthermore, the CADFM model in step (6) calculates the risk function and the energy consumption calculation formula input is as follows:

[0032]

[0033] Where T(pi,t) represents the temperature of battery area pi at time t. The temperature data is obtained by real-time monitoring by the sensor and is used to evaluate the thermal state of the battery. threshold (pi) represents the temperature safety threshold of battery area pi, which is determined according to battery material and design; P safe (pi) represents the power safety threshold of battery area pi; G(pi,t) represents the gas concentration of battery area pi at time t, which is monitored by the gas sensor; G safe represents the safety threshold of gas concentration; pi represents the i-th region or unit of the battery, and t represents time;

[0034] The risk function calculated by CADFM, u t+k To control the input vector cooling power and current regulation, F θ (p t+k ,t+k) are the temperature field and heat flow vector predicted by NeRF, w1, w2, w3 are dynamic weights, which are adjusted according to the thermal runaway stage, J(u t ) is to achieve the safe operation and thermal management goals with the lowest energy consumption by minimizing the energy consumption of the control input, and γ and μ are weight factors that adjust the importance of energy consumption and prediction consistency, respectively.

[0035] Furthermore, the formula for step (7) is as follows:

[0036]

[0037] Among them, F(M k ) To adjust the control strategy, ensure system safety and reduce risks, minimize energy consumption and avoid excessive temperature fluctuations, Is the risk function, which means that in a given model M k Next, based on the current state The assessed risk includes the degree to which temperature, power, and airflow factors deviate from safety thresholds, cost (M k ) is the same as the model M k The relevant cost functions include the computational complexity and energy consumption of the model, β and ν are weight factors that weigh the relative importance of different parts of the objective function. is the integral term, which represents the square integral of the gradient of the temperature field T(p,t).

[0038] Furthermore, the optimal blocking mode in step (7) is:

[0039]

[0040] If F(M * )<F threshold Or t=N, output the optimal mode Otherwise, return to step (3). If the condition is not met, return to the next round of iteration until the stop condition is met to achieve the optimal strategy selection for thermal runaway blocking.

[0041] Beneficial effects: Compared with the existing technology, the present invention has the following significant advantages: (1) The dynamic neural radiation field NeRF model is used to provide the best solution between the temperature, pressure and gas concentration data collected by the sensor, the thermal runaway path prediction and the blocking device adjustment, ensuring that the system responds quickly and efficiently blocks the risk of thermal runaway; (2) The use of a high-precision sensor group to monitor the battery status in real time, combined with the intelligent blocking device to dynamically adjust the operating mode, can effectively prevent the spread of thermal runaway, improve the safety and service life of the battery, and at the same time reduce the potential hazards caused by thermal runaway. The intelligent collaboration of sensors and algorithms enables the blocking device to always work in the optimal state; (3) The dynamic neural radiation field NeRF model adopted is simple and efficient in design. The path recognition capability is enhanced by optimizing the prediction process. At the same time, it can achieve high-precision prediction effects in a short time, providing reliable support for battery safety management and promoting the high-quality development of new energy technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a system structure diagram of the present invention.

[0043] Figure 2 This is a flowchart for optimizing the algorithm control module of the present invention. DETAILED DESCRIPTION

[0044] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0045] like Figure 1 As shown, a lithium-ion battery thermal runaway propagation path prediction and intelligent blocking system includes a lithium-ion battery pack, a high-precision sensor group, an intelligent blocking device, a central controller, an algorithm control unit, a data processing unit, and a monitoring and feedback module;

[0046] The lithium-ion battery pack is encapsulated with high-performance materials, optimized for heat dissipation and safety design, adaptable to high-load environments, and has a structure that facilitates sensor installation, ensuring accurate data collection and providing support for prediction and blocking.

[0047] The high-precision sensor group is distributed inside the battery, including temperature, pressure and gas sensors, which monitor changes in temperature, pressure and gas concentration in real time, and transmit the data to the data processing unit through a high-speed channel.

[0048] The thermal runaway propagation path prediction module adopts the dynamic neural radiation field NeRF model, combined with sensor data to simulate the changes in the thermal field and pressure field of the battery under high temperature conditions, and predicts the propagation path and speed of thermal runaway.

[0049] The algorithm control module uses a dynamic neural radiation field model to analyze sensor data and generate predictions of thermal runaway propagation paths. Through multiple rounds of optimization, it combines spatiotemporal information to quickly identify trends and generate blocking solutions.

[0050] The intelligent blocking module dynamically regulates battery status based on instructions from the central controller to prevent thermal runaway. Specifically, it activates local cooling when regional temperatures are too high, increasing cooling power to reduce temperatures. It also shuts off current or isolates modules in the event of pressure or gas abnormalities to prevent spread. The device responds quickly, completing interventions within a short period of time.

[0051] The central controller serves as the command center, receives the algorithm optimization plan and converts it into control instructions and sends them to the intelligent blocking module.

[0052] The data processing unit organizes and optimizes sensor data, removes noise, and provides stable input for the algorithm control unit to ensure real-time and reliable prediction.

[0053] The monitoring and feedback module tracks the blocking effect, analyzes the operation status, and adjusts the strategy. If the abnormality persists after blocking, the feedback signal is sent to the algorithm unit to trigger optimization, ensuring that the risk is completely eliminated and enhancing the adaptability and safety of the system.

[0054] The present invention combines sensors with a dynamic neural radiation field model to achieve efficient prediction and intelligent blocking of thermal runaway paths, which can issue alarms and take action in advance; the blocking device is fast and flexible, improving battery safety; data processing and feedback optimize resource utilization and ensure low-energy operation.

[0055] The specific process is as follows:

[0056] (1) Construct NeRF model, X=(x1,...,x d ) as a candidate solution represented by a d-dimensional vector, where d is the dimension of the model parameters; set the maximum number of iterations N of the algorithm, and define the range of the search space according to the number of iterations and the battery operating conditions; the initial objective function is as follows:

[0057]

[0058] Among them, F(M k ) is a multi-objective optimization function that comprehensively evaluates and optimizes the state parameters of the battery area, including temperature, voltage, current, power, and gas concentration, and accurately predicts the propagation path of thermal runaway before it occurs. i ,t) is the temperature of the battery area p, T threshold (p i ) is the temperature safety threshold of the battery area. V(p i ,t) is the voltage at the battery area. V safe (p i ) is the voltage safety threshold of the battery area, I(p i ,t) is the current in the battery area, I safe (p i ) is the current safety threshold of the battery area, P(p i ,t) is the power safety threshold of the battery area, P safe (p i ) is the power safety threshold of the battery area, G(p i ,t) is the gas concentration in the battery area, G safe (p i ) is the gas concentration safety threshold in the battery area. λ1 and λ2 are weighting coefficients that control the influence of state parameters in the objective function. N is the maximum number of iterations of the algorithm, which is the number of all possible paths considered by the model.

[0059] (2) Performance is evaluated based on the objective function. The 4D-TCN network is used to process the temporal-spatial features of sensor data, input position information (p, t), output NeRF prediction results, and combine the risk assessment of CADFM to define the comprehensive objective function as follows:

[0060]

[0061] Among them, F(x i ) is the state variable T of the battery considered comprehensively j ,q jChanges and NeRF model predictions can be used to optimize the operation of the battery management system, predict and effectively control thermal runaway risks in advance, and effectively control thermal runaway risks. j ,q j is the actual measured temperature field and heat flow, is the predicted value of temperature field and heat flux by NeRF, is the risk value of the CADFM model, is the physical constraint term of the equation, describing the electrolyte flow and gas diffusion, and α and β are weight factors.

[0062] (3) Set the model position with the lowest risk as the initial optimal position, and substitute the multi-source data sensor and spectral array into F(x i ), calculate the performance value of each model, and select the model with the lowest risk as the initial best position.

[0063]

[0064] in, is the estimated value of the battery temperature, voltage, current, power and gas concentration state of the i-th system at time t, A is the state transition matrix, which represents the transition rule from the current state to the next state, s t+1,i is the measured value of the battery temperature, voltage, current, power and gas concentration state of the i-th system at time t+1, is the estimated value of the battery temperature, voltage, current, power and gas concentration of the i-th system at time t+1, K t is the Kalman gain, which controls the weighting between prediction and observation, y t The real-time temperature, pressure, current, power, and gas concentration field measured by the observation data sensor are represented by C, which is the observation matrix that maps the system state variables (battery temperature, pressure, current, power, and gas concentration field) to the actual measured values.

[0065] (4) Execute NeRF chain prediction, calculate fitness value, and sort; process the current state of the battery through the NeRF model, predict the state at the future time t+1, and use the existing optimal state X best (t) is used as input for dynamic prediction. The fitness value F is calculated based on the difference between the state predicted by the NeRF model and each safety threshold. i The difference is measured using the L2 norm, and the goal is to evaluate different parameters through weighted coefficients to ensure that the deviation between the predicted value and the safety threshold is minimized, thereby improving the safety of the battery.

[0066]

[0067] in, is the value of battery state temperature, voltage, current, power and gas concentration predicted at time t+1, N is the NeRF neural radiation field model used to predict the future battery state based on the current battery state, X best (t) is the optimal battery state at the current time t, W k is the weight matrix, It is a nonlinear function used to process the input sensor data temperature, voltage, current, power, and gas concentration. i In order to optimize the prediction results of the battery state, the NeRF model and the constraints of multiple physical quantities are combined to ensure that the battery is always within the safe range in the future and avoid drastic changes in the internal parameters of the battery. j is the weighting coefficient, X safe (j) is the safety threshold of the jth path, which indicates the expected value of the path within the safety range and is used to measure the difference between the predicted value and the safety threshold. m is the weighted coefficient of the gradient term, which is used to control the influence of the gradient term on the fitness value calculation. M is the control variable that needs to be optimized in the system, which adjusts the control strategy of temperature, current, power, and gas concentration field.

[0068] (5) Predict the heat loss path and perform NeRF dynamic update and CADFM model collaborative optimization; the purpose of heat loss path prediction is to estimate the changes in temperature, voltage, gas concentration, etc. during the thermal runaway process inside the battery through the model, and then determine the possible heat loss path. It involves state prediction, optimization, and dynamic update. In order to ensure the accuracy of the system, the NeRF model and Kalman filter are combined to perform dynamic prediction of the heat loss path. The NeRF model is used to predict the future heat loss path based on the current state. The NeRF model generates a predicted path at a future moment by inputting the battery's state temperature, voltage, and external control current and power. Kalman filtering is used to dynamically optimize the path prediction, and the path prediction is updated based on the error covariance so that the new prediction result can better match the actual observation data. During the dynamic optimization process, the accuracy of each path is evaluated by the fitness function. Through the fitness function, we can evaluate the accuracy of each predicted path and further optimize the parameters of the NeRF model.

[0069]

[0070] K t =P t C T (CP t C T +R) -1

[0071] in, It represents the prediction of the system state at time t+1, B is the control matrix, which represents the impact of changes in external control input process temperature, voltage, gas concentration, etc. on the path, u t For external control input battery charging current and heat source input, y t The real-time temperature and pressure measured by the observation data sensor, C T is the transpose of the observation matrix, R is the measurement noise covariance matrix, which represents the noise or uncertainty generated during the measurement process, and P t is the state covariance matrix.

[0072] (6) Execute blocking strategy optimization, update position, and calculate fitness value; combine the pre-trained results of the NeRF model to predict the future thermal runaway propagation path cloud map and the state estimation of the CADFM model based on multi-source sensor data, and optimize the blocking strategy (such as cooling, current cutoff, isolation) to minimize risk and optimize energy consumption within the thermal runaway critical point. The optimization process must meet the requirements of prediction error and response time, and ensure consistency with the physical laws of gas diffusion and heat flow dynamics within the battery.

[0073]

[0074] Where T(pi,t) represents the temperature of battery area pi at time t, which is the temperature data obtained by real-time monitoring by the sensor and is used to evaluate the thermal state of the battery; T threshold (pi) represents the temperature safety threshold of the battery area pi, which is determined based on the battery material and design. Exceeding this temperature may cause thermal runaway. safe (pi) represents the power safety threshold of battery area pi. Power exceeding this threshold may cause battery overcharge or over-discharge, increasing the risk of thermal runaway. G(pi,t) represents the gas concentration in battery area pi at time t. The data obtained by gas sensor monitoring is used to evaluate the chemical state inside the battery. G safe represents the safe threshold for gas concentration. Exceeding this concentration may cause internal pressure in the battery to increase, increasing the risk of thermal runaway or explosion. pi represents the i-th region or cell of the battery. Batteries typically consist of multiple regions or cells, each with its own sensors and control strategies. t represents time, an important parameter used to track changes in battery status in dynamic monitoring and prediction.

[0075] The risk function calculated by CADFM, u t+k To control the input vector cooling power and current regulation, F θ (p t+k,t+k) are the temperature field and heat flow vector predicted by NeRF, w1, w2, w3 are dynamic weights, which are adjusted according to the thermal runaway stage, J(u t ) is to achieve the safe operation and thermal management goals with the lowest energy consumption by minimizing the energy consumption of the control input, and γ and μ are weight factors that adjust the importance of energy consumption and prediction consistency, respectively.

[0076] (7) Taking the battery state parameters temperature, voltage, current, power, and gas concentration field as input and the blocking mode selection as output, it is judged whether the stop condition is met; outputting an efficient blocking mode, comprehensively considering the state improvement, implementation cost, and uniformity of temperature distribution, and finally selecting the optimal blocking mode M* through the optimization algorithm. By analyzing the battery state parameters and optimizing the objective function, the optimal blocking mode is output to achieve early warning and effective blocking of thermal runaway.

[0077]

[0078] Among them, F(M k ) To adjust the control strategy, ensure system safety and reduce risks, minimize energy consumption and avoid excessive temperature fluctuations, and ensure the stability and efficient operation of the battery system, Is the risk function, which means that in a given model M k Next, based on the current state The risk of the assessment includes the degree to which the temperature, power, and airflow factors deviate from the safety threshold, cost (M k ) is the same as the model M k The relevant cost functions include the computational complexity and energy consumption of the model. β and ν are weight factors that weigh the relative importance of different parts of the objective function.

[0079] is an integral term, representing the square integral of the gradient of the temperature field T(p,t). This term calculates the smoothness of the temperature change. By minimizing this term, the temperature field can be made smoother and excessive changes can be reduced.

[0080] Among them, the optimal mode is:

[0081] If F(M * )<F threshold Or t=N, output the optimal mode Otherwise, return to step 3. If the condition is not met, return to the next round of iteration until the stop condition is met to achieve the optimal strategy selection for thermal runaway blocking.

Claims

1. A lithium-ion battery thermal runaway propagation path prediction and intelligent blocking system, characterized in that: include: High-precision sensor group: including temperature sensors, pressure sensors and gas sensors, used to monitor the temperature, pressure and gas concentration changes of the battery in real time and transmit the data to the data processing module; Thermal runaway propagation path prediction module: This module uses a dynamic neural radiation field model and combines sensor data to simulate the thermal and pressure field changes of batteries under high temperature conditions, and predicts the propagation path and speed of thermal runaway. Intelligent blocking module: used to adjust battery temperature, current, and isolation status according to the analysis results of the thermal runaway propagation path prediction module through the instructions of the central controller; Algorithm control module: Used to optimize the real-time data collected by sensors using the dynamic neural radiation field model, 4D-TCN algorithm, Kalman filter algorithm and CADFM risk assessment model, predict the probability of thermal runaway, and send instructions to the central controller to achieve dynamic control of battery temperature, current and isolation status in the intelligent blocking module.

2. A lithium-ion battery thermal runaway propagation path prediction and intelligent blocking system according to claim 1, characterized in that: The system also includes a data processing module for collecting sensor data and performing noise removal processing; The monitoring and feedback module is used to track the results of intelligent blocking. If an abnormality occurs after blocking, the feedback signal will be sent to the algorithm control module for optimization.

3. The lithium-ion battery thermal runaway propagation path prediction and intelligent blocking system according to claim 1, characterized in that: The algorithm control module includes: (1) Construct a dynamic neural radiation field model and set the initial objective function, X = (x1,…,x d ) as the candidate solution represented by the d-dimensional vector, set the maximum number of iterations of the algorithm, and define the range of the search space according to the number of iterations and the battery operating conditions; (2) Evaluate the battery performance indicators based on the initial objective function, use the 4D-TCN algorithm to process the temporal-spatial characteristics of the sensor data, input the position information, output the dynamic neural radiation field model prediction results, and define the comprehensive objective function in combination with the risk assessment of the CADFM model; (3) The model position with the lowest risk is set as the initial optimal position, the sensor data and spectral array are input into the comprehensive objective function, and the performance value of each model is calculated; (4) The current state of the battery is processed through the dynamic neural radiation field model to predict the state at the future time t+1. best (t) is used as input for dynamic prediction, and the fitness value F is calculated based on the difference between the state predicted by the dynamic neural radiation field model and each safety threshold. i ; (5) Use the dynamic neural radiation field model to predict the future heat loss path based on the current state. The dynamic neural radiation field model generates the predicted path at the future moment by inputting the battery state temperature, voltage and external control current and power; use Kalman filtering to dynamically optimize the path prediction, update the path prediction based on the error covariance, and evaluate the accuracy of each path through the fitness function; (6) Predicting the future thermal runaway propagation path cloud map and the state estimation of the CADFM model based on the pre-training results of the dynamic neural radiation field model. Optimizing the blocking strategy based on multi-source sensor data to achieve risk minimization and energy consumption optimization within the thermal runaway critical point. (7) Taking the battery state parameters temperature, pressure, and gas concentration as input and the blocking mode selection as output, it is judged whether the stop condition is met; outputting the efficient blocking mode, considering the state improvement, implementation cost, and uniformity of temperature distribution, and finally screening out the optimal blocking mode through the optimization algorithm. By analyzing the battery state parameters and optimizing the objective function, the optimal blocking mode is output.

4. A lithium-ion battery thermal runaway propagation path prediction and intelligent blocking system according to claim 3, characterized in that: The comprehensive objective function in step (2) is as follows: Among them, F(x i ) is the state variable T of the battery considered comprehensively j ,q j Prediction of changing and dynamic neural radiation field models; T j ,q j is the actual measured temperature field and heat flow; is the predicted value of temperature field and heat flux by the dynamic neural radiation field model, is the risk value of the CADFM model, ∫▽·(ρv)dV is the physical constraint term of the equation, describing the electrolyte flow and gas diffusion, and α and β are weight factors.

5. The lithium-ion battery thermal runaway propagation path prediction and intelligent blocking system according to claim 3, characterized in that: The calculation formula of step (3) is as follows: in, is the estimated value of the battery temperature, voltage, current, power and gas concentration state of the i-th system at time t, A is the state transfer matrix, s t+1,i is the measured value of the battery temperature, pressure, and gas concentration of the i-th system at time t+1, is the estimated value of the battery temperature, voltage, current, power and gas concentration of the i-th system at time t+1, K t is the Kalman gain, which controls the weighting between prediction and observation, y t The real-time temperature, voltage, current, power and gas concentration measured by the observation data sensor, C is the observation matrix, which maps the system state variables battery temperature, voltage, current, power and gas concentration to the actual measurement values.

6. A lithium-ion battery thermal runaway propagation path prediction and intelligent blocking system according to claim 3, characterized in that: The calculation formula of step (4) is as follows: in, is the value of battery state temperature, voltage, current, power and gas concentration predicted at time t+1, N is the dynamic neural radiation field model used to predict the future battery state based on the current battery state, X best (t) is the optimal battery state at the current time t, W k is the weight matrix, It is a nonlinear function used to process the input sensor data temperature, voltage, current, power and gas concentration. i To optimize the prediction results of battery status, λ j is the weighting coefficient, X safe (j) is the safety threshold of the jth path, which indicates the expected value of the path within the safety range and is used to measure the difference between the predicted value and the safety threshold. m is the weighted coefficient of the gradient term, which is used to control the influence of the gradient term on the fitness value calculation. M is the control variable that needs to be optimized in the system, which adjusts the control strategy of temperature, voltage, current, power and gas concentration.

7. The lithium-ion battery thermal runaway propagation path prediction and intelligent blocking system according to claim 3, characterized in that: The state prediction formula in step (5) is as follows: K t =P t C T (CP t C T +R) -1 in, It represents the prediction of the system state at time t+1, B is the control matrix, which represents the impact of changes in external control input process temperature, voltage, gas concentration, etc. on the path, u t For external control input battery charging current and heat source input, y t The real-time temperature and pressure measured by the observation data sensor, C T is the transpose of the observation matrix, R is the measurement noise covariance matrix, which represents the noise or uncertainty generated during the measurement process, and P t is the state covariance matrix.

8. The lithium-ion battery thermal runaway propagation path prediction and intelligent blocking system according to claim 3, characterized in that: The CADFM model in step (6) calculates the risk function and the energy consumption calculation formula input as follows: Where T(pi,t) represents the temperature of battery area pi at time t. The temperature data is obtained by real-time monitoring by the sensor and is used to evaluate the thermal state of the battery. threshold (pi) represents the temperature safety threshold of battery area pi, which is determined according to battery material and design; P safe (pi) represents the power safety threshold of battery area pi; G(pi,t) represents the gas concentration of battery area pi at time t, which is monitored by the gas sensor; G safe represents the safety threshold of gas concentration; pi represents the i-th region or unit of the battery, and t represents time; The risk function calculated by CADFM, u t+k To control the input vector cooling power and current regulation, F θ (p t+k ,t+k) are the temperature field and heat flow vector predicted by NeRF, w1, w2, w3 are dynamic weights, which are adjusted according to the thermal runaway stage, J(u t ) is to achieve the safe operation and thermal management goals with the lowest energy consumption by minimizing the energy consumption of the control input, and γ and μ are weight factors that adjust the importance of energy consumption and prediction consistency, respectively.

9. The lithium-ion battery thermal runaway propagation path prediction and intelligent blocking system according to claim 3, characterized in that: The formula for step (7) is as follows: Among them, F(M k ) To adjust the control strategy, ensure system safety and reduce risks, minimize energy consumption and avoid excessive temperature fluctuations, Is the risk function, which means that in a given model M k Next, based on the current state The assessed risk includes the degree to which temperature, power, and airflow factors deviate from safety thresholds, cost (M k ) is the same as the model M k The relevant cost functions include the computational complexity and energy consumption of the model, β and ν are weight factors that weigh the relative importance of different parts of the objective function. is the integral term, which represents the square integral of the gradient of the temperature field T(p,t).

10. The lithium-ion battery thermal runaway propagation path prediction and intelligent blocking system according to claim 3, characterized in that: The optimal blocking mode in step (7) is: If F(M * )<F threshold Or t=N, output the optimal mode Otherwise, return to step (3). If the condition is not met, return to the next round of iteration until the stop condition is met to achieve the optimal strategy selection for thermal runaway blocking.

Images