Sparse sensor optimization layout method for battery thermal safety monitoring

Abstract

The invention discloses a sparse sensor optimization layout method for battery thermal safety monitoring. The method comprises the following steps: constructing a full-working-condition temperature field data set through high-fidelity simulation data; defining a multi-task loss function fusing the overall reconstruction error, the hotspot monitoring error and the spatial dispersion; and with minimization of the loss function as a target, an optimal sensor distribution point is automatically searched by adopting an optimization algorithm. According to the method, the problems of sensing blind areas and early warning lag existing in empirical layout are solved, high-precision temperature field reconstruction and early thermal safety early warning can be achieved with the minimum number of sensors, the reliability is high, the cost is low, and the method is suitable for thermal management design of various lithium ion battery systems.

Description

Technical Field

[0001] This invention relates to the field of lithium-ion battery thermal management technology, specifically to a method for the layout of temperature monitoring sensors for battery packs or modules, and more particularly to an optimized layout method that utilizes high-fidelity simulation data and optimization algorithms to achieve high-precision, high-reliability temperature field reconstruction and thermal safety early warning with a minimum number of sensors. Background Technology

[0002] The performance, lifespan, and safety of lithium-ion batteries are highly dependent on their operating temperature. Uneven temperature distribution accelerates battery aging and can even trigger thermal runaway in extreme cases. Therefore, real-time and accurate monitoring of the battery temperature field is crucial. However, densely distributing a large number of temperature sensors within the battery pack is impractical in terms of cost, wiring complexity, and reliability. Currently, engineering practices typically rely on experience, placing a small number of sensors in limited locations (such as near the battery terminals or the geometric center). This experience-based layout has significant drawbacks. First, there are blind spots, potentially missing local hotspots that occur in variable locations during actual operation. Second, the reconstruction accuracy is low; based on sparse and suboptimal point measurements, it is difficult to accurately reconstruct the true overall temperature field through interpolation or models, especially in high-gradient regions. Finally, there is a lag in safety warnings; it is insensitive to small, localized abnormal temperature rises that may occur in the early stages of thermal runaway, leading to delayed warnings.

[0003] In existing technologies, although some studies have attempted to optimize sensor layout, most of them are based on simplified analytical models or single operating conditions, and fail to fully consider the dynamic thermal behavior of batteries under complex and varied electrical, thermal and aging conditions in reality, resulting in insufficient robustness and generalization ability of optimized layout. Summary of the Invention

[0004] To address the above problems, this invention provides a sparse sensor optimization layout method for battery thermal safety monitoring.

[0005] A method for optimizing the sparse sensor layout for battery thermal safety monitoring includes the following steps: Step 1: Establish a high-fidelity electrochemical thermo-simulation model and a full-temperature field dataset: A three-dimensional electrochemical-thermal coupled finite element model of the target battery was established as a high-fidelity electrochemical-thermal simulation model for simulation; temperature data of all grid nodes of the target battery under each simulation condition were collected to form a full temperature field dataset. ,in M This represents the total number of spatial grid points. N This represents the number of samples under operating conditions. R Represent real numbers; Step 2: Construct a task-oriented objective loss function for temperature sensor layout optimization. L(S) : ; S This represents the set of selected temperature sensor location indices. These are adjustable weighting coefficients; L recon (S) This represents the overall temperature field reconstruction error loss; L hotspot (S) This indicates the monitoring error loss in hotspot areas. L spread (S) This indicates the spatial dispersion penalty loss of the temperature sensor; Step 3: Use a greedy search algorithm to obtain the target loss function. L(S) The optimal set of temperature sensor location coordinates corresponding to the minimum value ; Step 4, according to Install temperature sensors.

[0006] Further improvements, ; in, Indicates the first Under various operating conditions, the battery's full range M The true temperature vector of each grid node Indicates the first Under each operating condition, the global temperature vector is reconstructed solely based on the measurements from the temperature sensors in layout S. Let L2 be the norm of the vector.

[0007] Further improvements, ; Represents the true temperature vector Extracted from hotspot areas Subvectors of internal temperature values; Indicates the reconstruction of the temperature vector Extracted corresponding hotspot areas Temperature subvector; No. Hotspot areas for individual working conditions For the first Under each operating condition, identify the front with the highest temperature. Grid points.

[0008] Further improvements, .

[0009] Further improvements,

[0010] in, It is a combination number, representing the number of combinations from... The number of combinations of selecting 2 temperature sensors from a given temperature sensor; dist( p , q () indicates the location of the temperature sensor p and q The three-dimensional Euclidean distance between them.

[0011] In a further improvement, step three employs a greedy search algorithm to search all... M Iteratively select from 1 candidate position K The positions constitute the optimal set. The steps are as follows: 3.1) Initialize the selected set The candidate set R = {1, 2, ..., M}; Represents the empty set; 3.2) For each candidate position Calculate the target loss function L(S∪{r}) corresponding to the trial set S∪{r}; where, for the first candidate position, L spread (S) Set to 0; 3.3) Select the candidate position r in the candidate set R that maximizes or minimizes the loss as the selected position. ; 3.4) Update the selected set S and the candidate set R: This will update the selected positions. Add elements to the selected set S, and remove elements from the candidate set R; 3.5) Repeat steps 3.2)-3.4) until the number of selected positions in S reaches a certain threshold. K The final optimal set of sensor position coordinates is output. .

[0012] In a further improvement, the target battery is a battery module or a battery pack.

[0013] Advantages of this invention: 1. High precision and high reliability: Based on high-fidelity data covering all operating conditions, the resulting layout can adapt to the complexity and dynamism of real battery operation, ensuring high-precision reconstruction of the overall temperature field, especially areas with potential safety hazards, in various scenarios.

[0014] 2. Task-oriented design: By explicitly incorporating "hotspot monitoring accuracy" and "spatial dispersion" into the optimization objectives, the layout directly serves the core engineering task of "thermal safety early warning," overcoming the shortcomings of simply pursuing the minimum overall reconstruction error and potentially ignoring local risks.

[0015] 3. Optimal cost-effectiveness: Under the premise of strictly limiting the number of sensors, the algorithm automatically finds the points with the most information, thereby maximizing the monitoring efficiency and achieving the highest level of security monitoring with the lowest hardware cost.

[0016] 4. Strong engineering applicability: The optimization process can be automated, and the results are output in intuitive 3D coordinates, making it easy to integrate into the design and manufacturing process of battery packs. This method is versatile and applicable to batteries with different packaging forms such as cylindrical, prismatic, and pouch cells. Attached Figure Description

[0017] Figure 1 This is an overall flowchart of the method of the present invention.

[0018] Figure 2a A three-dimensional layout diagram of temperature sensors optimized using the method of the present invention (example: 16 sensors). Figure 2b This is a planar projection layout diagram of a temperature sensor optimized using the method of the present invention.

[0019] Figure 3a To achieve the 3D temperature field reconstructed by the optimized layout and empirical layout of this invention under fast charging conditions.

[0020] Figure 3b To enable the use of the temperature field in the XY plane reconstructed by this invention under fast charging conditions. Detailed Implementation

[0021] The technical solution of the present invention will be specifically described below through specific embodiments and in conjunction with the accompanying drawings.

[0022] Example 1 Implementation Cases

[0023] Taking a certain vehicle power battery module as the optimization object, the module is composed of 12 square lithium-ion battery cells connected in series.

[0024] S1: Dataset Construction A three-dimensional geometric model of the battery pack module is created in finite element software. This model includes individual battery cells (distinguishing between positive and negative electrodes, separators, current collectors, and other layered structures), battery casing, electrode tabs, busbars, cooling system (liquid cooling plates or air ducts), and contact surfaces with the cooling medium. The internal thermal conductivity of the battery cell is anisotropic.

[0025] Design a simulation operating condition matrix including: charge / discharge rate (0.2C-3C), ambient temperature (-20°C-55°C), cooling flow rate (2-12 mL / min), and different levels of battery capacity decay (SOH 95%-80%). Run all combinations of the above operating condition matrix using finite element software.

[0026] Total N =500 valid operating condition samples. Extract the surface data of the battery pack under each operating condition. M =Temperature of 4000 grid nodes, construct dataset matrix T=M*N.

[0027] S2: Define the optimization problem From the M grid nodes where sensors can be installed, select K points ( K =16), making the sensor layout S formed by these K points optimal under the optimization objective evaluation:

[0028] Where S is the set of all candidate points, This represents the number of elements in set S (i.e., the number of sensors).

[0029] The weights of the loss function were initially tuned and set as follows: This gives higher priority to hotspot monitoring.

[0030] Overall temperature field reconstruction error L recon ( S Based on measurements from sensor layout S, the temperature field is reconstructed. To balance evaluation efficiency and accuracy, Gappy POD is chosen as a fast reconstruction method. The dataset T (M*N) is pre-decomposed into a set of basis vectors representing the main temperature distribution patterns, retaining the first 15 principal modes. For any layout S, a selection matrix is ​​defined. Each row of this matrix contains only one 1, corresponding to a sensor location in matrix S. Therefore, for the i-th operating condition, the actual measured sparse temperature vector is... Where t (i) This is the true full-field temperature vector. The reconstructed temperature field is calculated by solving the following least-squares problem using Gappy POD to find a set of coefficients a(i) that minimizes the error between the reconstructed and measured values ​​at the sensor location:

[0031] Solving ,in The reconstruction site is then... .

[0032] For all training conditions N, calculate the normalized root mean square error and take the average:

[0033] Hotspot area monitoring error L hotspot ( S ): For the firsti The actual temperature field under each working condition The temperature values ​​of all M points are calculated using a temperature field reconstruction method, sorted from highest to lowest, and hotspot regions are defined. This represents the set of nodes with the highest temperatures (top 5%). Temperature values ​​of these hotspot nodes are extracted from both the actual and reconstructed temperature fields to form sub-vectors. and Calculate the normalized root mean square error of the hotspot region:

[0034] Sensor spatial dispersion penalty term L spread ( S To avoid concentrating all sensors in a small, defined area, a spatially distributed distribution is encouraged to better capture changes in ambient temperature and reduce the risk of single-point failure. First, the sensor coordinates are obtained; let the three-dimensional spatial coordinates of each candidate sensor point j be... For layout S, obtain the specific coordinates of all K sensors. Calculate the Euclidean distance between all pairs of sensors in layout S and average them.

[0035] Next, we construct a penalty term. In order to maximize the average distance while minimizing the total loss, we define the penalty term as the negative of the average distance.

[0036] S3: Determine weighting coefficients and optimization algorithm: Initially set the weighting coefficients, prioritizing thermal safety, and assign higher weights to hotspot error terms: Pareto front analysis was employed to test the layout performance under different weight combinations on the validation set, and the optimal balance between overall error and hotspot error was selected.

[0037] Algorithm Selection and Parameters: Since selecting K combinatorial explosion points from M points is a problem requiring a heuristic algorithm, we employ a genetic algorithm. The inputs are the number of candidate points M, the number of sensors K, and the loss function. Starting from the initial set of values, we iterate N times, finding the point that minimizes the total loss function, and outputting the optimal set of sensor indices. S∗ .

[0038] S4: Sensor Optimization Layout Analysis Optimized sensor position as follows Figure 2a and Figure 2b As shown, they are not evenly distributed, but rather densely distributed in the upper part of the module (corresponding to the conventional heat generation concentration area) and on the sides near the cooling air inlet and outlet (corresponding to the area with a large temperature gradient).

[0039] To verify the effect, a set of extreme operating condition data simulating local cooling failure, which was not involved in the optimization, was used for testing. As shown in Figure 3, the temperature field reconstructed using the optimized layout of this invention ( Figure 3a and 3b This allows for more accurate detection of localized hot spots (red areas) caused by uneven cooling.

[0040] This indicates that the sensor layout obtained by the method of the present invention has a more sensitive ability to perceive potential thermal risks and can provide the battery management system with earlier and more accurate thermal anomaly warning signals.

[0041] The above is only one specific implementation method of the present invention, but the design concept of the present invention is not limited thereto. Any non-substantial modifications made to the present invention using this concept shall be considered as infringing the protection scope of the present invention.

Claims

1. A sparse sensor optimization layout method for battery thermal safety monitoring, characterized in that, Includes the following steps: Step 1: Establish a high-fidelity electrochemical thermo-simulation model and a full-temperature field dataset: A three-dimensional electrochemical-thermal coupled finite element model of the target battery was established as a high-fidelity electrochemical-thermal simulation model for simulation; temperature data of all grid nodes of the target battery under each simulation condition were collected to form a full temperature field dataset. ,in M This represents the total number of spatial grid points. N This represents the number of samples under operating conditions. R Represent real numbers; Step 2: Construct a task-oriented objective loss function for temperature sensor layout optimization. L(S) : ； S This represents the set of selected temperature sensor location indices. These are adjustable weighting coefficients; L recon (S) This represents the overall temperature field reconstruction error loss; L hotspot (S) This indicates the monitoring error loss in hotspot areas. L spread (S) This indicates the spatial dispersion penalty loss of the temperature sensor; Step 3: Use a greedy search algorithm to obtain the target loss function. L(S) The optimal set of temperature sensor location coordinates corresponding to the minimum value ; Step 4, according to Install temperature sensors.

2. The sparse sensor optimization layout method for battery thermal safety monitoring as described in claim 1, characterized in that: ； in, Indicates the first Under various operating conditions, the battery's full range M The true temperature vector of each grid node Indicates the first Under each operating condition, the global temperature vector is reconstructed solely based on the measurements from the temperature sensors in layout S. Let L2 be the norm of the vector.

3. The sparse sensor optimization layout method for battery thermal safety monitoring as described in claim 1, characterized in that: ； Represents the true temperature vector Extracted from hotspot areas Subvectors of internal temperature values; Indicates the reconstruction of the temperature vector Extracted corresponding hotspot areas Temperature subvector; No. Hotspot areas for individual working conditions For the first Under each operating condition, identify the front with the highest temperature. Grid points.

4. The sparse sensor optimization layout method for battery thermal safety monitoring as described in claim 4, characterized in that: 。 5. The sparse sensor optimization layout method for battery thermal safety monitoring as described in claim 1, characterized in that: ； in, It is a combination number, representing the number of combinations from... The number of combinations of selecting 2 temperature sensors from a given temperature sensor; dist( p , q () indicates the location of the temperature sensor p and q The three-dimensional Euclidean distance between them.

6. The sparse sensor optimization layout method for battery thermal safety monitoring as described in claim 1, characterized in that: In step three, a greedy search algorithm is used to search from all... M Iteratively select from 1 candidate position K The positions constitute the optimal set. The steps are as follows: 3.1) Initialize the selected set The candidate set R = {1, 2, ..., M}; Represents the empty set; 3.2) For each candidate position Calculate the target loss function L(S∪{r}) corresponding to the trial set S∪{r}; where, for the first candidate position, L spread (S) Set to 0; 3.3) Select the candidate position r in the candidate set R that maximizes or minimizes the loss as the selected position. ; 3.4) Update the selected set S and the candidate set R: This will update the selected positions. Add elements to the selected set S, and remove elements from the candidate set R; 3.5) Repeat steps 3.2)-3.4) until the number of selected positions in S reaches a certain threshold. K The final optimal set of sensor position coordinates is output. .

7. The sparse sensor optimization layout method for battery thermal safety monitoring as described in claim 1, characterized in that: The target battery is a battery module or a battery pack.

Images