Lithium battery all-parameter real-time intelligent monitoring method and system

Abstract

The invention relates to the field of data processing, in particular to a lithium battery all-parameter real-time intelligent monitoring method and system, and the method comprises the steps: collecting all parameters of a lithium battery pack in real time, carrying out the correlation analysis of the all parameters and the hot-spot temperature, and constructing a feature vector; presetting a sliding window, correcting a covariance matrix of the sliding window based on mutual information of each feature vector and the hot spot temperature and a Pearson's correlation coefficient, presetting a temperature sensitive projection coefficient, maximizing the covariance matrix and the sum of squares of the preset temperature sensitive projection coefficient and the hot spot temperature covariance; obtaining a projection matrix and calculating a temperature sensitive projection coefficient; taking the temperature-sensitive projection coefficient as single control input, scheduling the rotating speed of an exhaust motor and calculating cooling income; and by taking the cooling income as a supervision signal, iteratively updating the projection matrix through a gradient rising method, so that the temperature-sensitive projection coefficient is continuously aligned with the optimal direction of energy-saving temperature control, explosion-proof air cooling millisecond response and hot spot early warning are realized, and the power consumption of a fan is reduced.

Description

Technical Field

[0001] This invention relates to the field of data processing. In particular, it relates to a method and system for real-time intelligent monitoring of all parameters of a lithium battery. Background Technology

[0002] With the development of the new energy industry, lithium batteries are widely used in energy storage, electric vehicles and other fields. Lithium batteries are a type of battery that uses lithium metal or lithium alloy as positive / negative electrode materials and non-aqueous electrolyte solutions. However, the thermal safety of lithium batteries is a core pain point. If the heat generated during the operation of lithium batteries cannot be dissipated in time, it is easy to cause thermal runaway. In the existing technology, the temperature of lithium batteries is monitored and cooled by explosion-proof equipment. Explosion-proof equipment mostly uses exhaust fans to cool lithium batteries.

[0003] Existing monitoring systems for lithium batteries have the following problems: the monitoring parameters are too broad, and the high-dimensional data leads to real-time processing delays, which cannot meet the rapid response requirements of cooling systems in explosion-proof equipment; the motor speed of the air-cooling system in explosion-proof equipment is mostly controlled by a fixed threshold or only responds to changes in real-time temperature, resulting in energy waste or untimely adjustment. Summary of the Invention

[0004] To address the closed-loop optimization problems of existing lithium battery monitoring systems, such as delayed real-time processing of high-dimensional thermal safety parameters, poor temperature control targeting, and wasteful air-cooling energy consumption, this invention provides solutions in the following aspects.

[0005] In the first aspect, a real-time intelligent monitoring method for all parameters of a lithium battery includes: real-time acquisition of all parameters of the lithium battery pack and Pearson correlation analysis with hot spot temperature, followed by normalization to construct feature vectors for each parameter; pre-setting a sliding window, correcting the covariance matrix of the sliding window based on the mutual information between each feature vector and the hot spot temperature and the Pearson correlation coefficient as heat-sensitive weight coefficients; pre-setting temperature-sensitive projection coefficients, maximizing the sum of the covariance matrix and the square of the covariance of the pre-set temperature-sensitive projection coefficients and the hot spot temperature to obtain a projection matrix; calculating the temperature-sensitive projection coefficients of each parameter feature based on the projection matrix; using the temperature-sensitive projection coefficients as a single control input to linearly and steplessly schedule the exhaust motor speed, using the hot spot temperature corresponding to the lowest exhaust motor speed as the benchmark temperature control, taking the positive difference between the benchmark temperature control and the measured temperature control within the sliding window as the net cooling benefit, calculating the cooling benefit based on the net cooling benefit and the ratio of the current speed of the exhaust motor to the maximum speed; using the cooling benefit as a monitoring signal, iteratively updating the projection matrix through the gradient ascent method to continuously align the temperature-sensitive projection coefficients with the optimal direction of energy-saving temperature control until convergence.

[0006] All parameters are first filtered using a combination of Pearson and mutual information, retaining only the temperature-sensitive projection coefficient most relevant to the hotspot temperature as the sole control input. This enables the exhaust motor to adjust its speed within milliseconds, achieving a rapid response for lithium battery explosion prevention. After the sliding window covariance matrix is ​​corrected in real time by the heat-sensitive weights, the projection direction automatically aligns with the local pattern most prone to overheating, enabling early warning of hotspots. The projection matrix is ​​continuously iterated using gradient ascent, with cooling benefits as the monitoring signal, forcing the algorithm to find the optimal trade-off between effective cooling and minimum power consumption, effectively reducing the energy consumption of the exhaust motor.

[0007] Preferably, the thermal sensitivity weighting coefficient is calculated by multiplying the mutual information between the feature vectors of each parameter and the hotspot temperature and the Pearson correlation coefficient, and then normalizing the product to obtain the thermal sensitivity weighting coefficient.

[0008] Preferably, the method for adjusting the speed of the exhaust motor is as follows: calculate the difference between the maximum speed and the minimum speed of the exhaust motor, multiply it by the temperature-sensitive projection coefficient, and then add the minimum speed of the exhaust motor to obtain the adjusted speed of the exhaust motor.

[0009] It uses the minimum speed to continuously maintain the minimum ventilation volume, suppressing hot spot retention and nighttime noise, and releases the remaining speed range in a precise proportion according to the temperature-sensitive projection coefficient to avoid energy waste.

[0010] Preferably, the cooling benefit is calculated as follows: the difference between the adjusted speed of the exhaust fan motor and the maximum speed is calculated as the relative speed, and the net cooling benefit is multiplied by 1 after subtracting the relative speed.

[0011] Preferably, the method for updating the projection matrix is ​​as follows: the eigenvectors of the parameters are reduced in dimension and then increased in dimension to obtain the reconstructed vector; the error between the eigenvectors of the parameters at the current time and the reconstructed vector is calculated to obtain the reconstructed error vector, and the local covariance outer product is formed and pushed along the projection direction as the unsupervised term for updating the projection matrix; the square of the covariance between the temperature-sensitive projection coefficient and the cooling benefit is calculated and matrix differentiation is performed as the supervised term for updating the projection matrix; the cooling benefit is used as the supervised signal to perform a decaying gradient ascent on the projection matrix, the unsupervised term maintains the principal component direction, the gradient of the supervised term is fine-tuned, and orthogonalization and deviance from the initial value penalty are combined to iterate until convergence.

[0012] By updating the projection matrix through a dual-drive approach of reconstructing unsupervised error terms and supervised cooling benefit terms, the projection matrix can be slightly adaptively rotated along the energy-saving temperature control direction while maintaining the principal components of the original parameters. On the one hand, the external product of local covariance is used to lock in the structural changes caused by battery aging and sensor drift, preventing dimensionality reduction distortion. On the other hand, the gradient of the temperature-sensitive projection coefficient and the square of the cooling benefit covariance directly rewards the projection direction with high energy saving and high cooling. Combined with orthogonalization and deviating from the initial value penalty, large oscillations and overfitting are avoided.

[0013] Preferably, an initial value for the unsupervised learning rate is preset, and the square root of the number of iterations incremented by 1 is taken as the decay factor. The ratio of the initial value of the unsupervised learning rate to the decay factor is calculated to obtain an adaptive unsupervised learning rate and dynamically scale the step size of the unsupervised term.

[0014] Preferably, an initial value for the supervised learning rate is preset, the ratio of the cooling gain at the current moment to the standard deviation of all cooling gains within the sliding window is calculated as the signal-to-noise ratio gain, the square of the difference between the current projection matrix and the initial projection matrix is ​​calculated and subjected to a decay exponential function operation, and then multiplied by the initial value for the supervised learning rate and the signal-to-noise ratio gain to obtain the adaptive supervised learning rate and dynamically scale the step size of the supervised terms.

[0015] In a second aspect, a real-time intelligent monitoring system for all parameters of a lithium battery includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the real-time intelligent monitoring method for all parameters of a lithium battery described in any one of the claims is implemented.

[0016] The present invention has the following effects: 1. All parameters are first filtered by Pearson and mutual information, and only the temperature-sensitive projection coefficient most relevant to the hot spot temperature is retained as the sole control input. This enables the exhaust motor to complete speed adjustment in the millisecond, achieving a rapid response for lithium battery explosion prevention. After the sliding window covariance matrix is ​​corrected in real time by the heat-sensitive weights, the projection direction is automatically aligned with the local pattern most prone to overheating, realizing early warning of hot spots. The projection matrix is ​​continuously iterated using gradient ascent, with cooling benefits as the supervision signal, forcing the algorithm to find the optimal trade-off between effective cooling and minimum power consumption, effectively reducing the energy consumption of the exhaust motor.

[0017] 2. By reconstructing the unsupervised error term and the supervised cooling benefit term to update the projection matrix, the projection matrix is ​​slightly adaptively rotated along the energy-saving temperature control direction while maintaining the principal components of the original parameters. On the one hand, the external product of local covariance is used to lock in the structural changes caused by battery aging and sensor drift, preventing dimensionality reduction distortion. On the other hand, the gradient of the temperature-sensitive projection coefficient and the square of the cooling benefit covariance directly rewards the projection direction with high energy saving and high cooling. Combined with orthogonalization and deviating from the initial value penalty, large oscillations and overfitting are avoided. Attached Figure Description

[0018] Figure 1 This is a flowchart of steps S1-S4 in a real-time intelligent monitoring method for all parameters of a lithium battery according to an embodiment of the present invention.

[0019] Figure 2 This is a structural flowchart of a real-time intelligent monitoring system for all parameters of a lithium battery according to an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0021] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0022] Reference Figure 1 A real-time intelligent monitoring method for all parameters of a lithium battery includes steps S1-S4, as detailed below: S1: Real-time acquisition of all parameters of the lithium battery pack and Pearson correlation analysis with hot spot temperature, followed by normalization to construct feature vectors for each parameter.

[0023] The hotspot temperature of a lithium battery refers to the temperature of the smallest area inside the battery that is most prone to overheating. All parameters of the lithium battery pack are collected in real-time at a frequency of 1Hz. These parameters include, but are not limited to, the voltage of each individual battery cell, the total voltage, current, resistance, surface temperature, hotspot temperature, ambient temperature, and motor speed. First, historical data is statistically analyzed to calculate the Pearson correlation coefficient between each parameter and the hotspot temperature. The sign of the coefficient determines the positive or negative correlation between each parameter and the change in the lithium battery hotspot temperature. Then, the negative correlation parameters are inversely represented to align the parameters with the direction of temperature risk. Finally, Min-Max normalization is applied to the historical data to construct a normalized feature vector.

[0024] S2: A preset sliding window is used. The covariance matrix of the sliding window is modified based on the mutual information between each feature vector and the hot spot temperature and the Pearson correlation coefficient as the thermal sensitivity weight coefficient. A preset temperature sensitivity projection coefficient is used to maximize the sum of the covariance matrix and the preset temperature sensitivity projection coefficient and the square of the hot spot temperature covariance to obtain the projection matrix. The temperature sensitivity projection coefficient of each parameter feature is calculated based on the projection matrix.

[0025] The length of the sliding window is determined by the time it takes for the exhaust fan of the experimental explosion-proof equipment to achieve one air exchange when it is running at the lowest speed in historical data.

[0026] The formula for calculating the covariance matrix is: ; in, Represents the covariance matrix. Indicates the current moment. Represents a sliding window The starting time, for Time parameters eigenvectors, Represents a sliding window Intrinsic parameters The mean of the eigenvectors, Let be the deviation vector, representing Time parameters Feature vectors and parameters within the sliding window The difference between the means of the eigenvectors This indicates that the deviation vector is transposed. Used to create a diagonal matrix, where the elements on the diagonal are time intervals. Weights corresponding to each feature , This is the heat-sensitive weighting coefficient, used to represent the correlation strength between each parameter and the hotspot temperature.

[0027] The heat sensitivity weighting coefficient is calculated as follows: ; in, This is the heat-sensitive weighting coefficient. This represents the mutual information between the eigenvectors of each parameter and the hotspot temperature. The Pearson correlation coefficient represents the eigenvectors of each parameter and the hotspot temperature. The function implements automatic normalization and makes and The sum of is 1.

[0028] The objective function for constructing the projection matrix is ​​as follows: ; in, It is the covariance matrix via linear transformation The scalar variance obtained after mapping to a one-dimensional subspace The projected variance trace represents the variance against the scalar. Tracing, The temperature-sensitive projection coefficient represents the projection of the feature vector at each sampling time onto the direction. The resulting scalar score This indicates the measured hotspot temperature at the same moment. Describes the covariance operator. For temperature synchronization monitoring items, The adaptive balancing factor represents the ratio of the projection variance trace calculated from historical data to the temperature-synchronized monitoring term, with the aim of achieving a temperature-sensitive projection coefficient. and hot spot temperature To achieve balance and eliminate the influence of dimensions It is a constraint that makes the projection matrix vectors orthogonal. Represented by projection matrix To optimize the quantity, we need to find the maximum value of the objective function, that is, the maximum value of the objective function that satisfies all constraints. Searching in the expression The largest projection matrix.

[0029] Based on the final projection matrix, the temperature-sensitive projection coefficients of various parameters of the lithium battery at the current moment are calculated. .

[0030] S3: Using the temperature-sensitive projection coefficient as a single control input, the exhaust motor speed is linearly and steplessly adjusted. The hot spot temperature corresponding to the lowest speed of the exhaust motor is used as the reference temperature control. The positive difference between the reference temperature control and the measured temperature control within the sliding window is taken as the net cooling benefit. The cooling benefit is calculated based on the net cooling benefit and the ratio of the current speed of the exhaust motor to the maximum speed.

[0031] The formula for calculating the speed of the exhaust motor is as follows: ; in, This indicates the adjusted exhaust fan motor speed. This indicates the minimum speed of the exhaust fan motor. This indicates the maximum permissible speed of the exhaust fan motor. This is the temperature-sensitive projection coefficient.

[0032] After adjusting the exhaust fan motor speed, a new segment of real-time parameter data is extracted from the current moment according to the sliding window length in step S2. This serves two purposes: firstly, to implement delayed control and avoid frequent and excessively rapid adjustments to the motor speed; and secondly, to provide sufficient data support for further optimization of the full parameter dimensionality reduction process. Simultaneously, the cooling benefit at each moment is calculated. The formula is as follows: ; in, Indicates the cooling benefit. For baseline temperature control, this represents the normalized mean of the hot spot temperatures corresponding to the eigenvectors of each parameter when the exhaust fan is running at its lowest speed. For real-time temperature control, the temperature of the hotspot is displayed as the temperature collected in real time. This indicates that the positive value of the difference between the reference temperature control and the measured temperature control is taken, that is, in At that time, temperature control was ineffective. This is the current speed of the exhaust fan motor. This is the maximum permissible speed of the exhaust fan motor. This indicates the relative rotational speed, which effectively reflects the relative energy consumption of the exhaust fan motor. The larger the exhaust fan, the higher its energy consumption.

[0033] S4: Using the cooling benefit as a monitoring signal, the projection matrix is ​​iteratively updated through the gradient ascent method to keep the temperature-sensitive projection coefficient aligned with the optimal direction of energy-saving temperature control until convergence.

[0034] Based on the newly extracted lithium battery parameter sequence within the sliding time window described in step S3 and the calculated cooling benefits at each moment... The projection matrix is ​​iteratively updated using the gradient ascent method, and the formula is as follows: ; in, This represents the updated projection matrix. To orthogonalize the projection, ensure that the updated projection matrix still satisfies , This represents the projection matrix before the update. This represents the feature vector of the lithium battery parameters at the last moment before the sliding window is truncated. for Reconstructed vector after dimensionality reduction and then dimensionality increase ,express Using only the selected principal component directions The original data form that can be recovered is used to calculate the reconstruction error and assess the degree to which the dimensionality reduction space retains the original information. This represents the learning rate of the unsupervised term. Indicates the learning rate of the supervised term. This represents the cooling gain at the last moment before the sliding window is captured.

[0035] For unsupervised items, ensure The main information of the original parameters is still retained. The supervised term represents the gradient calculation based on the cooling benefit. Adjust the temperature-sensitive projection coefficient after dimensionality reduction. Further adapting to energy-saving temperature control requirements, gradient calculation ,in, This represents the mean of the parameter eigenvectors within the sliding window. This represents the average cooling benefit within the sliding window.

[0036] , Indicates the number of times the projection matrix has been updated. The decay factor representing the learning rate of the unsupervised term decreases as the number of iterations of the projection matrix updates increases, and the degree of decrease decreases with each iteration, meaning it becomes increasingly fine-tuned over time. This is the initial value for the unsupervised learning rate. The settings can be preset by staff according to the specific implementation situation. In this embodiment, The value can be 0.01.

[0037] ;in, For the supervised learning rate, The initial value for the learning rate of the supervised term. This ensures that the initial value of the supervised learning rate is lower than the initial value of the unsupervised learning rate, thus avoiding over-reliance on the supervised term. This represents the cooling benefit at the latest moment in the sliding window. This represents the standard deviation of cooling gains over all moments within the sliding window, indicating when cooling gains are better and volatility is lower. The value is larger. The larger the value, the stronger the enhancement. The connection, This represents the projection matrix before the update. This represents the initial projection matrix. This indicates that the part within the parentheses is subjected to a natural exponential function operation; the greater the difference between the two, the more exponential the function becomes. The smaller the value, the less likely the direction of the projection matrix adjustment will deviate too much.

[0038] Update projection matrix Then, based on the new projection matrix, the entire parameter is subjected to persistent dimensionality reduction, and the motor speed is intelligently controlled in step S3 of the loop. The new projection matrix is ​​then updated until the two consecutive nearest iterations are reached. The update of the projection matrix stops when the difference is less than a preset empirical threshold. The preset threshold can be any value. (Scientific notation, 1 multiplied by 10 to the power of negative 6).

[0039] The system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the real-time intelligent monitoring method for all parameters of a lithium battery according to the first aspect of the present invention.

[0040] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0041] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for real-time intelligent monitoring of all parameters of a lithium battery, characterized in that, include: Real-time acquisition of all parameters of the lithium battery pack and Pearson correlation analysis with hot spot temperature are performed, and normalization is performed to construct feature vectors for each parameter; A sliding window is preset. The covariance matrix of the sliding window is corrected based on the mutual information between each feature vector and the hot spot temperature and the Pearson correlation coefficient as the heat sensitivity weight coefficient. A temperature sensitivity projection coefficient is preset. The sum of the covariance matrix and the preset temperature sensitivity projection coefficient and the square of the hot spot temperature covariance is maximized to obtain the projection matrix. The temperature sensitivity projection coefficient of each parameter feature is calculated based on the projection matrix. The temperature-sensitive projection coefficient is used as the single control input to linearly and steplessly adjust the speed of the exhaust motor. The hot spot temperature corresponding to the lowest speed of the exhaust motor is used as the reference temperature control. The positive difference between the reference temperature control and the measured temperature control within the sliding window is taken as the net cooling benefit. The cooling benefit is calculated based on the net cooling benefit and the ratio of the current speed of the exhaust motor to the maximum speed. Using the cooling benefit as a monitoring signal, the projection matrix is ​​iteratively updated through the gradient ascent method, so that the temperature-sensitive projection coefficient is continuously aligned with the optimal direction of energy-saving temperature control until convergence.

2. The method for real-time intelligent monitoring of all parameters of a lithium battery according to claim 1, characterized in that, The heat sensitivity weighting coefficient is calculated as follows: Calculate the product of the mutual information between the eigenvectors of each parameter and the hotspot temperature and the Pearson correlation coefficient, and normalize it to obtain the heat sensitivity weighting coefficient.

3. The method for real-time intelligent monitoring of all parameters of a lithium battery according to claim 1, characterized in that, The method for adjusting the speed of the exhaust fan motor is as follows: The difference between the maximum and minimum speeds of the exhaust motor is calculated, multiplied by the temperature-sensitive projection coefficient, and then added to the minimum speed of the exhaust motor to obtain the adjusted speed of the exhaust motor.

4. The method for real-time intelligent monitoring of all parameters of a lithium battery according to claim 1, characterized in that, The cooling benefit is calculated as follows: The difference between the adjusted speed of the exhaust fan motor and its maximum speed is calculated as the relative speed. The relative speed is then subtracted from 1 and multiplied by the net cooling benefit to obtain the cooling benefit.

5. The method for real-time intelligent monitoring of all parameters of a lithium battery according to claim 1, characterized in that, The method for updating the projection matrix is ​​as follows: The reconstructed vector is obtained by reducing the dimensionality of the feature vector based on the parameters and then increasing the dimensionality. The error between the feature vector of the parameters at the current time and the reconstructed vector is calculated to obtain the reconstruction error vector. After forming a local covariance outer product, it is pushed along the projection direction as an unsupervised term to update the projection matrix. Calculate the square of the covariance between the temperature-sensitive projection coefficient and the cooling benefit, and perform matrix differentiation as a supervision term for updating the projection matrix; Using the cooling benefit as a monitoring signal, a decaying gradient ascent is performed on the projection matrix. The unsupervised term maintains the direction of the principal components, while the gradient of the supervised term is fine-tuned. Combined with orthogonalization and penalty for deviation from the initial value, the process iterates until convergence.

6. The method for real-time intelligent monitoring of all parameters of a lithium battery according to claim 5, characterized in that, An initial value for the unsupervised learning rate is preset. The square root of the number of iterations incremented by 1 is taken as the decay factor. The ratio of the initial value of the unsupervised learning rate to the decay factor is calculated to obtain the adaptive unsupervised learning rate and dynamically scale the step size of the unsupervised term.

7. The method for real-time intelligent monitoring of all parameters of a lithium battery according to claim 5, characterized in that, The initial value of the supervised learning rate is preset. The ratio of the cooling gain at the current moment to the standard deviation of all cooling gains within the sliding window is calculated as the signal-to-noise ratio gain. The square of the difference between the current projection matrix and the initial projection matrix is ​​calculated and subjected to a decay exponential function operation. This square is then multiplied by the initial value of the supervised learning rate and the signal-to-noise ratio gain to obtain the adaptive supervised learning rate and dynamically scale the step size of the supervised terms.

8. The method for real-time intelligent monitoring of all parameters of a lithium battery according to claim 1, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the real-time intelligent monitoring method for all parameters of a lithium battery according to any one of claims 1-7.

Images