Battery Defect Monitoring, Identification and Life Prediction Method

Abstract

The present invention discloses a method for monitoring, identifying defects and predicting the service life of a storage battery, which relates to the technical field of battery management, and includes: obtaining a set of operation monitoring indicators of the storage battery, randomly configuring weights, and constructing an initial weighted particle swarm; traversing the initial weighted particle swarm, training a storage battery life prediction model, and obtaining the verification accuracy rate of the initial weighted particle swarm; when the accuracy rates are all less than or equal to the convergence accuracy rate, using the verification accuracy rate of the initial weighted particle swarm as the fitness function, performing particle optimization to obtain an updated particle set, and training the storage battery life prediction model; when any particle of the verification accuracy rate of the initial weighted particle swarm is greater than the convergence accuracy rate, outputting the corresponding storage battery life prediction model, and performing the tasks of monitoring, identifying defects and predicting the service life of the storage battery. The present invention solves the technical problem that the prior art cannot accurately evaluate the operation state and remaining service life of the storage battery, and achieves the technical effect of accurately predicting the defects and remaining service life of the storage battery.

Description

Technical Field

[0001] The present invention relates to the technical field of battery management, and particularly to a method for monitoring and identifying battery defects and predicting battery life. Background Art

[0002] With the wide application of energy storage technology, the application scale of storage batteries is continuously expanding in the fields of new energy power generation, communication base stations, and data centers. However, during long-term operation, storage batteries may be affected by various factors, such as charge and discharge frequency, operating environment temperature, internal resistance change, etc., resulting in performance degradation and shortened life. Therefore, how to accurately monitor the operating state of storage batteries and predict their remaining service life has become an important technical requirement for ensuring the safe and stable operation of storage batteries.

[0003] Existing technologies usually evaluate the state of storage batteries using a single parameter (such as voltage or internal resistance), without fully considering the comprehensive impact of multi-dimensional operating parameters on life prediction. This method not only has insufficient monitoring accuracy but also cannot comprehensively reflect the health state of storage batteries. In addition, some methods rely on fixed models for life prediction and lack the dynamic adaptability to changes in the operating state of storage batteries, resulting in a large deviation between the life prediction results and the actual situation and being difficult to meet the requirements of complex application scenarios. Summary of the Invention

[0004] The present application provides a method for monitoring and identifying battery defects and predicting battery life, which is used to solve the technical problem that the existing technology cannot accurately evaluate the operating state and remaining service life of storage batteries.

[0005] In view of the above problems, the present application provides a method for monitoring and identifying battery defects and predicting battery life.

[0006] The present application provides a method for monitoring and identifying battery defects and predicting battery life, and the method includes:

[0007] Obtain a set of battery operation monitoring indicators; randomly configure weights for the set of battery operation monitoring indicators to construct an initial weight particle swarm, where any particle in the initial weight particle swarm represents a set of battery operation monitoring indicator weights, and the set of battery operation monitoring indicator weights corresponds one-to-one with the set of battery operation monitoring indicators; traverse the initial weight particle swarm, and based on the set of battery operation monitoring indicators, train a battery life prediction model to obtain the verification accuracy rate of the initial weight particle swarm, where each particle in the initial weight particle swarm corresponds one-to-one with the verification accuracy rate of the initial weight particle swarm; when the verification accuracy rates of the initial weight particle swarm are all less than or equal to the convergence accuracy rate, use the verification accuracy rate of the initial weight particle swarm as the fitness function to perform particle optimization on the initial weight particle swarm to obtain an updated particle set; based on the updated particle set, train a battery life prediction model based on the set of battery operation monitoring indicators; when any particle of the verification accuracy rate of the initial weight particle swarm is greater than the convergence accuracy rate, output the battery life prediction model corresponding to the particle and perform the tasks of battery defect monitoring and identification and life prediction.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] This application obtains a set of battery operation monitoring indicators; randomly configures weights for the set of battery operation monitoring indicators to construct an initial weight particle swarm, where any particle in the initial weight particle swarm represents a set of battery operation monitoring indicator weights, and the set of battery operation monitoring indicator weights corresponds one-to-one with the set of battery operation monitoring indicators; traverse the initial weight particle swarm, and based on the set of battery operation monitoring indicators, train a battery life prediction model to obtain the verification accuracy rate of the initial weight particle swarm, where each particle in the initial weight particle swarm corresponds one-to-one with the verification accuracy rate of the initial weight particle swarm; when the verification accuracy rates of the initial weight particle swarm are all less than or equal to the convergence accuracy rate, use the verification accuracy rate of the initial weight particle swarm as the fitness function to perform particle optimization on the initial weight particle swarm to obtain an updated particle set; based on the updated particle set, train a battery life prediction model based on the set of battery operation monitoring indicators; when any particle of the verification accuracy rate of the initial weight particle swarm is greater than the convergence accuracy rate, output the battery life prediction model corresponding to the particle and perform the tasks of battery defect monitoring and identification and life prediction. This invention solves the technical problem that the prior art cannot accurately evaluate the operation state and remaining service life of a battery. By collecting battery operation monitoring indicators through a sensor module, constructing an initial weight particle swarm and optimizing the weight configuration, training a life prediction model using a neural network, evaluating the particle effect through the verification accuracy rate, and outputting the best model after iterative optimization, the technical effect of accurate prediction of battery defect monitoring and remaining life is achieved. Brief Description of the Drawings

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0011] Figure 1 It is a schematic flow chart of the method for monitoring, identifying and predicting the life of battery defects provided by the embodiments of the present application;

[0012] Figure 2 It is a schematic flow chart of obtaining the verification accuracy rate of the initial weight particle swarm in the method for monitoring, identifying and predicting the life of battery defects provided by the embodiments of the present application. Detailed Embodiments

[0013] The present application provides a method for monitoring, identifying and predicting the life of battery defects, which is used to solve the technical problem that the prior art cannot accurately evaluate the operating state and remaining service life of the battery. By collecting the battery operation monitoring indicators through the sensor module, constructing the initial weight particle swarm and optimizing the weight configuration, training the life prediction model using the neural network, evaluating the particle effect through the verification accuracy rate, and iteratively optimizing to output the best model, the technical effect of accurately predicting the battery defects and the remaining life is achieved.

[0014] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0015] It should be noted that any variations of the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units does not necessarily have to be limited to those clearly listed steps or units, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products or devices.

[0016] Embodiments, as Figure 1 shown, the present application provides a method for monitoring, identifying and predicting the life of battery defects, and the method includes:

[0017] Step S100: Obtain the set of battery operation monitoring indicators.

[0018] In the embodiments of the present application, data collection is first performed through a sensor module installed in the battery pack and its operating environment. The sensor module includes a voltage sensor, an internal resistance sensor, a temperature sensor, a charge and discharge current sensor, etc., and real-time monitors key physical parameters of the battery operation. The battery operation monitoring indicators refer to a specific set of parameters collected by these sensors, including but not limited to battery voltage, internal resistance value, operating temperature, charge and discharge current, and depth of discharge.

[0019] The collected data is stored and processed to form a historical database. Finally, by extracting these stored historical data, a complete set of battery operation monitoring indicators is obtained.

[0020] Step S200: Randomly configure weights for the set of battery operation monitoring indicators to construct an initial weight particle swarm, where any particle in the initial weight particle swarm represents a set of battery operation monitoring indicator weights, and the set of battery operation monitoring indicator weights corresponds one-to-one with the set of battery operation monitoring indicators.

[0021] In the embodiments of the present application, for each indicator (such as voltage, internal resistance, temperature, charge and discharge current) in the set of battery operation monitoring indicators, a random proportion generation method is used to assign a weight value to each indicator, and it is ensured that these weight values satisfy the constraint condition that the sum is 1. In this way, a set of weight values is formed, and each weight value corresponds one-to-one with a monitoring indicator in the set, indicating the importance of the indicator in the subsequent life prediction task.

[0022] Each set of such assigned weight values constitutes a particle. For example, assume that the set of battery operation monitoring indicators includes four indicators: voltage, internal resistance, temperature, and current. Then a particle may be represented as 0.3, 0.4, 0.2, 0.1, where 0.3 represents the weight of voltage, 0.4 represents the weight of internal resistance, and 0.2 and 0.1 correspond to the weights of temperature and current respectively. The significance of the particle representation is that it comprehensively reflects the importance distribution of each indicator in the set of battery operation monitoring indicators in the form of weights.

[0023] By repeating the above steps, multiple particles are generated to construct an initial weight particle swarm. Each particle in the particle swarm represents a complete set of monitoring indicator weight configurations and corresponds one-to-one with each indicator in the set of battery operation monitoring indicators.

[0024] In this particle swarm, each particle corresponds to a complete set of monitoring indicator weight configurations, and these weight values strictly satisfy the one-to-one correspondence relationship, which not only reflects the relative importance of each monitoring indicator in the set but also provides a diverse set of initial weight configuration schemes for subsequent model training and optimization.

[0025] Step S300: Traverse the initial weight particle swarm, and based on the set of battery operation monitoring indicators, train a battery life prediction model to obtain the verification accuracy rate of the initial weight particle swarm, where each particle in the initial weight particle swarm corresponds one-to-one to the verification accuracy rate of the initial weight particle swarm.

[0026] In the embodiments of the present application, each particle in the initial weight particle swarm is extracted in sequence, and each particle contains a set of weight values that correspond one-to-one to the set of battery operation monitoring indicators. Based on the extracted particle weight set, configure the neural network topology of the battery life prediction model, where the input nodes correspond to each indicator in the monitoring indicator set, and the output node is used to predict the battery life, to obtain the life prediction model after weight configuration. Subsequently, use the data in the set of battery operation monitoring indicators as input features, and combine with the life identification data for supervised learning to train the model to fit the actual life data, and obtain the trained life prediction model corresponding to the current particle. Among them, the set of battery operation monitoring indicators and the corresponding life identification data are collected from the historical database, and the life identification data represents the remaining available duration of the battery in the current state.

[0027] Next, use an independent validation data set to evaluate the performance of the model and calculate its verification accuracy rate to reflect the prediction effect of the current particle weight configuration. Repeat the above steps for all particles in the particle swarm, and finally obtain a set of verification accuracy rates including all particles, that is, the verification accuracy rate of the initial weight particle swarm. This set of verification accuracy rates realizes the one-to-one correspondence between the particles and their weight configuration effects.

[0028] Further, as Figure 2 shown, in the method provided by the embodiments of the application, traversing the initial weight particle swarm, and based on the set of battery operation monitoring indicators, training a battery life prediction model to obtain the verification accuracy rate of the initial weight particle swarm further includes:

[0029] According to the initial weight particle swarm, extract a first particle, where the first particle has a first set of battery operation monitoring indicator weights; configure the topology of the battery life prediction model, where the topology of the battery life prediction model is a neural network topology, and the input nodes of the neural network topology correspond one-to-one to each node in the set of battery operation monitoring indicators, and the input node of the neural network topology is one and is used to output the predicted life; according to the first set of battery operation monitoring indicator weights, perform weight configuration on the input nodes of the topology of the battery life prediction model to obtain an initial battery life prediction model; according to the set of battery operation monitoring indicators, train the initial battery life prediction model to obtain a first particle battery life prediction model; verify the first particle battery life prediction model, and add the obtained verification accuracy rate to the training accuracy rate of the initial weight particle swarm.

[0030] In the embodiment of the present application, first, a first particle is extracted from the initial weight particle swarm. The first particle contains a first set of weights for battery operation monitoring indicators. Each weight value in this set of weights corresponds one-to-one with each indicator in the set of battery operation monitoring indicators (such as voltage, internal resistance, temperature, current, etc.), indicating the relative importance of each monitoring indicator in the prediction task. The first set of weights for battery operation monitoring indicators is obtained by extracting the first particle in order through the list indexing method.

[0031] Next, according to the first set of weights for battery operation monitoring indicators, the topology of the battery life prediction model is configured. The model topology is a neural network structure, where the number of input nodes is the same as the number of indicators in the set of battery operation monitoring indicators. Each input node corresponds to a monitoring indicator, and there is a unique output node for predicting the service life of the battery. The neural network topology is constructed by a multi-layer perceptron (MLP), and the input layer nodes, hidden layer structure, and output node are set to obtain the complete topology of the battery life prediction model.

[0032] Then, according to the first set of weights for battery operation monitoring indicators, weight configuration is performed on the input nodes of the battery life prediction model topology. The weight configuration uses the linear weighting method, multiplying each weight value in the set of weights in the first particle by the input feature corresponding to the monitoring indicator. For example, if the first set of weights for battery operation monitoring indicators is 0.3, 0.4, 0.2, 0.1, then the input features voltage, internal resistance, temperature, and current are weighted as 0.3, 0.4, 0.2, and 0.1 respectively. After completing the weight configuration, the initial battery life prediction model based on the first set of weights for battery operation monitoring indicators is obtained.

[0033] Next, using the data in the set of battery operation monitoring indicators (such as voltage, internal resistance, temperature, and current data in time series form) and the corresponding life identification data (such as actual service life), the initial battery life prediction model is trained. The mean squared error (MSE) is used as the loss function in the training process, and the model parameters are optimized through the batch gradient descent algorithm to minimize the gap between the predicted life and the actual life. After completing the training, the first particle battery life prediction model based on the first set of weights for battery operation monitoring indicators is obtained.

[0034] Subsequently, the first particle battery life prediction model is verified using an independent validation data set, where the independent validation data set is prepared in advance. By inputting the validation data set into the model, the error between the predicted value and the actual value is calculated, and the validation accuracy is calculated based on the error result. The validation accuracy quantifies the adaptability of the first set of weights for battery operation monitoring indicators in the life prediction task. After completing the validation, the validation accuracy of the first particle battery life prediction model is obtained.

[0035] Finally, record the verification accuracy rate of the first particle battery life prediction model and add it to the set of verification accuracy rates of the initial weight particle swarm.

[0036] Furthermore, in the method provided by the application embodiment, when training the initial battery life prediction model according to the set of battery operation monitoring indexes to obtain the first particle battery life prediction model, it further includes:

[0037] Using the set of battery operation monitoring indexes as input data and the battery life data as output data, collect the set of characteristic values of battery operation monitoring indexes and the battery life identification data; using the battery life identification data as supervision and the set of characteristic values of battery operation monitoring indexes as input, train the first-level first particle battery life prediction model; when the data volume of the first-level loss data set that does not meet the first accuracy threshold of the first-level first particle battery life prediction model is less than the data volume threshold, set the first-level first particle battery life prediction model as the first particle battery life prediction model.

[0038] In the embodiment of the present application, first, use the set of battery operation monitoring indexes as input data, and this set includes the characteristic values of monitoring indexes such as voltage, internal resistance, temperature, charge and discharge current collected from sensors; at the same time, use the battery life data as output data, and this data represents the actual battery life value corresponding to each group of monitoring indexes (characterized by the number of cycles or working hours, for example). Through the integration of historical data, collect and construct the set of characteristic values of battery operation monitoring indexes and the corresponding battery life identification data.

[0039] Then, use the supervised learning method, use the battery life identification data as the supervision signal, and use the set of characteristic values of battery operation monitoring indexes as the input features to start training the first-level first particle battery life prediction model. The topology of this model is constructed based on a neural network, where the number of nodes in the input layer corresponds to the number of features in the monitoring index set, and the output layer is a single node for predicting the battery life value. During the training process, the mean square error (MSE) is used as the loss function, and the model parameters are adjusted through the gradient descent optimization algorithm to minimize the gap between the predicted life and the actual life. After training, the first-level first particle battery life prediction model is obtained.

[0040] Subsequently, the life prediction model of the first-stage first-particle battery is verified, and the prediction accuracy of the model is calculated using an independent verification data set. By counting the scale of the first-stage loss data set, that is, the number of data points in the verification data set where the model prediction error exceeds the allowable range. Compare this loss data volume with a preset data volume threshold. If the data volume of the first-stage loss data set is less than the preset data volume threshold, it indicates that the prediction performance of the model has met the requirements in most cases, and the error points only account for a small range, and the overall performance of the model is considered acceptable. In this case, the life prediction model of the first-stage first-particle battery is set as the final life prediction model of the first-particle battery.

[0041] Furthermore, the method provided by the application embodiment further includes:

[0042] When the data volume of the first-stage loss data set of the life prediction model of the first-stage first-particle battery that does not meet the first accuracy threshold is greater than or equal to the data volume threshold, extract the life deviation vector set of the life prediction model of the first-stage first-particle battery; perform systematic error analysis on the life deviation vector set to obtain a systematic error vector; use the opposite number of the systematic error vector as supervised data, and use the set of battery operation monitoring index eigenvalue as input to train the second-stage first-particle battery life prediction model; until the data volume of the N-stage loss data set of the N-stage first-particle battery life prediction model that does not meet the first accuracy threshold is less than the data volume threshold, add and fully connect the outputs of the life prediction model of the first-stage first-particle battery, the second-stage first-particle battery life prediction model until the N-stage first-particle battery life prediction model to obtain the life prediction model of the first-particle battery.

[0043] In the embodiment of the present application, first, the life deviation vector set is extracted from the verification data set. By traversing the verification data set point by point to calculate the difference between the model prediction value and the actual life identification data, a deviation vector set is generated. For example, if the model predicts a life of 10 years and the actual life is 9 years, the deviation is 1. After extraction, a life deviation vector set reflecting the model error distribution is obtained.

[0044] Then, systematic error analysis is performed on the extracted life deviation vector set. By calculating the mean value of the deviation vectors, the central tendency and systematic characteristics of the model error are identified. For example, if the mean values are 0.2, -0.1, 0.15, it indicates that the prediction results of the model on some monitoring indicators are generally on the high or low side. After the analysis is completed, a systematic error vector describing the direction and magnitude of the model error is obtained.

[0045] Then, take the opposite of the system error vector to generate the supervised data for correcting the model. This supervised data reflects the direction and magnitude that need to be adjusted. For example, if the system error vector is 0.2, -0.1, 0.15, its opposite is -0.2, 0.1, -0.15, which is used to guide the model to correct the error. Through this step, the opposite of the system error vector for optimization is obtained.

[0046] Next, use the opposite of the system error vector as the supervised data, and combine it with the set of characteristic values of the battery operation monitoring indicators as the input to train the second-level first particle battery life prediction model. Through the supervised learning method, use the mean square error (MSE) as the loss function to optimize the model parameters to minimize the error between the output and the opposite. After training, the second-level first particle battery life prediction model for correcting the system error of the first-level model is obtained.

[0047] If the accuracy of the second-level first particle battery life prediction model still does not meet the accuracy threshold, and the data volume of the loss data set is greater than or equal to the data volume threshold, repeat the above steps. Extract a new set of deviation vectors, conduct system error analysis, generate the opposite of the new system error vector, and continue to train the third-level first particle battery life prediction model. Through multiple iterative optimizations, continuously correct the system error until the data volume of the loss data set of a certain level of model is less than the data volume threshold.

[0048] Finally, sum and integrate the outputs of all levels of models (such as the first level, the second level up to the Nth level) through a fully connected layer. The fully connected layer combines the correction results of each level of model into the final prediction value through a linear combination. After integration, the first particle battery life prediction model that has been comprehensively optimized is obtained.

[0049] Step S400: When the verification accuracies of the initial weight particle swarm are all less than or equal to the convergence accuracy, use the verification accuracies of the initial weight particle swarm as the fitness function to perform particle optimization on the initial weight particle swarm to obtain an updated particle set.

[0050] In the embodiment of the present application, compare the verification accuracies of the initial weight particle swarm with the preset convergence accuracy. When the verification accuracies of the initial weight particle swarm are all less than or equal to the convergence accuracy, perform particle optimization on the initial weight particle swarm to obtain an updated particle set.

[0051] Specifically, first, pairwise evaluation is performed on the distances between particles in the initial weight particle swarm, calculating the distances between each pair of particles to form a set of particle distance evaluation values. Then, based on the particle distance threshold, the particles are clustered according to the distance relationship to obtain multiple groups of initial weight particles. Next, according to the verification accuracy of each group of particles, they are sorted from high to low to generate multiple sets of sorting results of the initial weight particles. Finally, based on the sorting results, the particle positions are updated using a collaborative optimization strategy to generate a set of updated particles.

[0052] Further, in the method provided by the application embodiment, when the verification accuracies of the initial weight particle swarm are all less than or equal to the convergence accuracy, using the verification accuracy of the initial weight particle swarm as the fitness function, particle optimization is performed on the initial weight particle swarm to obtain a set of updated particles, which further includes:

[0053] Perform pairwise evaluation on the distances between particles in the initial weight particle swarm to obtain a set of particle distance evaluation values; based on the set of particle distance evaluation values, perform clustering based on the particle distance threshold to obtain multiple groups of initial weight particles; according to the verification accuracy of the initial weight particle swarm, sort the multiple groups of initial weight particles from large to small in terms of verification accuracy to obtain multiple sets of sorting results of the initial weight particles; according to the multiple sets of sorting results of the initial weight particles, perform collaborative particle optimization on the initial weight particle swarm to obtain a set of updated particles.

[0054] In the embodiment of the present application, first, pairwise evaluation is performed on the distances between particles in the initial weight particle swarm. Two particles are randomly extracted from the particle swarm, and their first set of weights of the battery operation monitoring indicators and the second set of weights of the battery operation monitoring indicators are respectively obtained, and then the Euclidean distance between the two sets of particle weights is calculated. The calculation result is used as the particle distance evaluation value and added to the set of particle distance evaluation values. By comparing all particle combinations pairwise, a complete set of particle distance evaluation values is finally obtained.

[0055] Then, based on the set of particle distance evaluation values, the particle swarm is clustered using a preset particle distance threshold as the standard, and the particles with closer distances are grouped together to form multiple groups of initial weight particles. For example, if the distance threshold is 0.1, all particles with distances less than 0.1 are grouped into one group. After clustering, multiple groups of initial weight particles are obtained.

[0056] Then, according to the verification accuracy of the initial weight particle swarm, each group of initial weight particles is sorted from high to low in terms of verification accuracy to generate multiple sets of sorting results of the initial weight particles. This sorting result identifies the particles with better performance in each group and provides a reference for guiding particle optimization. For example, the sorting of a certain group may be particle 1, particle 2, particle 3, where particle 1 has the highest verification accuracy.

[0057] Subsequently, based on the sorting results of multiple groups of initial weight particles, collaborative particle optimization is performed on the particle swarm. Randomly extract the sorting results of two groups of particles, respectively obtain the first particles of the first group and the second group, and use the first particles to guide the last particles of other groups to mutate. For example, the first particle of the first group guides the last particle of the second group, and updates the particle position by reducing the particle distance evaluation value between the two, generating the first particle update result. Similarly, the first particle of the second group guides the last particle of the first group to generate the second particle update result. All update results are added to the updated particle set.

[0058] Finally, through the above steps, the optimization of the particle swarm is completed, and an updated particle set is obtained.

[0059] Furthermore, in the method provided by the application embodiment, when pairwise evaluating the particle distances of the initial weight particle swarm to obtain a particle distance evaluation value set, it further includes:

[0060] Randomly obtain the first particle and the second particle of the initial weight particle swarm. Among them, the first particle has a first set of weights for battery operation monitoring indicators, and the second particle has a second set of weights for battery operation monitoring indicators; calculate the Euclidean distance between the first set of weights for battery operation monitoring indicators and the second set of weights for battery operation monitoring indicators, which is set as the first particle distance evaluation value, and add it to the particle distance evaluation value set.

[0061] In the embodiment of the present application, first, randomly extract two particles from the initial weight particle swarm, which are respectively called the first particle and the second particle. Each particle contains a set of weight values corresponding to the battery operation monitoring indicator set, and these weight values respectively reflect the importance of each monitoring indicator in the model.

[0062] Then, compare the weight sets of these two particles by the Euclidean distance method. The Euclidean distance is a commonly used method for measuring the distance between points in a multi-dimensional space, and it can effectively represent the difference in weight distribution between two particles. When calculating, compare the differences in the corresponding weight values of each monitoring indicator between the first particle and the second particle one by one, and synthesize these differences to obtain the distance size between the particles. This distance is called the first particle distance evaluation value, indicating the similarity degree of the two particles in weight configuration.

[0063] Finally, record the calculated first particle distance evaluation value and add it to the particle distance evaluation value set.

[0064] Furthermore, in the method provided by the application embodiment, when performing collaborative particle optimization on the initial weight particle swarm according to the sorting results of the multiple groups of initial weight particles to obtain an updated particle set, it further includes:

[0065] Randomly extract the first group of initial weight particle sorting results and the second group of initial weight particle sorting results from the multiple groups of initial weight particle sorting results; obtain the first leading initial weight particle of the first group of initial weight particle sorting results, guide a preset number of trailing initial weight particles of the second group of initial weight particle sorting results to mutate, reduce the particle distance evaluation value from the first leading initial weight particle, and obtain the first particle update result; obtain the second leading initial weight particle of the second group of initial weight particle sorting results, guide a preset number of trailing initial weight particles of the first group of initial weight particle sorting results to mutate, reduce the particle distance evaluation value from the second leading initial weight particle, and obtain the second particle update result; add the first particle update result and the second particle update result to the updated particle set.

[0066] In the embodiment of the present application, first, two groups are randomly extracted from multiple groups of initial weight particle sorting results, which are respectively called the first group of initial weight particle sorting results and the second group of initial weight particle sorting results. Select the leading particle of the first group of sorting results, that is, the first leading initial weight particle, as the optimization target.

[0067] Subsequently, according to the random value of the weight dimension, determine the number of dimensions k for mutation, and use this value to mutate a preset number of trailing particles in the second group of sorting results. The mutation is achieved by adjusting the partial weight values of these trailing particles to gradually approach the first leading initial weight particle, thereby reducing the particle distance evaluation value. After the mutation is completed, the first particle update result is generated.

[0068] Then, select the leading particle of the second group of initial weight particle sorting results, that is, the second leading initial weight particle, and perform a similar mutation operation on a preset number of trailing particles in the first group of sorting results. By adjusting the weight values, these particles gradually approach the second leading initial weight particle, thereby optimizing their configuration. After completion, the second particle update result is generated.

[0069] Finally, add the first particle update result and the second particle update result to the updated particle set to complete a process of collaborative particle optimization.

[0070] Further, in the method provided by the application embodiment, obtaining the first leading initial weight particle of the first group of initial weight particle sorting results, guiding a preset number of trailing initial weight particles of the second group of initial weight particle sorting results to mutate, reducing the particle distance evaluation value from the first leading initial weight particle, and obtaining the first particle update result further includes:

[0071] Based on the weight dimension, randomly obtain a value of k by random selection, where 1 ≤ k ≤ the weight dimension; according to the value of k, using the first initial weight particle as the target, mutate the preset number of trailing initial weight particles, reduce the particle distance evaluation value from the first initial weight particle, and obtain the first particle update result.

[0072] In the embodiment of the present application, first, according to the weight configuration characteristics of the particles, determine its weight dimension, that is, the number of weight values included in each particle. The weight dimension is derived from the number of features in the monitoring index set. For example, if the battery operation monitoring index set includes voltage, internal resistance, temperature, and current, then the weight dimension is 4. After determining the weight dimension, use a random number generation method (such as a uniform distribution random number generator) to randomly generate an integer value k within the range of 1 to the weight dimension. This value k represents the number of weight dimensions to be adjusted in the mutation operation. After this process is completed, the number k of adjustment dimensions for mutation is randomly obtained, laying a foundation for subsequent operations.

[0073] Then extract the first initial weight particle from the particle sorting result. This particle is the particle with the best performance in the current group, and its weight set (such as 0.3, 0.4, 0.2, 0.1) represents the optimal weight allocation scheme. This particle is selected as the target particle to guide the mutation of other particles. Through this operation, the target particle for mutation is clarified, providing a benchmark for the optimization of the trailing particles.

[0074] Subsequently, select a preset number of trailing particles (such as 3 particles) from the sorting result and mutate them. For each trailing particle, randomly select k dimensions (such as the 1st and 3rd dimensions) from its weight set, then calculate the weight difference between the trailing particle and the target particle in these dimensions, and gradually adjust these weight values according to a dynamic scaling factor (set to 0.5) to make it gradually approach the target particle. For example, if the weight set of a certain trailing particle is 0.2, 0.4, 0.3, 0.1, and the weight set of the target particle is 0.3, 0.4, 0.2, 0.1, and the 1st and 3rd dimensions are selected for mutation, then the mutated weight set is adjusted to 0.25, 0.4, 0.25, 0.1. Through this step of operation, the weight of each trailing particle gradually approaches the target particle in the selected dimensions, thereby reducing the distance between particles and optimizing the weight allocation.

[0075] After the mutation is completed, the new weight configurations of all trailing particles form an optimized particle set, which is called the first particle update result.

[0076] Step S500: Based on the updated particle set and the battery operation monitoring index set, train a battery life prediction model.

[0077] In the embodiment of the present application, based on the updated particle set and the battery operation monitoring index set, a battery life prediction model is trained. Specifically, using the particle weight configuration in the updated particle set, each index feature in the battery operation monitoring index set is weighted to highlight the relative importance of different monitoring indexes in life prediction. Subsequently, a battery life prediction model is constructed, where its input layer corresponds one-to-one with each monitoring index in the battery operation monitoring index set, and the output layer is used to predict the service life of the battery. By inputting historical monitoring data and life identification data, the model parameters are optimized in a supervised learning manner, and finally a battery life prediction model capable of accurately predicting the life based on the battery operation monitoring index set is obtained.

[0078] Step S600: When any particle of the verification accuracy rate of the initial weight particle swarm is greater than the convergence accuracy rate, output the battery life prediction model corresponding to the particle, and perform the battery defect monitoring, identification and life prediction tasks.

[0079] In the embodiment of the present application, first, the initial weight particle swarm is screened to find particles with a verification accuracy rate greater than the convergence accuracy rate. This process is the same as the calculation of the aforementioned verification accuracy rate. By applying the weight configuration of each particle to the model and calculating its prediction accuracy rate on the verification set. After screening, the qualified particles and their corresponding weight configurations are extracted. Then, the particle with the highest verification accuracy rate is selected from the qualified particles, and its optimal weight configuration is extracted. This operation is the same as the process of screening and sorting particles mentioned above. By sorting the verification accuracy rate from high to low, the optimal particle and its weight configuration are selected. Subsequently, the final battery life prediction model is constructed using the selected optimal weight configuration.

[0080] After the model is constructed, it enters the real-time execution stage of the battery defect monitoring, identification and life prediction tasks. Specifically, the real-time collected battery operation monitoring index data is input into the model. After weight weighting processing and model inference, the predicted life value of the battery is output. At the same time, combined with the anomaly detection algorithm, the life prediction result is analyzed in real time. If the predicted value is lower than the preset safety threshold, it is determined as a potential defect and an alarm is triggered, and a corresponding monitoring report is generated.

[0081] In the embodiment of the present application, in summary, the embodiment of the present application has at least the following technical effects:

[0082] This application obtains a set of battery operation monitoring indicators; randomly assigns weights to the set of battery operation monitoring indicators to construct an initial weight particle swarm, where any particle in the initial weight particle swarm represents a set of battery operation monitoring indicator weights, and the set of battery operation monitoring indicator weights corresponds one-to-one with the set of battery operation monitoring indicators; traverse the initial weight particle swarm, and based on the set of battery operation monitoring indicators, train a battery life prediction model to obtain the verification accuracy rate of the initial weight particle swarm, where each particle in the initial weight particle swarm corresponds one-to-one with the verification accuracy rate of the initial weight particle swarm; when the verification accuracy rates of the initial weight particle swarm are all less than or equal to the convergence accuracy rate, use the verification accuracy rate of the initial weight particle swarm as the fitness function to perform particle optimization on the initial weight particle swarm to obtain an updated particle set; based on the updated particle set, train a battery life prediction model based on the set of battery operation monitoring indicators; when any particle of the verification accuracy rate of the initial weight particle swarm is greater than the convergence accuracy rate, output the battery life prediction model corresponding to the particle and perform the tasks of battery defect monitoring and identification and life prediction. The present invention solves the technical problem that the prior art cannot accurately evaluate the operation state and remaining service life of a battery. By collecting battery operation monitoring indicators through a sensor module, constructing an initial weight particle swarm and optimizing the weight configuration, training a life prediction model using a neural network, evaluating the particle effect through the verification accuracy rate, and outputting the best model after iterative optimization, the technical effect of accurate prediction of battery defect monitoring and remaining life is achieved.

[0083] It should be noted that the above sequence of embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above specific embodiments of this specification have been described. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0084] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included within the protection scope of the present application.

[0085] This specification and the drawings are only exemplary descriptions of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A method for monitoring, identifying battery defects and predicting battery life, characterized in that Including: Obtain the set of battery operation monitoring indicators; Randomly configure weights for the set of battery operation monitoring indicators to construct an initial weight particle swarm. Among them, any particle in the initial weight particle swarm represents a set of battery operation monitoring indicator weights, and the set of battery operation monitoring indicator weights corresponds one-to-one with the set of battery operation monitoring indicators; Traverse the initial weight particle swarm, and based on the set of battery operation monitoring indicators, train the battery life prediction model to obtain the verification accuracy rate of the initial weight particle swarm. Among them, each particle in the initial weight particle swarm corresponds one-to-one with the verification accuracy rate of the initial weight particle swarm; When the verification accuracy rates of the initial weight particle swarm are all less than or equal to the convergence accuracy rate, use the verification accuracy rate of the initial weight particle swarm as the fitness function to perform particle optimization on the initial weight particle swarm to obtain an updated particle set; Based on the set of battery operation monitoring indicators, train the battery life prediction model according to the updated particle set; When any particle of the verification accuracy rate of the initial weight particle swarm is greater than the convergence accuracy rate, output the battery life prediction model of the corresponding particle and execute the battery defect monitoring and identification and life prediction tasks; The training of the battery life prediction model based on the set of battery operation monitoring indicators to obtain the verification accuracy rate of the initial weight particle swarm includes: Extract the first particle according to the initial weight particle swarm. Among them, the first particle has a first set of battery operation monitoring indicator weights; Configure the topology of the battery life prediction model. Among them, the topology of the battery life prediction model is a neural network topology. The input nodes of the neural network topology correspond one-to-one with each node of the set of battery operation monitoring indicators. The input node of the neural network topology is one and is used to output the predicted life; According to the first set of battery operation monitoring indicator weights, perform weight configuration on the input nodes of the battery life prediction model topology to obtain an initial battery life prediction model; According to the set of battery operation monitoring indicators, train the initial battery life prediction model to obtain the first particle battery life prediction model; Verify the first particle battery life prediction model and add the obtained verification accuracy rate to the verification accuracy rate of the initial weight particle swarm.

2. The method according to claim 1, characterized in that According to the set of battery operation monitoring indicators, training the initial battery life prediction model to obtain the first particle battery life prediction model includes: Using the set of battery operation monitoring indicators as input data and the battery life data as output data, collect the set of battery operation monitoring indicator characteristic values and battery life identification data; Using the battery life identification data as supervision and the set of battery operation monitoring indicator characteristic values as input, train the first-level first particle battery life prediction model; When the data volume of the first-level loss data set that does not meet the first accuracy threshold of the first-level first particle battery life prediction model is less than the data volume threshold, set the first-level first particle battery life prediction model as the first particle battery life prediction model.

3. The method according to claim 2, wherein It also includes: When the data volume of the first-level loss data set for which the first-level first particle battery life prediction model does not meet the first accuracy threshold is greater than or equal to the data volume threshold, extract the life deviation vector set of the first-level first particle battery life prediction model; Conduct a systematic error analysis on the life deviation vector set to obtain a systematic error vector; Use the opposite number of the systematic error vector as the supervised data, and use the set of battery operation monitoring index eigenvalue as the input to train the second-level first particle battery life prediction model; Until the data volume of the N-level loss data set for which the N-level first particle battery life prediction model does not meet the first accuracy threshold is less than the data volume threshold, perform a full connection of the outputs of the first-level first particle battery life prediction model, the second-level first particle battery life prediction model until the N-level first particle battery life prediction model to obtain the first particle battery life prediction model.

4. The method according to claim 1, wherein When the verification accuracies of the initial weight particle swarms are all less than or equal to the convergence accuracy, use the verification accuracies of the initial weight particle swarms as the fitness function to perform particle optimization on the initial weight particle swarms to obtain an updated particle set, including: Conduct pairwise evaluation of the particle distances of the initial weight particle swarms to obtain a set of particle distance evaluation values; Based on the set of particle distance evaluation values, perform clustering based on the particle distance threshold to obtain multiple groups of initial weight particles; According to the verification accuracies of the initial weight particle swarms, sort the multiple groups of initial weight particles from largest to smallest in terms of verification accuracy to obtain sorting results of multiple groups of initial weight particles; According to the sorting results of the multiple groups of initial weight particles, perform collaborative particle optimization on the initial weight particle swarms to obtain an updated particle set.

5. The method according to claim 4, characterized in that, Conduct pairwise evaluation of the particle distances of the initial weight particle swarms to obtain a set of particle distance evaluation values, including: Randomly obtain the first particle and the second particle of the initial weight particle swarms, where the first particle has a first set of battery operation monitoring index weights, and the second particle has a second set of battery operation monitoring index weights; Calculate the Euclidean distance between the first set of battery operation monitoring index weights and the second set of battery operation monitoring index weights, set it as the first particle distance evaluation value, and add it to the set of particle distance evaluation values.

6. The method according to claim 4, wherein According to the sorting results of the multiple groups of initial weight particles, perform collaborative particle optimization on the initial weight particle swarms to obtain an updated particle set, including: Randomly extract the first set of initial weight particle sorting results and the second set of initial weight particle sorting results from the sorting results of the multiple groups of initial weight particles; Obtain the first leading initial weight particle of the first set of initial weight particle sorting results, and guide a preset number of trailing initial weight particles of the second set of initial weight particle sorting results to mutate to reduce the particle distance evaluation value from the first leading initial weight particle to obtain a first particle update result; The second first initial weight particle that obtains the sorting result of the second group of initial weight particles guides the preset number of last initial weight particles in the sorting result of the first group of initial weight particles to mutate, reducing the particle distance evaluation value from the second first initial weight particle to obtain a second particle update result; Add the first particle update result and the second particle update result to the updated particle set.

7. The method according to claim 6, wherein The first first initial weight particle that obtains the sorting result of the first group of initial weight particles guides the preset number of last initial weight particles in the sorting result of the second group of initial weight particles to mutate, reducing the particle distance evaluation value from the first first initial weight particle to obtain a first particle update result, including: Based on the weight dimension, randomly obtain a value k, where 1 ≤ k ≤ the weight dimension; According to the value of k, using the first first initial weight particle as the target, mutate the preset number of last initial weight particles to reduce the particle distance evaluation value from the first first initial weight particle to obtain the first particle update result.

Images