Intelligent monitoring method based on dynamic modeling of charging screen and storage battery pack

Abstract

The invention provides an intelligent monitoring method based on dynamic modeling of a charging screen and a storage battery pack, and the method comprises the following steps: S1, obtaining a data sample of a storage battery in a charging and discharging process, carrying out the power storage abnormality detection based on a statistical parameter, and building a defect recognition model through a reference model; s2, through data preprocessing and health feature extraction, dividing training data and test data, training a time sequence prediction model and carrying out SOH estimation so as to construct a storage battery health state analysis model; and S3, constructing a residual service life prediction model of the state characteristics of the storage battery. According to the invention, the operation state information of the equipment can be accurately obtained in real time, the potential fault can be predicted in advance, and the reliability and stability of the power system can be improved.

Description

Technical Field

[0001] The present invention relates to the field of energy storage technology, and in particular to an intelligent monitoring method based on dynamic modeling of a charging screen and a battery pack. Background Art

[0002] In the power system, charging screens and battery packs are important backup power equipment, and their performance and stability are crucial to ensure the reliable operation of the power system. However, traditional monitoring methods often have problems such as low monitoring accuracy, poor real-time performance, and inability to accurately reflect the dynamic characteristics of the equipment.

[0003] At present, common monitoring methods mainly rely on regular manual inspections and simple parameter measurements, which is not only inefficient but also prone to missed detections and misjudgments. In addition, some existing monitoring systems lack the ability to dynamically model charging screens and battery packs, making it difficult to accurately predict performance changes and potential failures of equipment.

[0004] As the reliability requirements of power systems continue to increase, a more intelligent, accurate and real-time monitoring method is urgently needed to achieve effective management and maintenance of charging panels and battery packs. Summary of the invention

[0005] In order to solve the current technical problems, the main purpose of the present invention is to provide an intelligent monitoring method based on dynamic modeling of charging screens and battery packs, which can obtain the operating status information of the equipment in real time and accurately, predict potential faults in advance, and improve the reliability and stability of the power system.

[0006] The technical solution adopted by the present invention is: an intelligent monitoring method based on dynamic modeling of a charging screen and a battery pack, comprising the following steps: S1. Obtain data samples of the charging screen and the battery during the charging and discharging process, perform storage anomaly detection based on statistical parameters, and establish a defect recognition model through a benchmark model; S2, constructing a battery health status analysis model by preprocessing data and extracting health features, dividing training data and test data, training a time series prediction model and performing SOH estimation; S3. Construct a remaining service life prediction model for battery status characteristics.

[0007] The operating data of the charging screen and the battery are collected in real time. The operating data includes voltage, current, internal resistance and temperature. The collected operating data is preprocessed to obtain data samples.

[0008] The data samples are reduced in dimension, and the baseline model is established using the battery operation data under normal operating conditions.

[0009] The method of judging data samples through the benchmark model is as follows: Calculate T2 index: Use the benchmark model to calculate the T2 index of each new sample to determine whether it deviates from the normal state; Calculate the SPE index: Use the benchmark model to calculate the SPE index for each new sample to determine whether there is a prediction difference; Check whether the T2 and SPE indicators of the new sample exceed the preset thresholds to determine whether the battery may be abnormal. In S1, establishing a defect recognition model includes the following steps: The K-means clustering algorithm is used to group the battery data and classify similar data samples into the same group based on the Euclidean distance, thus forming different clusters. Calculate the center of each cluster and label each cluster with the battery defect type, including thermal runaway, open circuit, overcharge, and over discharge; Based on the cluster centers, the defect types of new batteries are calculated. Collect historical battery data and create defect labels for each data; Build and train a support vector machine (SVM) to map battery data to different defect types; In the training phase, sample data of known defect types are used for model training and parameter optimization; In the identification stage, new battery data samples are input into the SVM model to predict the defect type to which they belong.

[0010] During the performance evaluation phase of the battery defect type identification algorithm, cross-validation and evaluation indicators were used for testing, and the accuracy and reliability of the algorithm were evaluated by comparing the actual defect types with the algorithm prediction results.

[0011] In S2, the methods of data preprocessing and health feature extraction are as follows: Obtain real-time data of the charging screen and battery charging and discharging process, pre-process the collected data, extract the battery health feature sequence, analyze the correlation between the health feature sequence and battery capacity through the Pearson correlation coefficient, and screen the health features. The calculation formula is as follows: ; In the formula, X represents the sample population of health characteristic sequence, Y represents the sample population of capacity, and E represents the expected value of the data.

[0012] In S2, the method of dividing training data and test data is as follows: Split the dataset into a training set and a test set, use the training set to train the model, and use the test set to evaluate the performance of the model on unseen data.

[0013] In S2, the method for training the time series forecasting model and performing SOH estimation is as follows: A battery SOH estimation model based on the time series prediction model is constructed. The model is trained on the training set and the model's hyperparameters are optimized. Then, the data of the test set is used as the model input, and the SOH estimation result is obtained through the current maximum available capacity and the nominal capacity of the battery. The calculation formula is as follows: ; In the formula, Represents the current maximum available capacity. Represents the nominal capacity of the battery.

[0014] In S3, the method for constructing the remaining service life prediction model of the battery state characteristics is as follows: The number of cycles remaining before the battery reaches the end of its battery life is: ; In the formula, Represents the remaining battery life. The number of cycles representing end of life; Represents: the number of cycles actually used; In the RUL prediction process, a battery RUL joint prediction model is formed by using the time series prediction model; In the application stage, the real SOH data of the battery is used as the RUL model input, and the SOH is extrapolated to the failure threshold to achieve a joint prediction of the battery RUL. The output value is ; First, in order to measure the quality of the proposed prediction model, the root mean square error, mean absolute error, model determination coefficient and RUL prediction error are used for evaluation, among which: ; ; ; ; ; ; In the formula, stands for root mean square error; represents the mean absolute error; represents the model determination coefficient; and is the actual value and predicted value of the battery capacity in the i-th cycle; is the average value of the true value; RUL and PRUL are the true value and predicted value of the remaining service life of the battery respectively; PEr represents the RUL prediction error; represents the prediction error value.

[0015] The real value of the battery SOH and the estimated value of the SOH are taken as input to train the model and perform a joint prediction of the battery RUL; then, by dividing the model input data set into a training set and a test set, the RUL prediction model is trained on the training set, and the model hyperparameters are tuned. After that, the battery RUL prediction is performed on the test set, and the final RUL prediction result is output.

[0016] The present invention has the following beneficial effects: 1. The present invention integrates multi-source data and algorithms to establish an accurate dynamic model to achieve comprehensive and in-depth monitoring of the charging screen and battery pack; 2. The present invention updates model parameters in real time to adapt to changes in equipment performance and the influence of environmental factors, thereby improving the accuracy and reliability of monitoring; 3. The present invention combines a variety of prediction and diagnosis methods to detect potential faults in advance, effectively reducing equipment failure rate and maintenance costs; 4. The present invention adopts model prediction and intelligent alarm technology to improve operation and maintenance efficiency and response speed, and ensure the safe and stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 It is a flow chart of the present invention.

[0019] Figure 2 This is a flow chart of step one of the present invention.

[0020] Figure 3 This is a flow chart of step 2 of the present invention.

[0021] Figure 4 This is a flow chart of step three of the present invention. DETAILED DESCRIPTION

[0022] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] like Figure 1As shown, the present invention provides an intelligent monitoring method based on dynamic modeling of a charging screen and a battery pack, comprising the following steps: S1. Obtain data samples of the charging screen and the battery during the charging and discharging process, perform storage anomaly detection based on statistical parameters, and establish a defect recognition model through a benchmark model; S2, constructing a battery health status analysis model by preprocessing data and extracting health features, dividing training data and test data, training a time series prediction model and performing SOH estimation; S3. Construct a remaining service life prediction model for battery status characteristics.

[0024] First, the real-time data of the charging screen and battery charging and discharging process, such as voltage, current, resistance and temperature, are obtained through the online monitoring system, and then the pre-processed battery data is subjected to multi-dimensional and multi-state analysis. That is, based on the abnormal operation detection of battery based on statistical parameters, a battery defect type recognition model based on clustering algorithm and support vector machine SVM is established, aiming to timely detect possible abnormal conditions of the battery, such as lagging cells, open circuit, thermal runaway, etc., and accurately identify and classify the defect type.

[0025] Then, based on the existing battery charge and discharge history data set, the optimal life prediction model of the battery state characteristics is constructed. That is, the multi-source and multi-dimensional data in the data set are deeply extracted, and then the principal component analysis is performed on the feature data to obtain the principal component information with large contribution value, and a time series prediction model is constructed based on this to predict the remaining life of the battery.

[0026] Finally, an optimized iterative algorithm is used to update the weights, optimize the parameters and iterate the model of the life prediction model to improve the generalization ability and prediction accuracy of the model.

[0027] See also Figure 2 , real-time collection of charging screen and battery operation data, including voltage, current, internal resistance and temperature, etc., pre-processing the collected operation data to obtain data samples. The pre-processing of the collected data includes data cleaning, missing value supplementation and normalization.

[0028] In this embodiment, principal component analysis (PCA) is used to reduce the dimension of the data, and the battery operation data in a normal operation state is used to establish a benchmark model.

[0029] Specifically, the method for judging data samples through the benchmark model is as follows: Calculate T2 index: Use the benchmark model to calculate the T2 index of each new sample to determine whether it deviates from the normal state; Calculate the SPE index: Use the benchmark model to calculate the SPE index for each new sample to determine whether there is a prediction difference; Check whether the T2 and SPE indicators of the new sample exceed the preset thresholds to determine whether the battery may be abnormal. Among them, the T2 index includes whether the data such as voltage, current, and internal resistance deviate from the normal state; the SPE index includes small disturbances of the data such as voltage, current, and internal resistance.

[0030] See also Figure 3 ,In S1, establishing the defect recognition model includes the following steps: The K-means clustering algorithm is used to group the battery data and classify similar data samples into the same group based on the Euclidean distance, thus forming different clusters. Secondly, the center of each cluster is calculated and the battery defect type is labeled for each cluster, including thermal runaway, open circuit, overcharge, and over discharge. Then, based on the cluster centers, the defect type of the new battery is calculated; Again, collect historical battery data and create defect labels for each data; Finally, a support vector machine (SVM) is built and trained to map the battery data to different defect types.

[0031] In the training phase, sample data of known defect types are used for model training and parameter optimization; In the identification stage, new battery data samples are input into the SVM model to predict the defect type to which they belong.

[0032] During the performance evaluation stage of the battery defect type identification algorithm, cross-validation and evaluation indicators were used for testing. The evaluation indicators included accuracy, recall rate, etc. The accuracy and reliability of the algorithm were evaluated by comparing the actual defect types with the algorithm prediction results.

[0033] The state of health (SOH) of a battery refers to the ratio between the current capacity that can be discharged by the battery and the nominal capacity, reflecting the battery's ability to retain capacity. When the current capacity of the battery is less than 80% of the original nominal capacity, that is, SOH < 80%, the battery should be replaced. The prediction model based on the time series can directly analyze and predict the battery SOH based on the online monitoring data of the battery's operating status, understand the battery's health level, and perform maintenance in a timely manner, thereby extending the battery's service life and ensuring the stability and reliability of the battery throughout its life cycle. Figure 4 ,The implementation process of building a substation battery health status analysis and prediction model is as follows: 1. The methods of data preprocessing and health feature extraction are as follows: First, obtain the real-time data of the charging screen and the battery during charging and discharging, pre-process the collected data, extract the battery health feature sequence, analyze the correlation between the health feature sequence and the battery capacity through the Pearson correlation coefficient, and screen the health features. The calculation formula is as follows: ; In the formula, X represents the sample population of health characteristic sequence, Y represents the sample population of capacity, and E represents the expected value of the data.

[0034] 2. The method of dividing training data and test data is as follows: Split the dataset into a training set and a test set, use the training set to train the model, and use the test set to evaluate the performance of the model on unseen data.

[0035] The division of data into training data and test data is an important step in machine learning. By dividing the data set into training set and test set, the training set can be used to train the model, and the test set can be used to evaluate the performance of the model on unseen data, which helps to understand whether the model can be well generalized to new data, rather than just overfitting the training data. In order to make the deployed battery health status analysis model have a good prediction effect in real data, it is necessary to understand the generalization error of the model, that is, the error of the model in the real environment. However, due to the characteristics of large amounts of data and training in machine learning, we can neither directly use the generalization error as a signal to understand the generalization ability of the model, nor use the degree of fit of the model to the training data set as a signal to understand the generalization ability of the model. Therefore, by dividing the training data and test data, verification can be performed while training, and hyperparameters can be adjusted according to the model training results, so as to better optimize the model hyperparameters.

[0036] 3. The method of training the time series prediction model and estimating SOH is as follows: A battery SOH estimation model based on the time series prediction model is constructed. The model is trained on the training set and the model's hyperparameters are optimized. Then, the data of the test set is used as the model input, and the SOH estimation result is obtained through the current maximum available capacity and the nominal capacity of the battery. The calculation formula is as follows: ; In the formula, Represents the current maximum available capacity. Represents the nominal capacity of the battery.

[0037] The remaining useful life (RUL) is usually defined as the number of cycles remaining before the battery reaches the end of life (EOL), and the corresponding formula is: ; In the formula, Represents the remaining battery life. The number of cycles representing end of life; Represents: The number of cycles actually used.

[0038] See also Figure 4 ,In the RUL prediction process, since the input features at different moments contribute to the RUL prediction at the next moment to different degrees, a ,battery RUL joint prediction model is formed by using the time series ,prediction model.

[0039] In the application stage, the real SOH data of the battery is used as the RUL model input, and the SOH is extrapolated to the failure threshold to achieve a joint prediction of the battery RUL. The output value is ; First, in order to measure the quality of the proposed prediction model, the root mean square error, mean absolute error, model determination coefficient and RUL prediction error are used for evaluation, among which: ; ; ; ; ; ; In the formula, stands for root mean square error; represents the mean absolute error; represents the model determination coefficient; and is the actual value and predicted value of the battery capacity in the i-th cycle; is the average value of the true value; RUL and PRUL are the true value and predicted value of the remaining service life of the battery respectively; represents the RUL prediction error; represents the prediction error value.

[0040] The real value of the battery SOH and the estimated value of the SOH are taken as input to train the model and perform a joint prediction of the battery RUL; then, by dividing the model input data set into a training set and a test set, the RUL prediction model is trained on the training set, and the model hyperparameters are tuned. After that, the battery RUL prediction is performed on the test set, and the final RUL prediction result is output.

Claims

1. An intelligent monitoring method based on dynamic modeling of charging screen and battery pack, characterized in that: The following steps are involved: S1. Obtain data samples of the charging screen and the battery during the charging and discharging process, perform storage anomaly detection based on statistical parameters, and establish a defect recognition model through a benchmark model; S2, constructing a battery health status analysis model by preprocessing data and extracting health features, dividing training data and test data, training a time series prediction model and performing SOH estimation; S3. Construct a remaining service life prediction model for battery status characteristics.

2. According to claim 1, an intelligent monitoring method based on dynamic modeling of charging screen and battery pack is characterized in that: In S1, the operating data of the charging screen and the battery are collected in real time, the operating data including voltage, current, internal resistance and temperature, and the collected operating data are preprocessed to obtain data samples.

3. According to claim 1, an intelligent monitoring method based on dynamic modeling of charging screen and battery pack is characterized in that: In S1, the dimension of the data samples is reduced, and a baseline model is established using the battery operation data in a normal operating state.

4. According to claim 1, an intelligent monitoring method based on dynamic modeling of charging screen and battery pack is characterized in that: In S1, the method of judging data samples by the benchmark model is as follows: Calculate T2 index: Use the benchmark model to calculate the T2 index of each new sample to determine whether it deviates from the normal state; Calculate the SPE index: Use the benchmark model to calculate the SPE index for each new sample to determine whether there is a prediction difference; Check whether the T2 and SPE indicators of the new sample exceed the preset thresholds to determine whether the battery may be abnormal.

5. According to claim 1, an intelligent monitoring method based on dynamic modeling of charging screen and battery pack is characterized in that: In S1, establishing a defect recognition model includes the following steps: The K-means clustering algorithm is used to group the battery data and classify similar data samples into the same group based on the Euclidean distance, thus forming different clusters. Calculate the center of each cluster and label each cluster with the battery defect type, including thermal runaway, open circuit, overcharge, and over discharge; Based on the cluster centers, the defect types of new batteries are calculated. Collect historical battery data and create defect labels for each data; Build and train a support vector machine (SVM) to map battery data to different defect types; In the training phase, sample data of known defect types are used for model training and parameter optimization; In the identification stage, new battery data samples are input into the SVM model to predict the defect type to which they belong.

6. The intelligent monitoring method based on dynamic modeling of charging screen and battery pack according to claim 5 is characterized in that: During the performance evaluation phase of the battery defect type identification algorithm, cross-validation and evaluation indicators were used for testing, and the accuracy and reliability of the algorithm were evaluated by comparing the actual defect types with the algorithm prediction results.

7. The intelligent monitoring method based on dynamic modeling of charging screen and battery pack according to claim 1 is characterized in that: In S2, the methods of data preprocessing and health feature extraction are as follows: Obtain real-time data of the charging screen and battery charging and discharging process, pre-process the collected data, extract the battery health feature sequence, analyze the correlation between the health feature sequence and battery capacity through the Pearson correlation coefficient, and screen the health features. The calculation formula is as follows: ； In the formula, X represents the sample population of health characteristic sequence, Y represents the sample population of capacity, and E represents the expected value of the data.

8. The intelligent monitoring method based on dynamic modeling of charging screen and battery pack according to claim 7 is characterized in that: In S2, the method of dividing training data and test data is as follows: Split the dataset into a training set and a test set, use the training set to train the model, and use the test set to evaluate the performance of the model on unseen data.

9. The intelligent monitoring method based on dynamic modeling of charging screen and battery pack according to claim 8 is characterized in that: In S2, the method for training the time series forecasting model and performing SOH estimation is as follows: A battery SOH estimation model based on the time series prediction model is constructed. The model is trained on the training set and the model's hyperparameters are optimized. Then, the data of the test set is used as the model input, and the SOH estimation result is obtained through the current maximum available capacity and the nominal capacity of the battery. The calculation formula is as follows: ； In the formula, Represents the current maximum available capacity. Represents the nominal capacity of the battery.

10. The intelligent monitoring method based on dynamic modeling of charging screen and battery pack according to claim 1 is characterized in that: In S3, the method for constructing the remaining service life prediction model of the battery state characteristics is as follows: The number of cycles remaining before the battery reaches the end of its battery life is: ； In the formula, Represents the remaining battery life. The number of cycles representing end of life; Represents the number of cycles actually used; In the RUL prediction process, a battery RUL joint prediction model is formed by using the time series prediction model; In the application stage, the real SOH data of the battery is used as the RUL model input, and the SOH is extrapolated to the failure threshold to achieve a joint prediction of the battery RUL. The output value is ; First, in order to measure the quality of the proposed prediction model, the root mean square error, mean absolute error, model determination coefficient and RUL prediction error are used for evaluation, among which: ； ； ； ； ； ； In the formula, stands for root mean square error; represents the mean absolute error; represents the model determination coefficient; and is the actual value and predicted value of the battery capacity in the i-th cycle; is the average value of the true value; RUL and PRUL are the true value and predicted value of the remaining service life of the battery respectively; PEr represents the RUL prediction error; represents the prediction error value; The real value of the battery SOH and the estimated value of the SOH are taken as input to train the model and perform a joint prediction of the battery RUL; then, by dividing the model input data set into a training set and a test set, the RUL prediction model is trained on the training set, and the model hyperparameters are tuned. After that, the battery RUL prediction is performed on the test set, and the final RUL prediction result is output.

Images