A comprehensive safety protection method and system for base station energy storage batteries

Abstract

The present invention discloses an all-round safety protection method and system for base station energy storage batteries, including: obtaining multi-source monitoring data of base station energy storage batteries, performing online identification based on actual working conditions, and screening battery status parameters; using a residual generator to obtain residual data with respect to normal working conditions of energy storage batteries to analyze abnormal conditions, and classifying abnormal conditions into abnormal categories, and setting different types of abnormal monitoring branches for base station energy storage batteries; constructing an abnormal joint detection model, coupling different monitoring matrices, and analyzing the functional relationship between different types of abnormal conditions, and performing joint estimation of the battery status of base station energy storage batteries; providing early warning based on the estimated battery status combined with the actual working conditions of the base station energy storage batteries, and optimizing battery management. The present invention makes battery status evaluation more comprehensive by analyzing the relationship between different abnormal battery states, which helps to prevent battery pack safety issues in advance, and has good estimation accuracy and robustness.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage batteries, and more specifically, to an all-round safety protection method and system for base station energy storage batteries. Background Art

[0002] As the penetration rate of 5G base stations increases rapidly, the energy storage battery capacity of the base station group cannot be underestimated. At the same time, the internal energy storage lithium batteries of 5G base stations have the characteristics of fast response speed and long access to the system, and have great potential for flexible scheduling. In addition, considering the immobility and long idle time of base station backup energy storage, using 5G communication base stations as energy storage mechanisms can improve energy utilization efficiency, reduce carbon emissions, reduce the peak and valley electricity consumption differences of the macro power system, improve electricity quality, and thus reduce base station energy consumption and operating costs.

[0003] The energy storage monitoring system is an important component of the lithium battery energy storage system. It is the control center of the energy storage system and plays a role in uploading and controlling optimization. It is used to monitor the alarm and fault status of the battery system, the current chargeable and dischargeable power, the statistical energy changes during the battery charging and discharging process, and the health status of the battery system. At present, the lithium battery energy storage system is composed of multiple single cells connected in series and parallel to form a large-capacity battery system. The battery monitoring module can effectively monitor and control the lithium battery system, energy storage converter, switch status, etc., to ensure the safety of the energy storage battery in the energy storage system, and make the entire energy storage system operate safely, reliably and economically. However, the existing monitoring lacks in-depth analysis of the relationship between different factors, making it difficult to prevent battery safety problems and performance degradation in advance. Therefore, how to achieve all-round safety protection based on the multi-source monitoring data of the base station energy storage battery and improve the response capability of the energy storage battery abnormality is an urgent problem to be solved. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes an all-round safety protection method and system for base station energy storage batteries.

[0005] A first aspect of the present invention provides an all-round safety protection method for a base station energy storage battery, comprising:

[0006] Obtain and preprocess the multi-source monitoring data of the base station energy storage battery, perform online identification on the pre-processed multi-source monitoring data based on actual working conditions, and select monitoring parameters as battery status parameters of the base station energy storage battery;

[0007] Aggregate the battery status parameters and use a residual generator to obtain residual data from the normal working condition of the energy storage battery to analyze the abnormal state, and classify the abnormal state into abnormal categories, and set different types of abnormal monitoring branches for base station energy storage batteries;

[0008] Acquire monitoring matrices of different types of abnormal monitoring branches, construct an abnormal joint detection model, couple the monitoring matrices, analyze the interaction relationship between different types of abnormal states, and perform joint estimation of the battery state of the base station energy storage battery;

[0009] According to the battery estimation status output by the abnormal joint detection model and the actual working condition of the base station energy storage battery, an early warning is issued, and the battery management of the base station energy storage battery is optimized based on the early warning information.

[0010] In this solution, the pre-processed multi-source monitoring data is identified online based on the actual working conditions, and the monitoring parameters are selected as the battery status parameters of the base station energy storage battery, specifically:

[0011] Generate multi-source monitoring data according to the environmental data and charging and discharging data of the base station energy storage battery, perform outlier processing, normalization processing and windowing processing on the multi-source monitoring data, and obtain pre-processed multi-source monitoring data;

[0012] Preliminarily clustering the multi-source monitoring data to achieve charging and discharging condition division, obtaining a multi-source monitoring data sequence with condition labels, retrieving battery state monitoring and SOC estimation instances, and extracting characteristic parameters involved in each instance;

[0013] Constructing and generating a characteristic parameter sequence according to the characteristic parameters, clustering the multi-source monitoring data sequence and the characteristic parameter sequence again, evaluating the similarity between the multi-source monitoring data sequence under different working conditions and each characteristic parameter sequence by using a dynamic time warping method, and obtaining the cluster to which the multi-source monitoring data sequence belongs by clustering;

[0014] Extract all the sequences in the cluster where the multi-source monitoring data sequence belongs, and use multivariate empirical mode decomposition to decompose all the sequences, obtain several IMF component groups, and divide the IMF component groups into high-frequency IMF component groups and low-frequency IMF component groups;

[0015] In the high-frequency IMF component group, the variable selection coefficient is calculated by using LASSO variable selection, the optimal characteristic parameter combination is selected in combination with the preset lag value, and the monitoring parameter is selected as the battery state parameter of the base station energy storage battery using the optimal characteristic parameter combination.

[0016] In this solution, the battery state parameters are aggregated and a residual generator is used to obtain residual data from the normal working condition of the energy storage battery to analyze the abnormal state, specifically:

[0017] Based on the battery status parameters, the battery status parameter sequence of the base station energy storage battery under normal working conditions is screened, the screened battery status parameter sequence is aggregated, and a normal state model of the base station energy storage battery is constructed using a deep belief network;

[0018] Divide the normal working condition intervals of the base station energy storage battery according to the parameter characteristics of the battery state parameter sequence, calculate the occurrence frequencies of different normal working condition intervals, and set the sampling quantity according to the occurrence frequencies to obtain training data samples, and train the normal state model;

[0019] Read the abnormal working condition data of the base station energy storage battery, and obtain the battery state parameter sequence corresponding to the abnormal working condition data, and import the battery state parameter sequence with the abnormal working condition label into the normal state model to obtain the parameter estimation value;

[0020] Calculate the residual data of the parameter estimation value and the battery state parameter sequence with abnormal working condition labels, obtain the relative entropy of the residual distribution, use the relative entropy to set the abnormality judgment thresholds of different types of abnormalities, and set different types of abnormality monitoring branches of the base station energy storage battery based on the abnormality judgment thresholds.

[0021] In this solution, an abnormal joint detection model is constructed, specifically:

[0022] Obtain the abnormality category of the base station energy storage battery, analyze the abnormality mechanisms corresponding to different abnormality categories, calculate the correlation between the abnormality mechanisms, and read the degree of association between the abnormality categories based on the frequency and time of occurrence of different abnormality categories;

[0023] Different anomaly categories are used as anomaly entities for graph mapping. The edge structure between different anomaly entities is obtained by using the correlation and association degree of different anomaly categories, and an undirected graph consisting of anomaly entities and edge structures is constructed.

[0024] Reading the adjacency matrix according to the undirected graph, obtaining the monitoring matrix and abnormal monitoring results output by different types of abnormal monitoring branches of the base station energy storage battery, and assigning values ​​to the adjacency matrix using each monitoring matrix;

[0025] A graph convolutional neural network is used to learn the undirected graph to construct an abnormality joint detection model, and the model is trained. The assigned adjacency matrix is ​​used as the model input to perform joint estimation of the battery state of the base station energy storage battery.

[0026] In this solution, the battery status of the base station energy storage battery is jointly estimated, specifically:

[0027] The assigned adjacency matrix is ​​imported into the graph convolutional neural network for encoding to obtain the embedded representation of each abnormal node. A multi-head attention mechanism is introduced to obtain the implicit relationship between the embedded representations of different abnormal nodes, and the implicit association weight is set according to the implicit relationship.

[0028] According to the implicit association weights, weighted aggregation is performed using the neighbor aggregation mechanism of the graph convolutional neural network to update the embedded representation of each abnormal node, and the abnormal nodes that are abnormal at the current moment are marked by marking the abnormal monitoring results output by different types of abnormal monitoring branches;

[0029] The inner product is calculated based on the updated embedding representation of the marked abnormal node and the updated embedding representations of other abnormal nodes, and the abnormal state is determined using the preset inner product threshold to obtain the abnormal score. When the inner product is greater than the preset inner product threshold, it means that the base station energy storage battery will have an abnormal state of the corresponding type at the preset prediction time.

[0030] In this solution, an early warning is issued based on the estimated battery status output by the abnormal joint detection model and the actual working condition of the base station energy storage battery, specifically:

[0031] According to the battery state joint estimation result output by the anomaly joint detection model, the abnormal state category of the base station energy storage battery in the future preset time step is obtained, low-level warning information is generated, and the abnormal score of the abnormal state category is obtained to evaluate the degree of abnormality in combination with the historical abnormal instances of the energy storage battery;

[0032] Performing principal component analysis on the battery status parameters according to the abnormal status category, obtaining the principal component parameters, obtaining the actual operating condition distribution of the base station energy storage battery in a daily cycle, and obtaining the principal component parameter variation interval according to the actual operating condition distribution;

[0033] The influence coefficient is generated by the abnormality degree to weight the principal component parameters under normal working conditions, and the variation interval of the principal component parameters under the influence of the abnormality is obtained. The two principal component parameter variation intervals are compared, and when the deviation is greater than the preset deviation threshold, a high warning information is generated;

[0034] The warning information within the preset time range is aggregated, and the relevant knowledge graph is connected to optimize the flexible charging and discharging strategy according to the principal component parameters corresponding to the abnormal state category. If the warning information does not decrease in the future period, the warning information will be fed back to the expert end to obtain feedback information, and the warning threshold will be optimized based on the feedback information.

[0035] The second aspect of the present invention further provides an all-round safety protection system for a base station energy storage battery, the system comprising: a memory, a processor, the memory comprising an all-round safety protection method program for a base station energy storage battery, the all-round safety protection method program for a base station energy storage battery being executed by the processor to implement the following steps:

[0036] Obtain and preprocess the multi-source monitoring data of the base station energy storage battery, perform online identification on the pre-processed multi-source monitoring data based on actual working conditions, and select monitoring parameters as battery status parameters of the base station energy storage battery;

[0037] Aggregate the battery status parameters and use a residual generator to obtain residual data from the normal working condition of the energy storage battery to analyze the abnormal state, and classify the abnormal state into abnormal categories, and set different types of abnormal monitoring branches for base station energy storage batteries;

[0038] Acquire monitoring matrices of different types of abnormal monitoring branches, construct an abnormal joint detection model, couple the monitoring matrices, analyze the interaction relationship between different types of abnormal states, and perform joint estimation of the battery state of the base station energy storage battery;

[0039] According to the battery estimation status output by the abnormal joint detection model and the actual working condition of the base station energy storage battery, an early warning is issued, and the battery management of the base station energy storage battery is optimized based on the early warning information.

[0040] The present invention discloses an all-round safety protection method and system for base station energy storage batteries, including: obtaining multi-source monitoring data of base station energy storage batteries, performing online identification based on actual working conditions, and screening battery status parameters; using a residual generator to obtain residual data with respect to normal working conditions of energy storage batteries to analyze abnormal conditions, and classifying abnormal conditions into abnormal categories, and setting different types of abnormal monitoring branches for base station energy storage batteries; constructing an abnormal joint detection model, coupling different monitoring matrices, and analyzing the functional relationship between different types of abnormal conditions, and performing joint estimation of the battery status of base station energy storage batteries; providing early warning based on the estimated battery status combined with the actual working conditions of the base station energy storage batteries, and optimizing battery management. The present invention makes battery status evaluation more comprehensive by analyzing the relationship between different abnormal battery states, which helps to prevent battery pack safety issues in advance, and has good estimation accuracy and robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flow chart of a comprehensive safety protection method for a base station energy storage battery according to the present invention is shown;

[0042] Figure 2 A flow chart of a method for setting different types of abnormal monitoring branches according to the present invention is shown;

[0043] Figure 3 A flow chart of a method for jointly estimating the battery status of a base station energy storage battery according to the present invention is shown;

[0044] Figure 4 A block diagram of an all-round safety protection system for base station energy storage batteries according to the present invention is shown. DETAILED DESCRIPTION

[0045] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0046] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0047] Figure 1 A flow chart of an all-round safety protection method for a base station energy storage battery of the present invention is shown.

[0048] like Figure 1 As shown, the first aspect of the present invention provides an all-round safety protection method for a base station energy storage battery, comprising:

[0049] S102, obtaining and preprocessing multi-source monitoring data of the base station energy storage battery, performing online identification on the pre-processed multi-source monitoring data based on actual working conditions, and selecting monitoring parameters as battery status parameters of the base station energy storage battery;

[0050] S104, aggregating the battery status parameters and using a residual generator to obtain residual data from a normal working condition of the energy storage battery to analyze the abnormal state, and classifying the abnormal state into abnormal categories, and setting different types of abnormal monitoring branches for the base station energy storage battery;

[0051] S106, obtaining monitoring matrices of different types of abnormal monitoring branches, constructing an abnormal joint detection model, coupling the monitoring matrices, and analyzing the functional relationship between different types of abnormal states, and performing a joint estimation of the battery state of the base station energy storage battery;

[0052] S108, issuing an early warning based on the estimated battery status output by the abnormal joint detection model and the actual working condition of the base station energy storage battery, and optimizing the battery management of the base station energy storage battery based on the early warning information.

[0053] It should be noted that digital control, power electronics and wireless communication technologies are used to realize all-round safety data monitoring, charge and discharge management, energy control, as well as cloud-based telesignaling, telemetering, remote control and remote adjustment of base station energy storage batteries. Multi-source monitoring data is generated based on the environmental data and charge and discharge data of the base station energy storage battery, including temperature and humidity, charge and discharge rate, internal resistance, battery capacity, charge and discharge current voltage and energy density, etc. The multi-source monitoring data is processed for outliers, normalized and windowed to obtain pre-processed multi-source monitoring data; the multi-source monitoring data is preliminarily clustered using the K-means algorithm to realize the division of charge and discharge working conditions, and a multi-source monitoring data sequence with working condition labels is obtained; battery status monitoring and SOC estimation examples are retrieved using data retrieval or literature retrieval engines, and the characteristic parameters involved in each example are extracted; the characteristic parameters are constructed based on the characteristic parameters. A characteristic parameter sequence is constructed, the multi-source monitoring data sequence and the characteristic parameter sequence are clustered again, the clustering process is optimized using the dynamic time warping method, the similarity between the multi-source monitoring data sequence and each characteristic parameter sequence under different working conditions is evaluated, and the cluster to which the multi-source monitoring data sequence belongs is obtained through clustering; all sequences in the cluster to which the multi-source monitoring data sequence belongs are extracted, and all sequences are decomposed by multivariate empirical mode decomposition to obtain several IMF component groups, and each component in the intrinsic mode eigenfunction (IMF) grouping obtained by the multivariate empirical mode decomposition has the same length, the same frequency distribution and the same complexity ordering. The energy entropy calculation method is used to divide the IMF component group into a high-frequency IMF component group and a low-frequency IMF component group. The more useful information the corresponding component contains in the high-frequency IMF component group; the least absolute shrinkage and selection operator (LASSO) is used to select variables in the high-frequency IMF component group, the lag value is preset, the variable selection coefficient is calculated, the coefficient that is not zero is selected, and the characteristic parameters under the lag value are selected to construct the best characteristic parameter combination; the monitoring parameters are selected as the battery state parameters of the base station energy storage battery using the best characteristic parameter combination; by determining the characteristic parameters closely related to the multi-source monitoring data, the stability and accuracy of the subsequent joint detection of the base station energy storage battery are ensured.

[0054] Figure 2 A flow chart of a method for setting different types of abnormal monitoring branches according to the present invention is shown.

[0055] According to an embodiment of the present invention, the battery state parameters are aggregated and a residual generator is used to obtain residual data from a normal working condition of the energy storage battery to analyze the abnormal state, specifically:

[0056] S202, screening a battery state parameter sequence of a base station energy storage battery under normal working conditions based on the battery state parameter, aggregating the screened battery state parameter sequence, and constructing a normal state model of the base station energy storage battery using a deep belief network;

[0057] S204, dividing the normal operating condition intervals of the base station energy storage battery according to the parameter characteristics of the battery state parameter sequence, calculating the occurrence frequencies of different normal operating condition intervals, and setting the sampling quantity according to the occurrence frequencies to obtain training data samples, and training the normal state model;

[0058] S206, reading abnormal operating condition data of the base station energy storage battery, and obtaining a battery state parameter sequence corresponding to the abnormal operating condition data, and importing the battery state parameter sequence with the abnormal operating condition label into the normal state model to obtain parameter estimation values;

[0059] S208, calculating the residual data of the parameter estimation value and the battery state parameter sequence with the abnormal operating condition label, obtaining the relative entropy of the residual distribution, using the relative entropy to set the abnormality determination thresholds of different types of abnormalities, and setting different types of abnormality monitoring branches for the base station energy storage battery based on the abnormality determination thresholds.

[0060] It should be noted that the normal state model of the base station energy storage battery is constructed using a deep belief network, in which the basic component unit of the deep belief network is a restricted Boltzmann machine. The main structure includes two layers of neurons. There is a connection relationship between any two neurons in different layers. The input module of the normal state model is obtained by connecting multiple restricted Boltzmann machines in series, and the fully connected layer is set as the output module of the model. The unsupervised method is used for training to obtain the parameter distribution characteristics of the battery state parameter sequence and optimize the data sampling, which effectively improves the training efficiency. When using the normal state model for parameter estimation, a sliding window is introduced to reduce random noise, and the Mahalanobis distance is used as a metric function to obtain the residual data of the parameter estimation value and the battery state parameter sequence with abnormal working condition labels, so as to better utilize the original data with abnormal working condition labels. The relative entropy of the residual distribution of the parameter estimation value and the battery state parameter sequence with abnormal working condition labels is calculated in different types of abnormal monitoring branches to indicate the degree of deviation, and a reasonable abnormal judgment threshold is obtained as the threshold of different types of abnormal monitoring branches to determine whether an abnormal state occurs.

[0061] Figure 3 A flow chart of a method for jointly estimating the battery status of a base station energy storage battery according to the present invention is shown.

[0062] According to an embodiment of the present invention, the battery status of the base station energy storage battery is jointly estimated, specifically:

[0063] S302, importing the assigned adjacency matrix into the graph convolutional neural network for encoding, obtaining the embedded representation of each abnormal node, introducing a multi-head attention mechanism to obtain the implicit relationship between the embedded representations of different abnormal nodes, and setting the implicit association weight according to the implicit relationship;

[0064] S304, performing weighted aggregation using the neighbor aggregation mechanism of the graph convolutional neural network according to the implicit association weight, updating the embedded representation of each abnormal node, and marking the abnormal node that has an abnormality at the current moment through the abnormal monitoring results output by different types of abnormal monitoring branches;

[0065] S306, performing inner product calculation based on the updated embedded representation of the marked abnormal node and the updated embedded representations of other abnormal nodes, and using a preset inner product threshold to determine the abnormal state to obtain an abnormal score. When the inner product is greater than the preset inner product threshold, it indicates that the base station energy storage battery will have an abnormal state of the corresponding type at the preset prediction time.

[0066] It should be noted that the abnormal category of the base station energy storage battery is obtained, the abnormal mechanisms corresponding to different abnormal categories are analyzed, the correlation between the abnormal mechanisms is calculated, and the degree of association of the abnormal categories is read according to the frequency and time of occurrence of different abnormal categories. When the occurrence frequency of two abnormalities is consistent and the occurrence time is in a similar time range, it means that the two abnormal types are highly dependent on each other; different abnormal categories are used as abnormal entities for graph mapping, and the edge structure between different abnormal entities is obtained by using the correlation and association degree of different abnormal categories, and an undirected graph composed of abnormal entities and edge structures is constructed; the adjacency matrix is ​​read according to the undirected graph, and the monitoring matrix and abnormal monitoring results output by different types of abnormal monitoring branches of the base station energy storage battery are obtained, and the adjacency matrix is ​​assigned using each monitoring matrix; the graph convolutional neural network is used to learn the undirected graph to construct an abnormal joint detection model, and the model is trained. The assigned adjacency matrix is ​​used as the model input to perform joint estimation of the battery state of the base station energy storage battery.

[0067] Since there are multiple dependent relationships between different abnormal states of energy storage batteries, abnormal charging and discharging temperatures, short circuits, overcurrents and other abnormalities can easily lead to thermal failure of energy storage batteries. The Pearson correlation coefficient is used to calculate the correlation between abnormal mechanisms and the degree of association to obtain the explicit relationship of abnormal nodes. However, there are not only explicit relationships between abnormal nodes, but also implicit implicit relationships. The multi-head self-attention mechanism is used to model the complex implicit association relationship between abnormal nodes from the spatial dimension, and the implicit association weights are generated through linear layer processing. The explicit and implicit relationships between abnormal nodes are fused through the neighbor aggregation mechanism to obtain the final embedded representation. According to the inner product between the final embedded representations of abnormal nodes, the associated anomalies that are easily caused under the existing abnormal state are screened, and the deep relationship between different anomalies is analyzed to achieve early warning of abnormalities, which improves the early warning response capability of base station energy storage batteries and realizes all-round safety warnings for different abnormal states.

[0068] It should be noted that, according to the battery state joint estimation result output by the anomaly joint detection model, the abnormal state category of the base station energy storage battery in the future preset time step is obtained, and low-level warning information is generated. The abnormal score is characterized according to the degree of deviation between the inner product calculation result and the inner product threshold in the battery state joint estimation, and the abnormal score of the abnormal state category is obtained. Combined with the historical abnormal examples of the energy storage battery, the degree of abnormality is evaluated; according to the abnormal state category, principal component analysis is performed in the battery state parameters to obtain the principal component parameters, and the actual working condition distribution of the base station energy storage battery in the daily cycle is obtained. According to the actual working condition distribution, the variation range of the principal component parameters is obtained; the influence coefficient on the normal working condition is generated by the abnormal degree. The principal component parameters under the abnormal conditions are weighted to obtain the principal component parameter variation interval under the influence of the abnormality, and the two principal component parameter variation intervals are compared. When the deviation is greater than the preset deviation threshold, a high warning information is generated; the warning information within the preset time range is aggregated, and the relevant lithium battery protection strategy knowledge graph is connected to optimize the flexible charging and discharging strategy according to the principal component parameters corresponding to the abnormal state category, and the battery management is optimized to effectively avoid a series of battery application safety issues such as external short circuit, overcurrent, impact current, and battery circulation; if the warning information does not decrease in the future period, the warning information will be fed back to the expert end to obtain feedback information, and the warning threshold will be optimized according to the feedback information.

[0069] Construct a battery status database for the base station energy storage battery, store the multi-source monitoring data and the battery status joint estimation result in the battery status database, obtain the energy storage battery charging data of the target base station to analyze the charging preference time period, evaluate according to the charging deviation time period, actual working conditions and the time interval corresponding to the valley peak electricity price, store the data in the battery status database, analyze the economy of the energy storage battery charging, and when the economy does not meet the preset standard, obtain base stations with similar actual working conditions and similar battery status in the battery status database, extract the charging strategy corresponding to the base station with the economy meeting the preset standard, and optimize the charging time period of the energy storage battery of the target base station.

[0070] Figure 4 A block diagram of an all-round safety protection system for base station energy storage batteries according to the present invention is shown.

[0071] The second aspect of the present invention further provides a comprehensive safety protection system 4 for a base station energy storage battery, the system comprising: a memory 41, a processor 42, the memory comprising a comprehensive safety protection method program for a base station energy storage battery, the comprehensive safety protection method program for a base station energy storage battery being executed by the processor to implement the following steps:

[0072] Obtain and preprocess the multi-source monitoring data of the base station energy storage battery, perform online identification on the pre-processed multi-source monitoring data based on actual working conditions, and select monitoring parameters as battery status parameters of the base station energy storage battery;

[0073] Aggregate the battery status parameters and use a residual generator to obtain residual data from the normal working condition of the energy storage battery to analyze the abnormal state, and classify the abnormal state into abnormal categories, and set different types of abnormal monitoring branches for base station energy storage batteries;

[0074] Acquire monitoring matrices of different types of abnormal monitoring branches, construct an abnormal joint detection model, couple the monitoring matrices, analyze the interaction relationship between different types of abnormal states, and perform joint estimation of the battery state of the base station energy storage battery;

[0075] According to the battery estimation status output by the abnormal joint detection model and the actual working condition of the base station energy storage battery, an early warning is issued, and the battery management of the base station energy storage battery is optimized based on the early warning information.

[0076] The third aspect of the present invention also provides a computer-readable storage medium, which includes a comprehensive safety protection method program for base station energy storage batteries. When the comprehensive safety protection method program for base station energy storage batteries is executed by a processor, the steps of the comprehensive safety protection method for base station energy storage batteries as described in any one of the above items are implemented.

[0077] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0078] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0079] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0080] A person of ordinary skill in the art can understand that: all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: a mobile storage device, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0081] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0082] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A comprehensive safety protection method for base station energy storage batteries, characterized in that: The following steps are involved: Obtain multi-source monitoring data of the base station energy storage battery and perform preprocessing, perform online identification on the preprocessed multi-source monitoring data based on actual working conditions, and select monitoring parameters as battery status parameters of the base station energy storage battery; Aggregate the battery status parameters and use a residual generator to obtain residual data from the normal working condition of the energy storage battery to analyze the abnormal state, and classify the abnormal state into abnormal categories, and set different types of abnormal monitoring branches for base station energy storage batteries; Acquire monitoring matrices of different types of abnormal monitoring branches, construct an abnormal joint detection model, couple the monitoring matrices, analyze the functional relationships between different types of abnormal states, and perform joint estimation of the battery state of the base station energy storage battery; Early warning is issued based on the estimated battery status output by the abnormal joint detection model and the actual working condition of the base station energy storage battery, and the battery management of the base station energy storage battery is optimized based on the early warning information; Based on the actual working conditions, the pre-processed multi-source monitoring data is identified online, and the monitoring parameters are selected as the battery status parameters of the base station energy storage battery, specifically: Generate multi-source monitoring data according to the environmental data and charging and discharging data of the base station energy storage battery, perform outlier processing, normalization processing and windowing processing on the multi-source monitoring data, and obtain pre-processed multi-source monitoring data; Preliminarily clustering the multi-source monitoring data to achieve charging and discharging condition division, obtaining a multi-source monitoring data sequence with condition labels, retrieving battery state monitoring and SOC estimation instances, and extracting characteristic parameters involved in each instance; Constructing and generating a characteristic parameter sequence according to the characteristic parameters, clustering the multi-source monitoring data sequence and the characteristic parameter sequence again, evaluating the similarity between the multi-source monitoring data sequence under different working conditions and each characteristic parameter sequence by using a dynamic time warping method, and obtaining the cluster to which the multi-source monitoring data sequence belongs by clustering; Extract all the sequences in the cluster where the multi-source monitoring data sequence belongs, and use multivariate empirical mode decomposition to decompose all the sequences, obtain several IMF component groups, and divide the IMF component groups into high-frequency IMF component groups and low-frequency IMF component groups; In the high-frequency IMF component group, the variable selection coefficient is calculated by using LASSO variable selection, the optimal characteristic parameter combination is selected in combination with the preset lag value, and the monitoring parameters are selected as the battery state parameters of the base station energy storage battery using the optimal characteristic parameter combination.

2. The all-round safety protection method for base station energy storage batteries according to claim 1, characterized in that: The battery state parameters are aggregated and the residual generator is used to obtain the residual data of the normal working condition of the energy storage battery to analyze the abnormal state, specifically: Based on the battery status parameters, the battery status parameter sequence of the base station energy storage battery under normal working conditions is screened, the screened battery status parameter sequence is aggregated, and a normal state model of the base station energy storage battery is constructed using a deep belief network; Divide the normal working condition intervals of the base station energy storage battery according to the parameter characteristics of the battery state parameter sequence, calculate the occurrence frequencies of different normal working condition intervals, and set the sampling quantity according to the occurrence frequencies to obtain training data samples, and train the normal state model; Read the abnormal working condition data of the base station energy storage battery, and obtain the battery state parameter sequence corresponding to the abnormal working condition data, and import the battery state parameter sequence with the abnormal working condition label into the normal state model to obtain the parameter estimation value; Calculate the residual data of the parameter estimation value and the battery state parameter sequence with abnormal working condition labels, obtain the relative entropy of the residual distribution, use the relative entropy to set the abnormality judgment thresholds of different types of abnormalities, and set different types of abnormality monitoring branches of the base station energy storage battery based on the abnormality judgment thresholds.

3. The all-round safety protection method for base station energy storage batteries according to claim 1, characterized in that: Construct an anomaly joint detection model, specifically: Obtain the abnormality category of the base station energy storage battery, analyze the abnormality mechanisms corresponding to different abnormality categories, calculate the correlation between the abnormality mechanisms, and read the degree of association between the abnormality categories based on the frequency and time of occurrence of different abnormality categories; Different anomaly categories are used as anomaly entities for graph mapping. The edge structure between different anomaly entities is obtained by using the correlation and association degree of different anomaly categories, and an undirected graph consisting of anomaly entities and edge structures is constructed. Reading the adjacency matrix according to the undirected graph, obtaining the monitoring matrix and abnormal monitoring results output by different types of abnormal monitoring branches of the base station energy storage battery, and assigning values ​​to the adjacency matrix using each monitoring matrix; A graph convolutional neural network is used to learn the undirected graph to construct an abnormality joint detection model, and the model is trained. The assigned adjacency matrix is ​​used as the model input to perform joint estimation of the battery state of the base station energy storage battery.

4. The all-round safety protection method for base station energy storage batteries according to claim 3, characterized in that: Perform joint estimation of the battery status of the base station energy storage battery, specifically: The assigned adjacency matrix is ​​imported into the graph convolutional neural network for encoding to obtain the embedded representation of each abnormal node. A multi-head attention mechanism is introduced to obtain the implicit relationship between the embedded representations of different abnormal nodes, and the implicit association weight is set according to the implicit relationship. According to the implicit association weights, weighted aggregation is performed using the neighbor aggregation mechanism of the graph convolutional neural network to update the embedded representation of each abnormal node, and the abnormal nodes that are abnormal at the current moment are marked by marking the abnormal monitoring results output by different types of abnormal monitoring branches; The inner product is calculated based on the updated embedding representation of the marked abnormal node and the updated embedding representations of other abnormal nodes, and the abnormal state is determined using the preset inner product threshold to obtain the abnormal score. When the inner product is greater than the preset inner product threshold, it means that the base station energy storage battery will have an abnormal state of the corresponding type at the preset prediction time.

5. The all-round safety protection method for base station energy storage batteries according to claim 1, characterized in that: According to the battery estimation status output by the abnormal joint detection model and the actual working condition of the base station energy storage battery, early warning is carried out, specifically: According to the battery state joint estimation result output by the anomaly joint detection model, the abnormal state category of the base station energy storage battery in the future preset time step is obtained, low-level warning information is generated, and the abnormal score of the abnormal state category is obtained to evaluate the degree of abnormality in combination with the historical abnormal instances of the energy storage battery; Performing principal component analysis on the battery status parameters according to the abnormal status category, obtaining the principal component parameters, obtaining the actual operating condition distribution of the base station energy storage battery in a daily cycle, and obtaining the principal component parameter variation interval according to the actual operating condition distribution; The influence coefficient is generated by the abnormality degree to weight the principal component parameters under normal working conditions, and the variation interval of the principal component parameters under the influence of the abnormality is obtained. The two variation intervals of the principal component parameters are compared, and when the deviation is greater than the preset deviation threshold, a high warning information is generated; The warning information within the preset time range is aggregated, and the relevant knowledge graph is connected to optimize the flexible charging and discharging strategy according to the principal component parameters corresponding to the abnormal state category. If the warning information does not decrease in the future period, the warning information will be fed back to the expert end to obtain feedback information, and the warning threshold will be optimized based on the feedback information.

6. An all-round safety protection system for base station energy storage batteries, characterized in that: The system includes: a memory and a processor, wherein the memory includes a comprehensive safety protection method program for a base station energy storage battery, and when the comprehensive safety protection method program for a base station energy storage battery is executed by the processor, the following steps are implemented: Obtain multi-source monitoring data of the base station energy storage battery and perform preprocessing, perform online identification on the preprocessed multi-source monitoring data based on actual working conditions, and select monitoring parameters as battery status parameters of the base station energy storage battery; Aggregate the battery status parameters and use a residual generator to obtain residual data from the normal working condition of the energy storage battery to analyze the abnormal state, and classify the abnormal state into abnormal categories, and set different types of abnormal monitoring branches for base station energy storage batteries; Acquire monitoring matrices of different types of abnormal monitoring branches, construct an abnormal joint detection model, couple the monitoring matrices, analyze the functional relationships between different types of abnormal states, and perform joint estimation of the battery state of the base station energy storage battery; Early warning is issued based on the estimated battery status output by the abnormal joint detection model and the actual working condition of the base station energy storage battery, and the battery management of the base station energy storage battery is optimized based on the early warning information; Based on the actual working conditions, the pre-processed multi-source monitoring data is identified online, and the monitoring parameters are selected as the battery status parameters of the base station energy storage battery, specifically: Generate multi-source monitoring data according to the environmental data and charging and discharging data of the base station energy storage battery, perform outlier processing, normalization processing and windowing processing on the multi-source monitoring data, and obtain pre-processed multi-source monitoring data; Preliminarily clustering the multi-source monitoring data to achieve charging and discharging condition division, obtaining a multi-source monitoring data sequence with condition labels, retrieving battery state monitoring and SOC estimation instances, and extracting characteristic parameters involved in each instance; Constructing and generating a characteristic parameter sequence according to the characteristic parameters, clustering the multi-source monitoring data sequence and the characteristic parameter sequence again, evaluating the similarity between the multi-source monitoring data sequence under different working conditions and each characteristic parameter sequence by using a dynamic time warping method, and obtaining the cluster to which the multi-source monitoring data sequence belongs by clustering; Extract all the sequences in the cluster where the multi-source monitoring data sequence belongs, and use multivariate empirical mode decomposition to decompose all the sequences, obtain several IMF component groups, and divide the IMF component groups into high-frequency IMF component groups and low-frequency IMF component groups; In the high-frequency IMF component group, the variable selection coefficient is calculated by using LASSO variable selection, the optimal characteristic parameter combination is selected in combination with the preset lag value, and the monitoring parameter is selected as the battery state parameter of the base station energy storage battery using the optimal characteristic parameter combination.

7. The all-round safety protection system for base station energy storage batteries according to claim 6, characterized in that: The battery state parameters are aggregated and the residual generator is used to obtain the residual data of the normal working condition of the energy storage battery to analyze the abnormal state, specifically: Based on the battery status parameters, the battery status parameter sequence of the base station energy storage battery under normal working conditions is screened, the screened battery status parameter sequence is aggregated, and a normal state model of the base station energy storage battery is constructed using a deep belief network; Divide the normal working condition intervals of the base station energy storage battery according to the parameter characteristics of the battery state parameter sequence, calculate the occurrence frequencies of different normal working condition intervals, and set the sampling quantity according to the occurrence frequencies to obtain training data samples, and train the normal state model; Read the abnormal working condition data of the base station energy storage battery, and obtain the battery state parameter sequence corresponding to the abnormal working condition data, and import the battery state parameter sequence with the abnormal working condition label into the normal state model to obtain the parameter estimation value; Calculate the residual data of the parameter estimation value and the battery state parameter sequence with abnormal working condition labels, obtain the relative entropy of the residual distribution, use the relative entropy to set the abnormality judgment thresholds of different types of abnormalities, and set different types of abnormality monitoring branches of the base station energy storage battery based on the abnormality judgment thresholds.

8. The all-round safety protection system for base station energy storage batteries according to claim 6, characterized in that: Construct an anomaly joint detection model, specifically: Obtain the abnormality category of the base station energy storage battery, analyze the abnormality mechanisms corresponding to different abnormality categories, calculate the correlation between the abnormality mechanisms, and read the degree of association between the abnormality categories based on the frequency and time of occurrence of different abnormality categories; Different anomaly categories are used as anomaly entities for graph mapping. The edge structure between different anomaly entities is obtained by using the correlation and association degree of different anomaly categories, and an undirected graph consisting of anomaly entities and edge structures is constructed. Reading the adjacency matrix according to the undirected graph, obtaining the monitoring matrix and abnormal monitoring results output by different types of abnormal monitoring branches of the base station energy storage battery, and assigning values ​​to the adjacency matrix using each monitoring matrix; A graph convolutional neural network is used to learn the undirected graph to construct an abnormality joint detection model, and the model is trained. The assigned adjacency matrix is ​​used as the model input to perform joint estimation of the battery state of the base station energy storage battery.

9. The all-round safety protection system for base station energy storage batteries according to claim 8, characterized in that: Perform joint estimation of the battery status of the base station energy storage battery, specifically: The assigned adjacency matrix is ​​imported into the graph convolutional neural network for encoding to obtain the embedded representation of each abnormal node. A multi-head attention mechanism is introduced to obtain the implicit relationship between the embedded representations of different abnormal nodes, and the implicit association weight is set according to the implicit relationship. According to the implicit association weights, weighted aggregation is performed using the neighbor aggregation mechanism of the graph convolutional neural network to update the embedded representation of each abnormal node, and the abnormal nodes that are abnormal at the current moment are marked by marking the abnormal monitoring results output by different types of abnormal monitoring branches; The inner product is calculated based on the updated embedding representation of the marked abnormal node and the updated embedding representations of other abnormal nodes, and the abnormal state is determined using the preset inner product threshold to obtain the abnormal score. When the inner product is greater than the preset inner product threshold, it means that the base station energy storage battery will have an abnormal state of the corresponding type at the preset prediction time.

Images