Method, device and equipment for positioning hidden danger battery cell in energy storage power station and readable storage medium

Abstract

The embodiment of the invention relates to a method, a device and equipment for positioning hidden danger battery cells in an energy storage power station and a readable storage medium, and the method comprises the steps: obtaining the pressure data of a first battery cell, carrying out the quick detection of a battery pack according to a pressure signal, and screening out a high-risk battery pack; according to a preset battery cell diagnosis model, diagnosing all battery cells of the high-risk battery pack to obtain hidden danger battery cells; screening the hidden danger battery cells through a preset clustering algorithm to obtain a high-risk battery cell cluster; and predicting the high-risk cell cluster according to a preset hybrid prediction model to obtain the remaining life of the cell. Through the scheme, the calculation amount and the energy consumption of the energy storage power station are greatly reduced. Moreover, according to the prediction result, the energy storage power station can carry out graded early warning on the hidden danger battery cells in advance, replacement can be completed before a fault occurs, and the energy utilization and use safety of the energy storage power station is greatly improved.

Description

TECHNICAL FIELD

[0001] The embodiment of the present application relates to the technical field of energy storage, in particular to a hidden danger battery positioning method, device and equipment in an energy storage power station and a readable storage medium. BACKGROUND

[0002] With the promotion of the "carbon peak and carbon neutral" strategy, the scale of electrochemical energy storage power stations taking lithium ion batteries as the core has rapidly expanded, becoming a key facility to support the stable operation of new energy power grids. Such power stations are usually composed of tens of thousands to millions of battery cells. Due to differences in production and manufacturing, material batches, transportation and installation, and operating conditions, there is inevitable initial inconsistency between the battery cells, which gradually intensifies in the long-term cycling process, forming significant performance differentiation, which seriously restricts the overall life and safe operation of the system. Inconsistent battery cells are more prone to overcharging or overdischarging, triggering chain reactions such as lithium precipitation, gas production, and internal short circuit, and even leading to thermal runaway and fire and explosion accidents. Therefore, how to prospectively, accurately and efficiently identify and manage abnormal battery cells has become a core challenge to improve the safety, life and economic returns of the power station.

[0003] The current industry commonly used method has obvious limitations: the fixed threshold alarm method is lagging behind and cannot detect early slow degradation faults; the full amount of high frequency data transmission to the cloud analysis method has high cost, large delay and low signal-to-noise ratio, and is difficult to practically promote; even if multiple independent diagnostic algorithms are combined, full unit calculation is still required, which is inefficient and cannot meet the comprehensive needs of large-scale energy storage systems for real-time, economy and accuracy. The existing technology lacks a new type of solution that can consider predictability, accuracy, real-time and low cost in a large-scale energy storage system. SUMMARY

[0004] In view of the above problems, the present application provides a hidden danger battery positioning method, device and equipment in an energy storage power station and a readable storage medium, which solves the problems of the prior art that the fixed threshold alarm method is lagging behind and cannot detect early slow degradation faults; the full amount of high frequency data transmission to the cloud analysis method has high cost, large delay and low signal-to-noise ratio, and is difficult to practically promote; even if multiple independent diagnostic algorithms are combined, full unit calculation is still required, which is inefficient and cannot meet the comprehensive needs of large-scale energy storage systems for real-time, economy and accuracy.

[0005] According to one aspect of the present application, a hidden danger battery positioning method in an energy storage power station is provided, which is applied in an energy storage power station, the energy storage power station has at least one battery pack, the battery pack includes a first battery cell and a second battery cell, and the first battery cell includes a pressure sensor; the method comprises:

[0006] obtaining pressure data of the first battery cell, and performing rapid detection on the battery pack according to the pressure signal to screen out high-risk battery packs;

[0007] diagnose all the battery cells of the high-risk battery pack according to a preset battery cell diagnosis model, to obtain a hidden danger battery cell;

[0008] screen the hidden danger battery cell through a preset clustering algorithm, to obtain a high-risk battery cell cluster;

[0009] predict the high-risk battery cell cluster according to a preset hybrid prediction model, to obtain a battery cell remaining life.

[0010] In some optional embodiments, a pressure signal of the battery cell is obtained, and a quick detection is performed on the battery pack according to the pressure signal, to screen out a high-risk battery pack, specifically including:

[0011] obtain pressure data of each first battery cell in the battery pack, and perform a dispersion analysis on the pressure data, to obtain a first dispersion;

[0012] determine whether the first dispersion is within a preset first threshold range;

[0013] if yes, the current battery pack is determined as a non-risk battery;

[0014] otherwise, detection data of the first battery cell and a second battery cell in the current battery pack is obtained, a composite dispersion analysis is performed on the current battery pack according to the detection data, to obtain a second dispersion, and the second dispersion is compared with a preset second threshold range, to determine whether the current battery pack is a high-risk battery pack;

[0015] wherein, the detection data at least includes battery cell voltage data and battery cell temperature data.

[0016] In some optional embodiments, all the battery cells of the high-risk battery pack are diagnosed according to a preset battery cell diagnosis model, to obtain a hidden danger battery cell, specifically including:

[0017] diagnose the first battery cell and the second battery cell through the battery cell diagnosis model, to obtain a fluctuation parameter;

[0018] compare the fluctuation parameter with a dynamic threshold of the battery pack where the battery cell is located, to determine whether the current battery cell is a hidden danger battery cell.

[0019] In some optional embodiments, the hidden danger battery cell is screened through a preset clustering algorithm, to obtain a high-risk battery cell cluster, specifically including:

[0020] obtain battery cell information of the hidden danger battery cell, and establish a high-dimensional feature file of the hidden danger battery cell according to the battery cell information;

[0021] analyze the high-dimensional feature file of the hidden danger battery cell according to the clustering algorithm, to obtain the high-risk battery cell cluster.

[0022] In some optional embodiments, the clustering algorithm is any one of a DBSCAN algorithm, a K-Means algorithm, and an Isolation Forest algorithm.

[0023] In some optional embodiments, the high-risk battery cell cluster is predicted according to a preset hybrid prediction model to obtain the remaining life of the battery cell, specifically including:

[0024] The first time sequence information of each hidden danger battery cell in the high-risk battery cell cluster is obtained, and the second time sequence information of the battery cluster in which the hidden danger battery cell is located is obtained.

[0025] The percentage energy is calculated according to the first time sequence information and the second time sequence information, and the original window data is obtained by dividing the percentage energy by a preset rated charge and discharge energy.

[0026] The health state of the hidden danger battery cell is obtained by performing sliding window filtering on the original window data.

[0027] The health states of the N hidden danger battery cells are linearly fitted to obtain a decay rate.

[0028] The remaining life of the battery cell is calculated and obtained according to the decay rate and a preset safety energy retention threshold.

[0029] In some optional embodiments, the linear fitting of the health states of the N hidden danger battery cells specifically includes:

[0030] Any one of a least square method, a weighted least square method, a Lasso regression algorithm, a ridge regression algorithm, an elastic network algorithm, a Huber algorithm, a Theil-Sen algorithm, a Quantile algorithm, a Gaussian process algorithm, a RANSAC algorithm, a least median square algorithm, and a linear discriminant analysis algorithm is used to linearly fit the health states of the N hidden danger battery cells.

[0031] According to another aspect of the example of the present application, a hidden danger battery cell positioning device in an energy storage power station is provided, which is applied in an energy storage power station, the energy storage power station has at least one battery pack, the battery pack includes a first battery cell and a second battery cell, the first battery cell includes a pressure sensor; the device includes:

[0032] A first screening module is configured to obtain pressure data of the first battery cell, and perform rapid detection on the battery pack according to the pressure signal to screen out a high-risk battery pack.

[0033] A second screening module is configured to diagnose all battery cells of the high-risk battery pack according to a preset battery cell diagnosis model to obtain a hidden danger battery cell.

[0034] A third screening module is configured to screen the hidden danger battery cell by a preset clustering algorithm to obtain a high-risk battery cell cluster.

[0035] and a life prediction module configured to predict the high-risk battery cell cluster according to a preset hybrid prediction model and obtain a remaining life of the battery cell.

[0036] According to another aspect of the present application, a hidden danger battery cell positioning device in an energy storage power station is provided, comprising a processor, a memory, a communication interface and a communication bus, the processor, the memory and the communication interface complete communication with each other through the communication bus;

[0037] The memory is configured to store at least one executable instruction, and the executable instruction causes the processor to perform the operations of the hidden danger battery cell positioning method in the energy storage power station.

[0038] According to another aspect of the present application, a readable storage medium is provided, the storage medium stores at least one executable instruction, and the executable instruction causes the life optimization device of the multi-cluster battery system to perform the operations of the hidden danger battery cell positioning method in the energy storage power station when the executable instruction runs on the hidden danger battery cell positioning device in the energy storage power station.

[0039] The present application provides a hidden danger battery cell positioning method, device, equipment and readable storage medium in an energy storage power station, which has the beneficial effects of:

[0040] The present application obtains the pressure data of the first battery cell, quickly detects the battery pack according to the pressure signal, screens out high-risk battery packs, diagnoses all battery cells of the high-risk battery packs according to a preset battery cell diagnosis model, obtains hidden danger battery cells, screens the hidden danger battery cells through a preset clustering algorithm, obtains a high-risk battery cell cluster, and predicts the high-risk battery cell cluster according to a preset hybrid prediction model to obtain the remaining life of the battery cell. Through the above scheme, the present application ensures that the high-cost calculation is only applied to a small number of most critical battery cells, greatly reduces the calculation amount and energy consumption of the energy storage power station. Moreover, according to the prediction result, the energy storage power station can realize early classification warning for the hidden danger battery cells and can complete replacement before failure, greatly improving the energy utilization and use safety of the energy storage power station.

[0041] The above description is only a summary of the technical solutions of the present application, in order to more clearly understand the technical means of the embodiments of the present application, the embodiments can be implemented according to the content of the specification, and in order to make the above and other purposes, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application are described below. BRIEF DESCRIPTION OF DRAWINGS

[0042] The accompanying drawings are only used to illustrate the embodiments and are not considered as limiting the present application. Moreover, the same reference signs are used to represent the same parts throughout the drawings. In the drawings:

[0043] Figure 1 A first embodiment flowchart of a method for positioning hidden danger battery cells in a power storage station according to the present application is shown, which is the method for positioning hidden danger battery cells in a power storage station according to the present application in embodiment 1;

[0044] Figure 2 A flowchart of step 110 according to the present application is shown, which is step 110 according to the present application in embodiment 1;

[0045] Figure 3 A flowchart of step 120 according to the present application is shown, which is step 120 according to the present application in embodiment 1;

[0046] Figure 4 A flowchart of step 140 according to the present application is shown, which is step 140 according to the present application in embodiment 1;

[0047] Figure 5 A structural diagram of a device for positioning hidden danger battery cells in a power storage station according to the present application is shown, which is the device for positioning hidden danger battery cells in a power storage station according to the present application in embodiment 2;

[0048] Figure 6 A structural diagram of an equipment for positioning hidden danger battery cells in a power storage station according to the present application is shown, which is the equipment for positioning hidden danger battery cells in a power storage station according to the present application in embodiment 3. DETAILED DESCRIPTION

[0049] Exemplary embodiments of the present application will be described in greater detail below, with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it is to be understood that the present application can be embodied in various forms without being limited by the embodiments set forth herein.

[0050] Embodiment 1:

[0051] Figure 1 An embodiment of a method for positioning hidden danger battery cells in a power storage station according to the present application is shown, which is applied in a power storage station, the power storage station has at least one battery pack, the battery pack includes a first battery cell and a second battery cell, the first battery cell includes a pressure sensor; the method includes:

[0052] 110, obtaining pressure data of the first battery cell, and performing rapid detection on the battery pack according to the pressure signal to screen out high-risk battery packs; in step 110, the pressure signal of the battery cell is obtained, and the battery pack is rapidly detected according to the pressure signal to screen out high-risk battery packs, specifically including: obtaining the pressure data of each first battery cell in the battery pack, and performing dispersion analysis on the pressure data to obtain a first dispersion; judging whether the first dispersion is within a preset first threshold range; if yes, the current battery pack is determined as a non-risk battery; otherwise, obtaining the detection data of the first battery cell and the second battery cell in the current battery pack, performing composite dispersion analysis on the current battery pack according to the detection data to obtain a second dispersion, and comparing the second dispersion with a preset second threshold range to judge whether the current battery pack is a high-risk battery pack; wherein the detection data at least includes battery cell voltage data and battery cell temperature data.

[0053] 120, diagnosing all battery cells of the high-risk battery pack according to a preset battery cell diagnosis model to obtain a hidden danger battery cell; in step 120, all battery cells of the high-risk battery pack are diagnosed according to a preset battery cell diagnosis model to obtain a hidden danger battery cell, specifically including: diagnosing the first battery cell and the second battery cell through the battery cell diagnosis model to obtain a fluctuation parameter; comparing the fluctuation parameter with the dynamic threshold of the battery pack where the battery cell is located to judge whether the current battery cell is a hidden danger battery cell.

[0054] 130, screening the hidden danger battery cell through a preset clustering algorithm to obtain a high-risk battery cell cluster; in step 130, the hidden danger battery cell is screened through a preset clustering algorithm to obtain a high-risk battery cell cluster, specifically including: obtaining battery cell information of the hidden danger battery cell, and establishing a high-dimensional feature file of the hidden danger battery cell according to the battery cell information; analyzing the high-dimensional feature file of the hidden danger battery cell according to the clustering algorithm to obtain a high-risk battery cell cluster.

[0055] 140, predicting the high-risk battery cell cluster according to a preset hybrid prediction model to obtain the remaining life of the battery cell. In step 140, the high-risk battery cell cluster is predicted according to a preset hybrid prediction model to obtain the remaining life of the battery cell, specifically including: obtaining the first time sequence information of each hidden danger battery cell in the high-risk battery cell cluster, and simultaneously obtaining the second time sequence information of the battery cluster where the hidden danger battery cell is located; calculating the percentage energy according to the first time sequence information and the second time sequence information, and dividing the percentage energy by the preset rated charge and discharge energy to obtain the original window data; obtaining the health state of the hidden danger battery cell by sliding window filtering on the original window data; linearly fitting the health states of N hidden danger battery cells to obtain a decay rate; calculating and obtaining the remaining life of the battery cell according to the decay rate and a preset safety energy retention threshold.

[0056] The application obtains the pressure data of the first battery cell, performs rapid detection on the battery pack according to the pressure signal, and screens out high-risk battery packs; diagnoses all battery cells of the high-risk battery packs according to a preset battery cell diagnosis model, obtains hidden danger battery cells; screens the hidden danger battery cells through a preset clustering algorithm, obtains a high-risk battery cell cluster; and predicts the high-risk battery cell cluster according to a preset hybrid prediction model, and obtains the remaining life of the battery cell. Through the above scheme, the application ensures that the high-cost calculation is only applied to a few most critical battery cells, greatly reducing the calculation amount and energy consumption of the energy storage power station. Moreover, according to the prediction result, the energy storage power station can realize early classification warning for the hidden danger battery cells, and can complete replacement before failure occurs, greatly improving the energy utilization and use safety of the energy storage power station.

[0057] In step 110, referring to Figure 2 , the pressure signal of the battery cell is obtained, and rapid detection is performed on the battery pack according to the pressure signal to screen out high-risk battery packs, specifically including:

[0058] 210, obtaining the pressure data of each first battery cell in the battery pack, and performing dispersion analysis on the pressure data to obtain a first dispersion;

[0059] 220, judging whether the first dispersion is within a preset first threshold range;

[0060] 230, if yes, the current battery pack is determined as a non-risk battery;

[0061] 240, otherwise, obtaining the detection data of the first battery cell and the second battery cell in the current battery pack, performing composite dispersion analysis on the current battery pack according to the detection data to obtain a second dispersion, and comparing the second dispersion with a preset second threshold range to judge whether the current battery pack is a high-risk battery pack;

[0062] In steps 210-240, the detection data at least includes battery cell voltage data and battery cell temperature data. The steps 210-240 aim to complete the first range contraction with the lowest calculation cost. The operation object is all battery packs PACK in the energy storage power station. The application introduces conditional priority screening logic for screening.

[0063] For systems equipped with internal gas pressure sensors: Given that gas pressure is the most sensitive indicator of early safety anomalies in the battery cell, the system will prioritize discrete analysis of gas pressure data alone. If the discrete degree of gas pressure is within the normal range, it is considered that there is no obvious group risk in this macroscopic scanning stage, and there is no need to judge other parameters such as voltage and temperature, and it is directly determined to pass. Only when the discrete degree of gas pressure is abnormal, the system will start step 240 to analyze the composite discrete degree of other parameters such as voltage and temperature, in order to further confirm the risk. This priority judgment mechanism makes use of the most sensitive indicator, greatly reducing unnecessary calculations.

[0064] For systems without internal gas pressure sensors, each PACK is considered as a whole, and the "health status" of the PACK is quickly evaluated by calculating a composite discrete degree index that can macroscopically represent the consistency of the battery cell group inside the PACK, for example, the standard deviation and range of its internal voltage, temperature and other parameters.

[0065] Through the above path, this step can complete the scanning of thousands of PACKs in the whole station at the minute level, especially in the early and middle stages of the operation of the power station, which directly narrows the analysis object from all PACKs in the whole station to at least a few, such as 1% ~ 5%; "High-risk battery packs" that show abnormal discrete characteristics directly reduce the amount of data for subsequent analysis by an order of magnitude.

[0066] In step 120, see Figure 3 , diagnose all battery cells of the high-risk battery pack according to the preset battery cell diagnosis model to obtain the hidden danger battery cell, specifically including:

[0067] 310, diagnose the first battery cell and the second battery cell by the battery cell diagnosis model to obtain the fluctuation parameter;

[0068] 320, compare the fluctuation parameter with the dynamic threshold of the battery pack where the battery cell is located to determine whether the current battery cell is a hidden danger battery cell.

[0069] In steps 310-320, the present application aims to perform more detailed calculations within the range of high-risk battery packs that have been greatly narrowed. Its operation object is all the cells inside the high-risk battery packs screened out in step 110. Specifically, the core logic is to enable a more computationally intensive refined diagnostic model, to calculate a fluctuation composite parameter that can reflect the microscopic abnormal behavior of a single cell, such as the volatility of its own gas pressure time series and its stability in the gas pressure ranking within the PACK, and to determine whether the cell is a "troubled cell" by comparing it with a threshold dynamically generated according to the PACK it is in. Since the calculation is only performed in a small number of high-risk PACKs, the total computational burden is still very low. This step will analyze the granularity from the "battery pack level" to the "cell level", and further locate a smaller number of troubled cells with real early failure characteristics from a large number of cells in the battery packs of interest.

[0070] In step 130, the high-risk cell cluster is obtained by screening the troubled cells through a preset clustering algorithm, specifically including: obtaining cell information of the troubled cells, and establishing a high-dimensional feature file of the troubled cells according to the cell information; analyzing the high-dimensional feature file of the troubled cells according to the clustering algorithm to obtain the high-risk cell cluster.

[0071] In an embodiment, the present embodiment is a process of confirming and classifying the troubled cells found in step 120, aiming to exclude single cells with low risk relevance and provide clues for the real high-risk failure traceability. Its operation object is only the troubled cells located in step 120. The core logic is to establish a high-dimensional feature file containing multi-dimensional information such as working conditions, thermodynamics and environmental correlations for these troubled cells, and to apply a clustering algorithm to analyze these high-dimensional files to find a group of cells with similar abnormal characteristics. Since the number of samples processed is extremely small, the clustering analysis is fast, and this step focuses on the high-risk cell cluster with clear common characteristics after cross-validation, instead of the scattered troubled cells. The clustering algorithm is any one of DBSCAN algorithm, K-Means algorithm and Isolation Forest algorithm.

[0072] In step 140, referring to Figure 4 , the high-risk cell cluster is predicted according to a preset hybrid prediction model to obtain the cell remaining life, specifically including:

[0073] 410, obtaining the first time series information of each troubled cell in the high-risk cell cluster, and obtaining the second time series information of the battery cluster where the troubled cell is located;

[0074] 420, calculating the percentage energy according to the first time series information and the second time series information, and obtaining the original window data by dividing the percentage energy by the preset rated charge and discharge energy;

[0075] 430, obtaining the health state of the hidden danger battery cell by performing sliding window filtering on the original window data;

[0076] 440, performing linear fitting on the health states of the N hidden danger battery cells to obtain the decay rate;

[0077] 450, calculating the remaining life of the battery cell according to the decay rate and the preset safety energy retention threshold.

[0078] In steps 410-450, the operation object of the embodiment is the hidden danger battery cell in the high-risk battery cell cluster identified in step 130. The core logic is to start the most complex health state or safety state estimation, and to establish the remaining service life or remaining safe service life based thereon, using a hybrid prediction model of service life, such as a combination of a mechanism model and a machine learning model, to accurately predict the end of life or safety threshold.

[0079] Specifically, the first time sequence information is battery cell voltage data, battery cell temperature data, and battery cell SOC data, and the second time sequence information is cluster current data and cluster SOC data of the battery cluster where the hidden danger battery cell is located. The above time sequence information must contain a time stamp, and the collection period of the same frame of data must comply with the national standard to avoid the influence of large differences between the hidden danger battery cell data due to different collection times on the calculation result.

[0080] The calculation of the rated energy can be divided into two cases: the first case is to use the cycle test battery cells of the same batch as the project cluster currently used by the power station to avoid large tolerances caused by different batches, and to obtain the rated charge and discharge energy at a specific temperature based on the cycle test data. The specific steps are as follows:

[0081] The charge capacity, discharge capacity, charge energy, and discharge energy of each cycle aging test of the batch at a specified temperature are recorded to form a data array. When the discharge capacity is equal to the rated capacity, the corresponding charge capacity, charge energy, and discharge energy are recorded as the "rated charge capacity", "rated charge energy", and "rated discharge energy" at the temperature.

[0082] The "rated charge capacity" at the current temperature is referred to as C disc-rated T The "rated charge capacity" is C char-rated T The "rated charge energy" is E disc-rated T The "rated discharge energy" is E char-rated T

[0083] The second case: record the charge capacity, discharge capacity, charge energy, and discharge energy of the battery during capacity sorting as the C disc-ratedT , C char-rated T , E disc-rated T , E char-rated T Note that in this way, the partition information under multiple temperature conditions needs to be recorded, preferably two different temperatures are used for partitioning, and the partition information of the two temperatures is recorded, preferably the temperatures used are 25°C and 40°C, and the MES data with the partition information and the battery identity code is uploaded to the cloud platform database.

[0084] After obtaining the rated charging energy and the rated discharging energy, the E disc-reted T , E char-rated T is divided by 100, that is, the E per-disc-rated T , E per-char-rated T .

[0085] For temperature points without setting cycle test samples, based on the number of test temperature samples, the E per-disc-rated T and E per-char-rated T are calculated by piecewise linear interpolation method. Since this method is commonly used, it is not described in detail. The temperature selection can be the average temperature of the charging period in the window time and the average temperature of the discharging period in the window time. The E per-disc-rated T , E per-char-rated T of the hidden danger battery cells in the current batch can be obtained. Preferably, the E per-disc-rated T .

[0086] According to the first time sequence information and the second time sequence information, the percentage energy is calculated, and the battery cluster in actual operation is

[0087] The time sequence data obtained or calculated by the BCU includes the current, SOC of the battery cluster, the voltage, SOC, and temperature information of each hidden danger battery cell in the cluster.

[0088] According to these time sequence information, a window time is selected, which needs to include at least one of the charging or discharging system states, and the window time needs to include the change of the battery cell voltage data, the battery cell temperature data, and the battery cell SOC data.

[0089] When there is no battery cell SOC data, the cluster SOC data can be used instead, but this will reduce the percentage charging and discharging energy of the hidden danger battery cells. Preferably, the charging energy Echar T , discharge energy E disc T , and the SOC change ΔSOC when the single cell is charged char T and the SOC change ΔSOC when discharged disc T to calculate the percentage energy charge energy E per-char T and the percentage discharge energy E per-disc T , and record the single cell SOC change in the window time

[0090] In the current window of the hidden cell E per-char T i and E per-disc T i , divided by the rated charge and discharge energy E per-disc-rated T or e per-char-rated T , the charge energy retention rate E retention-char-i or the electrical energy retention rate E retention-disc-i of the hidden cell in the window time is obtained. Preferably, the discharge energy retention rate is used as the health degree measurement index of the battery hidden cell.

[0091] The data of the percentage energy in the calculation can already be used as the health degree of the battery hidden cell in the window time under ideal conditions, but considering various uncertain factors and cumulative tolerances in the actual operation of the project, it is necessary to perform an outlier screening and interpolation filling to ensure the stability of the window timing data.

[0092] The specific screening and interpolation process is as follows:

[0093] Step one: record the missing values in the hidden cell data sequence as "outliers".

[0094] Step two: select a sliding window W for each hidden cell data sequence, and the number of original data sequence should be greater than or equal to W, and select a threshold N according to the window size threshold , and perform outlier detection on the data y i in each window.

[0095]

[0096] Wherein,

[0097] Preferably, W=50, N threshold= 2.33, which means the number of original data sequence should be greater than or equal to 50.

[0098] When there are currently A data points, the number of sliding windows is A-1, and the sliding window traverses the time series data, and the data points in each window are normalized by Z-score. For the normalized data y Zi , if its absolute value |y Zi |>N threshold , the original data point y i is recorded as an "outlier". The set of outlier sequences is denoted as Ω abnormal , and the set of non-outlier sequences is denoted as Ω normal . Note that the processed data sequence here is aligned with the index of the original sequence to ensure the integrity of the data.

[0099] Step three: obtain the neighborhood valid data of each "outlier", select the neighborhood window W s , then the neighborhood data N i of the outlier y i = {y i -W s , y i -W s +1,..., y i +W s -1 s , y i +W s}, if there are j y i ∈Ω normal in N i , then The preferred W s = 20, note that when the outlier data point is at both ends of the data, the number of data points in its left or right neighborhood does not meet the condition of W s , at least one end of the neighborhood needs to meet the condition, for example, when i = 1, N i = {y i +1, y i +W s -1 s , y i +W s}.

[0100] After the above "sliding window filling outlier" process, the stability of the data is ensured.

[0101] The last data of the data sequence can be used as the health degree of the battery hidden danger cell in this window time, preferably, the discharge energy retention rate after the sliding window filling outlier is used as the health degree of the battery hidden danger cell. Denote the charge energy retention rate E retention-char-ior electrical energy retention rate E retention-disc-i At this time, the efficiency of the window time of the hidden battery cell can be obtained by dividing the interpolated discharge energy retention rate by the charge electrical energy retention rate:

[0102] For the health degree of the above battery hidden cell in multiple consecutive window times, multiple linear fitting methods can be used for fitting, including but not limited to common least squares method, weighted least squares method, Lasso regression, ridge regression, elastic network, Huber, Theil-Sen, Ouantile, Gaussian process, RANSAC, least median squares, linear discriminant analysis algorithm. Preferably, a linear kernel is used for fitting. Since the data has been subjected to the interpolation process described above, there are no obvious outliers, and the least squares method is preferably used for linear fitting. The calculation process is transparent, simple, easy to deploy and implement.

[0103] Preferably, the health degree of the latest M hidden cell consecutive window times is selected for linear fitting, preferably M = 200. When the data set to be fitted does not meet this condition, the full set of no less than the fitting threshold M0 is used for fitting calculation. Here, the fitting data must be greater than or equal to the fitting threshold M0, preferably M0 = 100. For the input quantity of fitting, the health degree of the battery hidden cell in multiple consecutive window times is required as the dependent variable y, and the natural number of the current window is required as the independent variable x. The last frame natural number is denoted as iL.

[0104] For the result of linear fitting, it is denoted as y i = m i x + b i At this time, if the slope parameter satisfies m i <0, the coefficient of determination R i 2 >R0, P i <P0 is normal, and it is considered that the result meets the expectation. At this time, m i is denoted as the energy retention rate decay rate of the cell with the number i. Preferably, R0 = 0.85 and P0 = 1‰.

[0105] The safe charging energy retention rate threshold Threshold E-char and the safe discharge energy retention rate threshold Threshold E-disc are selected, preferably Threshold E-disc The remaining available window time of a hidden cell is calculated according to the following formula:

[0106]

[0107] The preferred fixed window time is, for example, 12H or 24H, by which the remaining safe use years R can be calculated available-time-i .

[0108] Taking the window time h hours as an example, the remaining safe use days of the hidden danger battery cell is

[0109] The remaining safe use years, which can also be calculated by another scheme:

[0110] Calculate the average value of the full amount of window ΔSOC T , that is, the ΔSOC from 1 to the last frame iL T The average value or mode is denoted as

[0111]

[0112] At this time, R available-cycle-i The remaining safe cycle number. According to the actual running cycle number of each window, the real-time remaining available cycle number can be simply calculated.

[0113] Among them, the remaining available window and the remaining safe use years can be used as the remaining life of the battery cell.

[0114] Embodiment 2:

[0115] Figure 5 An embodiment of a hidden danger battery cell positioning device in an energy storage power station is shown, which is applied in an energy storage power station, the energy storage power station has at least one battery pack, the battery pack includes a first battery cell and a second battery cell, the first battery cell includes a pressure sensor; the hidden danger battery cell positioning device 500 in the energy storage power station includes a first screening module 510, a second screening module 520, a third screening module 530 and a life prediction module 540.

[0116] Among them, the first screening module is used to execute step 110 in embodiment 1, and specifically includes: acquiring the pressure signal of the battery cell, and performing rapid detection on the battery pack according to the pressure signal, and screening out high-risk battery packs, specifically including: acquiring the pressure data of each first battery cell in the battery pack, and performing dispersion analysis on the pressure data to obtain the first dispersion; determine whether the first dispersion is within the preset first threshold range; if yes, determine the current battery pack as a non-risk battery; otherwise, acquire the detection data of the first battery cell and the second battery cell in the current battery pack, perform composite dispersion analysis on the current battery pack according to the detection data to obtain the second dispersion, and compare the second dispersion with the preset second threshold range to determine whether the current battery pack is a high-risk battery pack; wherein the detection data at least includes battery cell voltage data and battery cell temperature data.

[0117] The second screening module is configured to perform step 120 in embodiment 1, and specifically includes diagnosing all battery cells of the high-risk battery pack according to a preset battery cell diagnosis model to obtain a hidden danger battery cell, specifically including diagnosing the first battery cell and the second battery cell through the battery cell diagnosis model to obtain a fluctuation parameter; and comparing the fluctuation parameter with a dynamic threshold of the battery pack where the battery cell is located to determine whether the current battery cell is a hidden danger battery cell.

[0118] The third screening module is configured to perform step 130 in embodiment 1, and specifically includes screening the hidden danger battery cell through a preset clustering algorithm to obtain a high-risk battery cell cluster, specifically including obtaining battery cell information of the hidden danger battery cell, and establishing a high-dimensional feature file of the hidden danger battery cell according to the battery cell information; and analyzing the high-dimensional feature file of the hidden danger battery cell according to the clustering algorithm to obtain the high-risk battery cell cluster.

[0119] The life prediction module is configured to perform step 140 in embodiment 1, and specifically includes predicting the high-risk battery cell cluster according to a preset hybrid prediction model to obtain a battery cell remaining life, specifically including obtaining first time sequence information of each hidden danger battery cell in the high-risk battery cell cluster, and simultaneously obtaining second time sequence information of a battery cluster where the hidden danger battery cell is located; calculating a percentage energy according to the first time sequence information and the second time sequence information, and dividing the percentage energy by a preset rated charge and discharge energy to obtain original window data; obtaining a health state of the hidden danger battery cell by performing sliding window filtering on the original window data; performing linear fitting on the health states of N hidden danger battery cells to obtain a decay rate; and calculating and obtaining the battery cell remaining life according to the decay rate and a preset safety energy retention threshold.

[0120] Embodiment 3

[0121] Figure 6 An embodiment of a hidden danger battery cell positioning device in a power storage power station is shown in a structural schematic diagram. The specific embodiments of the present application do not limit the specific implementation of the hidden danger battery cell positioning device in a power storage power station.

[0122] As shown in Figure 6 , the hidden danger battery cell positioning device in a power storage power station can include a processor, a communications interface, a memory, and a communications bus.

[0123] The processor 610, the communications interface 640, and the memory 620 can communicate with each other through the communications bus 630. The communications interface is configured to communicate with network elements such as clients or other servers. The processor is configured to execute the program 650, and specifically can execute the related steps in the above-mentioned embodiment of the method for positioning a hidden danger battery cell in a power storage power station.

[0124] In particular, the program can comprise program code comprising computer-executable instructions.

[0125] The processor can be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to perform the functions of an embodiment of the application. The one or more processors of the battery storage system internal dangerous battery cell positioning device can be the same type of processor, such as one or more CPUs; or different types of processors, such as one or more CPUs and one or more ASICs.

[0126] The memory is used to store the program. The memory can include a high-speed RAM memory, and can also include a non-volatile memory, such as at least one disk memory.

[0127] The program can be specifically executed by a battery storage system internal dangerous battery cell positioning device Figure 1 between step 110 and step 140.

[0128] Embodiment 4:

[0129] According to another aspect of the present application, a readable storage medium is provided, and the storage medium stores at least one executable instruction, which, when executed on a life optimization device of a battery storage system internal dangerous battery cell positioning device as described above, causes the life optimization device of the multi-cluster battery system to perform the operations of the battery storage system internal dangerous battery cell positioning method as described above.

[0130] The refinement strategy of the present application ensures that this high-cost calculation is only applied to a small number of the most critical battery cells, making it completely feasible in practical applications. According to the prediction results, the system can generate a hierarchical early warning and proactive maintenance work order weeks or even months in advance, guiding maintenance personnel to complete replacement before failure occurs. The present application uses a "gradual refinement" sparse screening method to gradually decrease the analysis range, and is expected to reduce the cloud computing power demand by more than 50% throughout the life cycle, significantly saving servers and bandwidth. The hierarchical screening architecture can be smoothly applied to battery storage systems of different sizes without causing a computing power bottleneck.

[0131] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other apparatus. Furthermore, embodiments of the present application are not described in terms of any particular programming language.

[0132] In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the application can be practiced without these specific details. In other instances, well-known methods, structures and techniques have not been described in detail in order to avoid obscuring the understanding of this description. Like reference numerals refer to like elements throughout. Similarly, while operations can be depicted in the drawings in a particular order, this should not be understood as requiring or

[0133] It is understood by those skilled in the art that modules in the apparatus of the embodiments can be adapted and placed in one or more apparatuses other than the embodiments. Modules or units or components in the embodiments can be combined into one module or unit or component, and further can be divided into multiple sub-modules or sub-units or sub-components. Except that at least some of such features and / or processes or units are mutually exclusive.

[0134] It is noted that the foregoing examples have been provided merely for the purpose of explanation and are in no way to be construed as limiting of the present application. While the application has been described with reference to preferred embodiments, it is understood that the words which have been used herein are words of description, and that details of the preferred embodiments are not intended to limit the scope of the application. In fact, it is recognized that changes can be made in the details, especially in matters of shape, size, arrangement of parts and the like, without departing from the spirit of the application. Accordingly, whatever the scope of the application, it is submitted that it covers all alternatives, modifications and equivalents. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps other than those listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The application can be implemented by means of both hardware and software, and any combination thereof. In a unitary claim, several devices, apparatuses or means can be listed, even though they are not all mutually exclusive. The use of the term "at least" followed by a list of one or more items should be interpreted as including at least one of the items but it does not exclude the presence of others not listed. The use of the term "one" followed by a list of one or more items should be interpreted as including one of the items but it does not exclude the presence of others not listed. The use of the term "first", "second" and "third", etc. does not limit the quantity and / or order of those items. The use of the terms "first", "second", "third", and / or "fourth" and / or the like, merely identifies one or more of the enumerated or other items as different from another. The suffix "(s)" as used herein is only used as an abbreviation of "one or more".

Claims

1. A method for locating potentially hazardous battery cells within an energy storage power station, characterized in that, Applied to an energy storage power station, the energy storage power station having at least one battery pack, the battery pack including a first battery cell and a second battery cell, the first battery cell including a pressure sensor; the method includes: The pressure data of the first battery cell is obtained, and the battery pack is quickly detected based on the pressure signal to screen out high-risk battery packs. The high-risk battery pack is diagnosed based on a preset cell diagnostic model to identify potentially hazardous cells. The potential battery cells are screened using a pre-defined clustering algorithm to obtain high-risk battery cell clusters; Based on a preset hybrid prediction model, the high-risk cell cluster is predicted to obtain the remaining lifespan of the cells.

2. The method for locating hidden battery cells in an energy storage power station according to claim 1, characterized in that, Acquire the pressure signal of the battery cell, and perform rapid detection of the battery pack based on the pressure signal to screen out high-risk battery packs, specifically including: The pressure data of each first cell in the battery pack is obtained, and the pressure data is subjected to dispersion analysis to obtain the first dispersion. Determine whether the first dispersion is within a preset first threshold range; If so, the current battery pack will be determined as a risk-free battery; Otherwise, obtain the detection data of the first cell and the second cell in the current battery pack, perform composite dispersion analysis on the current battery pack based on the detection data to obtain the second dispersion, and compare the second dispersion with the preset second threshold range to determine whether the current battery pack is a high-risk battery pack. The detection data includes at least cell voltage data and cell temperature data.

3. The method for locating hidden battery cells in an energy storage power station according to claim 1, characterized in that, Based on a preset cell diagnostic model, all cells in the high-risk battery pack are diagnosed to identify cells with potential problems. Specifically, this includes: The first and second cells were diagnosed using a cell diagnostic model to obtain fluctuation parameters. The fluctuation parameters are compared with the dynamic threshold of the battery pack in which the cell is located to determine whether the current cell is a potential problem cell.

4. The method for locating hidden battery cells in an energy storage power station according to claim 1, characterized in that, The potential battery cells are screened using a pre-defined clustering algorithm to obtain high-risk cell clusters, specifically including: Obtain the cell information of the potentially hazardous battery cell, and establish a high-dimensional feature profile of the potentially hazardous battery cell based on the cell information; The high-dimensional feature profiles of the potentially hazardous battery cells are analyzed using a clustering algorithm to obtain clusters of high-risk battery cells.

5. The method for locating hidden battery cells in an energy storage power station according to claim 4, characterized in that, The clustering algorithm is any one of the following: DBSCAN algorithm, K-Means algorithm, and Isolation Forest algorithm.

6. The method for locating hidden battery cells in an energy storage power station according to claim 1, characterized in that, Based on a preset hybrid prediction model, the high-risk cell clusters are predicted to obtain the remaining lifespan of the cells, specifically including: Obtain the first time-series information of each potentially hazardous cell in a high-risk cell cluster, and simultaneously obtain the second time-series information of the battery cluster containing the potentially hazardous cell. Calculate the percentage energy based on the first timing information and the second timing information, and obtain the original window data by dividing the percentage energy by the preset rated charge and discharge energy; The health status of potentially hazardous battery cells can be obtained by applying sliding window filtering to the raw window data. The degradation rate is obtained by linearly fitting the health status of N potentially hazardous battery cells. The remaining lifespan of the battery cell is calculated based on the decay rate and the preset safe energy retention threshold.

7. The method for locating hidden battery cells in an energy storage power station according to claim 6, characterized in that, A linear fit was performed on the health status of N potentially hazardous battery cells, specifically including: The health status of N potentially hazardous battery cells is linearly fitted using any one of the following methods: least squares, weighted least squares, Lasso regression, ridge regression, elastic network algorithm, Huber algorithm, Theil-Sen algorithm, Ouantile algorithm, Gaussian process algorithm, RANSAC algorithm, least median square algorithm, and linear discriminant analysis algorithm.

8. A device for locating potentially hazardous battery cells in an energy storage power station, characterized in that, Applied in an energy storage power station, the energy storage power station has at least one battery pack, the battery pack including a first battery cell and a second battery cell, the first battery cell including a pressure sensor; the device includes: The first screening module is used to acquire the pressure data of the first battery cell and to quickly detect the battery pack based on the pressure signal, thereby screening out high-risk battery packs. The second screening module is used to diagnose all the cells of the high-risk battery pack according to the preset cell diagnosis model and identify the cells with potential problems. The third screening module is used to screen the potential battery cells using a preset clustering algorithm to obtain high-risk battery cell clusters; And a lifespan prediction module, used to predict the remaining lifespan of the high-risk cell clusters based on a preset hybrid prediction model.

9. A device for locating potentially hazardous battery cells in an energy storage power station, characterized in that, include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction, which causes the processor to perform the operation of the method for locating hidden battery cells in an energy storage power station as described in any one of claims 1-7.

10. A readable storage medium, characterized in that, The storage medium stores at least one executable instruction, which, when executed on the hidden cell locating device in the energy storage power station as described in claim 9, causes the life optimization device of the multi-cluster battery system to perform the operation of the hidden cell locating method in the energy storage power station as described in any one of claims 1-7.

Images