Lithium battery data dynamic acquisition method and system of energy storage power station

Abstract

The invention provides a lithium battery data dynamic acquisition method and system of an energy storage power station, and the method comprises the steps: obtaining fault data of a target energy storage power station in a current data acquisition period when the current data acquisition period is ended; determining a plurality of target fault types according to the fault data and each fault frequency threshold; according to each target fault type and a preset parameter-fault correlation degree matrix, performing first compensation on a basic acquisition level of each characteristic parameter of the target energy storage power station, and determining a first acquisition level of each characteristic parameter; according to the variation amplitude of each feature parameter in the corresponding preset time window, performing second compensation on the first acquisition level of each feature parameter, and determining the second acquisition level of each feature parameter; and performing real-time data acquisition on each characteristic parameter of the target energy storage power station in the next data acquisition period according to each second acquisition level, thereby improving pertinence and efficiency of data acquisition.

Description

Technical Field

[0001] This application relates to the fields of data analysis and data acquisition technology, and in particular to a method and system for dynamic acquisition of lithium battery data in an energy storage power station. Background Technology

[0002] With the acceleration of energy transition, lithium battery energy storage power stations are increasingly widely used in power systems due to their advantages such as high energy density, long cycle life, and fast response speed, in areas such as peak shaving and valley filling, frequency and voltage regulation, and backup power. However, lithium battery energy storage power stations also face many safety hazards during operation, such as thermal runaway, capacity decay, and battery short circuits. These faults not only affect the normal operation of the energy storage power station but may also cause serious safety accidents such as fires and explosions, resulting in huge losses to people and property.

[0003] Currently, online monitoring of energy storage power stations typically involves collecting data on different characteristic parameters of lithium batteries at a manually set sampling frequency. However, because different energy storage power stations have different operating conditions, such as station size and operating history, the frequency and types of failures that are prone to occur vary. Furthermore, different failure types are related to different characteristic parameters. Therefore, the importance of each characteristic parameter to an energy storage power station varies, necessitating targeted and effective data collection to obtain more valuable fault information. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides a method and system for dynamic acquisition of lithium battery data in energy storage power stations, which improves the relevance and efficiency of data acquisition and provides strong support for the long-term stable operation of energy storage power stations.

[0005] In a first aspect, embodiments of this application provide a method for dynamically acquiring lithium battery data in an energy storage power station, including: At the end of the current data acquisition cycle, acquire the fault data of the target energy storage power station during the current data acquisition cycle; Based on the fault data and various fault frequency thresholds, several target fault types are determined. The various fault frequency thresholds are calculated based on the operating data of the target energy storage power station in the previous data collection cycle. Based on each of the target fault types and the preset parameter-fault correlation matrix, the basic acquisition level of each characteristic parameter of the target energy storage power station is compensated first to determine the first acquisition level of each characteristic parameter. Based on the variation range of each of the feature parameters within the corresponding preset time window, a second compensation is performed on the first acquisition level of each of the feature parameters to determine the second acquisition level of each of the feature parameters. Based on each of the second acquisition levels, determine the acquisition frequency of each feature parameter in the next data acquisition cycle; In the next data acquisition cycle, real-time data acquisition is performed on each of the characteristic parameters of the target energy storage power station according to the respective acquisition frequencies.

[0006] This application provides a method for dynamically acquiring lithium battery data in an energy storage power station. First, at the end of the current data acquisition cycle, fault data of the target energy storage power station is acquired. Then, based on historical operating data, Bayesian inference is used to set thresholds for the occurrence frequency of various faults, thereby identifying high-frequency target fault types. At the end of each cycle, the actual fault data is compared with the predicted data from Bayesian conjugate inference. Based on the comparison results, the acquisition level and frequency of characteristic parameters are dynamically adjusted. Furthermore, this application uses a preset parameter-fault correlation matrix to initially compensate the basic acquisition level, forming a first acquisition level. Then, a second compensation is performed based on the variation amplitude of characteristic parameters within a preset time window, forming a second acquisition level. Finally, the acquisition frequency for the next cycle is determined based on the second acquisition level, and real-time data acquisition is performed. Therefore, this embodiment possesses a high degree of adaptability and intelligence, enabling dynamic adjustment of the acquisition strategy according to the actual operating status of the energy storage power station. This more effectively captures potential fault information, significantly improving the targeting and efficiency of data acquisition, overcoming the resource waste or key data omission problems common in traditional fixed-frequency acquisition modes, and providing strong support for the long-term stable operation of energy storage power stations.

[0007] Furthermore, based on the fault data and various fault frequency thresholds, several target fault types are determined, including: The number of occurrences for each fault type is determined based on the fault data. The occurrence count of each fault is compared with the corresponding fault count threshold. Fault types with a fault occurrence count greater than the fault count threshold are identified as the target fault types, thereby determining several target fault types.

[0008] This application further clarifies the mechanism for determining target fault types. Specifically, it includes counting the occurrence frequency of each fault type and comparing it with a preset fault frequency threshold to filter out target fault types that require special attention. Through quantitative analysis, it achieves objectivity and accuracy in fault identification, effectively reducing errors that may be introduced by subjective human judgment. By comparing the number of fault occurrences with the corresponding thresholds, the system can quickly identify frequently occurring fault types and subsequently adjust the collection frequency of various characteristic parameters to improve the targeting and efficiency of data collection.

[0009] In one possible implementation, when calculating the fault count threshold for any fault type based on the operating data of the target energy storage power station in the previous data acquisition cycle, the calculation of the respective fault count thresholds based on the operating data of the target energy storage power station in the previous data acquisition cycle includes: Obtain the operating data of the target energy storage power station in the previous data acquisition cycle. The operating data includes the number of faults of the fault type and the lithium battery operating energy data. If there is no prior distribution of the failure rate of the fault type under unit operating energy in the current data acquisition period, then the prior modeling of the failure rate of the fault type under unit operating energy is performed based on the historical data of the target energy storage power station to obtain the prior distribution. Based on the operational data and the prior distribution, Bayesian conjugate inference is performed to determine the posterior distribution in the Bayesian estimation method; Based on the posterior distribution, the posterior number of fault occurrences of the fault type in the current data collection period is predicted using a Bayesian estimation method, and the posterior number of fault occurrences is used as the fault count threshold for the fault type. The posterior distribution is used as the prior distribution of the failure rate of the fault type under unit operating energy in the next data acquisition cycle.

[0010] This application provides a method for determining a fault count threshold. At the end of the current data acquisition cycle, the prior distribution of the current data acquisition cycle is obtained. Then, combined with operational data, Bayesian conjugate inference is used to obtain the posterior distribution of the current data acquisition cycle. Finally, the number of faults in the current data acquisition cycle is predicted based on the posterior distribution, and this predicted value is used as the fault count threshold for the current data acquisition cycle. This application introduces Bayesian inference theory from statistics, making the setting of the fault threshold more scientific and reasonable, and effectively addressing uncertainties in the operating environment. Furthermore, this embodiment also uses the posterior distribution of each cycle as the prior distribution of the next data acquisition cycle, realizing iterative updates of Bayesian parameters. This enables Bayesian conjugate inference to have self-learning and continuous optimization capabilities, dynamically adjusting the prediction model based on continuously accumulated operational data, thereby improving the adaptive capability and robustness of the data acquisition system and providing solid data support for the long-term reliable operation of energy storage power stations. Furthermore, this application embodiment also considers the scenario of performing Bayesian estimation for the first time. In this case, the prior distribution cannot be obtained directly. Therefore, based on the historical data of the target energy storage power station, the initial prior distribution is obtained through prior modeling, which ensures the smooth progress of Bayesian parameter iteration and update.

[0011] Furthermore, based on historical data from the target energy storage power station, prior modeling is performed on the failure rate of the fault type per unit operating energy to obtain the prior distribution, including: Based on the length of the data collection cycle, the historical operating data is divided into several sub-historical operating data. A priori model is performed on the failure rate of the aforementioned failure type under unit operating energy to construct the initial prior distribution in the Bayesian estimation method; Based on the historical data of each sub-sub-distribution, the method of moments is used to estimate the parameters of the initial prior distribution, determine the parameter values ​​of the initial prior distribution, and then obtain the prior distribution.

[0012] This application provides a modeling process for a prior distribution, which involves dividing historical operational data into several subsets, constructing an initial prior distribution, and estimating parameters using the method of moments to obtain the final prior distribution. This embodiment, through refined processing of historical data, ensures the accuracy and representativeness of the prior distribution, laying a reliable foundation for subsequent Bayesian inference. This makes the calculation of the fault threshold more robust and reliable, and improves the targeting and efficiency of subsequent data collection.

[0013] Furthermore, the lithium battery data dynamic acquisition method also includes updating the posterior distribution and the fault count threshold if the target energy storage power station has undergone non-quantitative optimization in the current data acquisition cycle, specifically as follows: Based on the prior distribution, the number of first prior fault occurrences of the fault type in the current data acquisition period is predicted using a Bayesian estimation method. The parameters of the prior distribution and the posterior distribution are updated by a preset adjustment coefficient to obtain the corrected prior distribution and the corrected posterior distribution. Based on the modified prior distribution, the number of second prior fault occurrences of the fault type in the current data acquisition period is predicted using a Bayesian estimation method. Based on the modified posterior distribution, the first posterior fault occurrence count of the fault type in the current data acquisition period is predicted using a Bayesian estimation method. Based on the difference between the first prior fault occurrence count and the second prior fault occurrence count, the first posterior fault occurrence count is adjusted to obtain the second posterior fault occurrence count, and the second posterior fault occurrence count is used as the fault occurrence count threshold for the fault type. The modified posterior distribution is used as the prior distribution of the failure rate of the fault type under unit operating energy in the next data acquisition cycle.

[0014] This application further considers a threshold update mechanism for non-quantitative optimization scenarios. After the energy storage power station undergoes non-quantitative optimization, an adjustment coefficient is introduced to express the expected reduction in failure rate brought about by this technical improvement. Then, after updating the parameters based on the adjustment coefficient, the first posterior failure occurrence count is adjusted according to the difference between the first and second prior failure occurrence counts before and after the parameter update. This achieves a forward-looking threshold reduction for high-probability risks while maintaining the historical trend of the prior distribution, shifting and tightening the threshold towards a lower failure rate. This makes the corrected failure count threshold closer to the actual failure frequency of the target energy storage power station, ensuring that the data acquisition system remains efficient and sensitive and can dynamically adjust based on changes in the external environment. This further improves the targeting and efficiency of data acquisition, providing strong support for the long-term stable operation of the energy storage power station.

[0015] In one possible implementation, the step of performing a first compensation on the basic acquisition level of each characteristic parameter of the target energy storage power station based on each of the target fault types and a preset parameter-fault correlation matrix, and determining the first acquisition level of each of the characteristic parameters, includes: Iterate through each of the feature parameters and determine the first compensation coefficient for each of the feature parameters based on the parameter-fault correlation matrix and the preset weights of each of the target fault types. During the traversal process, for any feature parameter, the correlation degree between the feature parameter and each target fault type is determined according to the parameter-fault correlation degree matrix. The corresponding correlation degrees are weighted and summed according to the preset weights of each target fault type to obtain the first compensation coefficient of the feature parameter. The first acquisition level of each feature parameter is determined by performing a first compensation on the basic acquisition level of each feature parameter based on each of the first compensation coefficients.

[0016] This application provides a method for first compensation of the basic acquisition level. By traversing each characteristic parameter, calculating the weighted sum of correlation degrees, and determining the first compensation coefficient, the first acquisition level of each characteristic parameter is finally determined. This application dynamically adjusts the acquisition level by quantitatively analyzing the correlation between each parameter and the fault, thereby optimizing the allocation of data resources, improving the targeting and efficiency of data acquisition, and providing strong support for the long-term stable operation of energy storage power stations.

[0017] In one possible implementation, when determining a second acquisition level of a feature parameter by performing a second compensation on the first acquisition level of each feature parameter based on the variation amplitude of each feature parameter within the corresponding preset time window, the step of performing a second compensation on the first acquisition level of each feature parameter based on the variation amplitude of each feature parameter within the corresponding preset time window to determine the second acquisition level of each feature parameter includes: Obtain several sampled values ​​of the feature parameter within the corresponding preset time window, and further determine the initial value, final value, and average value of the feature parameter within the preset time window; The rate of change index of the feature parameter is calculated based on the initial value, the final value, and the length of the preset time window; The fluctuation range index of the feature parameter is calculated based on the sampled values ​​and the average value. If the rate of change index is greater than the corresponding preset rate of change threshold or the fluctuation amplitude index is greater than the corresponding preset fluctuation amplitude threshold, then the first acquisition level of the feature parameter is compensated for a second time according to the dynamic compensation coefficient to determine the second acquisition level of the feature parameter. The dynamic compensation coefficient is dynamically generated based on a first difference and / or a second difference. The first difference is the difference between the rate of change index and the preset rate of change threshold, and the second difference is the difference between the fluctuation amplitude index and the preset fluctuation amplitude threshold.

[0018] This application further proposes a second compensation mechanism based on the variation amplitude of characteristic parameters. By calculating the rate of change and fluctuation amplitude indicators, and comparing them with preset thresholds, a second compensation is applied to the first acquisition level, thereby determining the final second acquisition level of the characteristic parameters. This application embodiment statistically analyzes data changes within a time window to calculate the rate of change and fluctuation amplitude indicators. This allows for the sensitive capture of instantaneous anomalies and trend changes in parameters, enabling targeted adjustments to the acquisition level. This prevents insufficient acquisition or data distortion due to the dynamic characteristics of parameters, improving the targeting and efficiency of data acquisition. Furthermore, this application embodiment incorporates a dynamic compensation mechanism, adjusting the compensation coefficient based on the difference between the indicator value and the threshold. The greater the degree of exceeding the threshold, the stronger the compensation, ensuring minimum compensation when the parameter just exceeds the threshold. This avoids drastic changes in acquisition frequency due to small fluctuations, maintaining system stability. Simultaneously, it ensures timely increases in acquisition frequency during periods of drastic parameter fluctuations, thereby guaranteeing data integrity and timeliness.

[0019] In one possible implementation, the lithium battery data dynamic acquisition method further includes: sequentially performing noise removal and missing point interpolation operations on each feature parameter data obtained through the real-time data acquisition to obtain each effective feature parameter data after data processing.

[0020] This application's embodiments incorporate post-processing steps such as noise removal and missing point interpolation after data acquisition, further improving the quality of the original data. By effectively eliminating environmental interference and filling in data gaps, the accuracy and usability of the data are significantly improved. Furthermore, this method can provide a cleaner and more complete dataset for subsequent data analysis and fault diagnosis, offering strong support for the long-term stable operation of energy storage power stations.

[0021] Furthermore, the lithium battery data dynamic acquisition method also includes: performing quality assessment on each of the effective feature parameter data according to preset evaluation rules to obtain corresponding data quality assessment results, wherein the quality assessment includes data consistency assessment and data integrity assessment.

[0022] This application further proposes a mechanism for quality assessment of valid characteristic parameter data, including data consistency assessment and integrity assessment. By systematically assessing the quality of the data, data problems can be identified and corrected in a timely manner, thereby ensuring that the collected data meets the preset standards and providing strong support for the long-term stable operation of energy storage power stations.

[0023] Secondly, embodiments of this application provide a lithium battery data dynamic acquisition system for an energy storage power station, including an acquisition module, a fault type determination module, a first compensation module, a second compensation module, an acquisition frequency determination module, and a data acquisition module; The acquisition module is used to acquire fault data of the target energy storage power station during the current data acquisition cycle at the end of the current data acquisition cycle. The fault type determination module is used to determine several target fault types based on the fault data and various fault frequency thresholds. The various fault frequency thresholds are calculated based on the operating data of the target energy storage power station in the previous data acquisition cycle. The first compensation module is used to perform a first compensation on the basic acquisition level of each characteristic parameter of the target energy storage power station according to each of the target fault types and the preset parameter-fault correlation matrix, and to determine the first acquisition level of each of the characteristic parameters. The second compensation module is used to perform a second compensation on the first acquisition level of each feature parameter based on the change range of each feature parameter within the corresponding preset time window, and to determine the second acquisition level of each feature parameter. The acquisition frequency determination module is used to determine the acquisition frequency of each feature parameter in the next data acquisition cycle based on each of the second acquisition levels. The data acquisition module is used to collect data in real time on each of the characteristic parameters of the target energy storage power station according to the acquisition frequency in the next data acquisition cycle. Attached Figure Description

[0024] Figure 1 A flowchart illustrating a method for dynamically acquiring lithium battery data in an energy storage power station, as provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a lithium battery dynamic data acquisition system for an energy storage power station, provided in an embodiment of this application. Detailed Implementation

[0025] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0026] It should be noted that the step numbers in this document are only for the convenience of explaining the specific embodiments and are not intended to limit the order in which the steps are performed. In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of that feature.

[0027] Example 1: like Figure 1 As shown, Embodiment 1 provides a method for dynamically acquiring lithium battery data in an energy storage power station, including steps S1-S6: Step S1: At the end of the current data acquisition cycle, acquire the fault data of the target energy storage power station in the current data acquisition cycle; Step S2: Based on the fault data and each fault frequency threshold, determine several target fault types. Each fault frequency threshold is calculated based on the operating data of the target energy storage power station in the previous data collection cycle. Step S3: Based on each of the target fault types and the preset parameter-fault correlation matrix, perform a first compensation on the basic acquisition level of each characteristic parameter of the target energy storage power station to determine the first acquisition level of each characteristic parameter. Step S4: Based on the variation range of each feature parameter within the corresponding preset time window, perform a second compensation on the first acquisition level of each feature parameter to determine the second acquisition level of each feature parameter. Step S5: Determine the acquisition frequency of each feature parameter in the next data acquisition cycle based on each of the second acquisition levels. Step S6: In the next data acquisition cycle, real-time data acquisition is performed on each of the characteristic parameters of the target energy storage power station according to the acquisition frequency.

[0028] This application provides a method for dynamically acquiring lithium battery data in an energy storage power station. First, at the end of the current data acquisition cycle, fault data of the target energy storage power station is acquired. Then, based on historical operating data, Bayesian inference is used to set thresholds for the occurrence frequency of various faults, thereby identifying high-frequency target fault types. At the end of each cycle, the actual fault data is compared with the predicted data from Bayesian conjugate inference. Based on the comparison results, the acquisition level and frequency of characteristic parameters are dynamically adjusted. Furthermore, this application uses a preset parameter-fault correlation matrix to initially compensate the basic acquisition level, forming a first acquisition level. Then, a second compensation is performed based on the variation amplitude of characteristic parameters within a preset time window, forming a second acquisition level. Finally, the acquisition frequency for the next cycle is determined based on the second acquisition level, and real-time data acquisition is performed. Therefore, this embodiment possesses a high degree of adaptability and intelligence, enabling dynamic adjustment of the acquisition strategy according to the actual operating status of the energy storage power station. This more effectively captures potential fault information, significantly improving the targeting and efficiency of data acquisition, overcoming the resource waste or key data omission problems common in traditional fixed-frequency acquisition modes, and providing strong support for the long-term stable operation of energy storage power stations.

[0029] Furthermore, in step S2, determining several target fault types based on the fault data and various fault frequency thresholds includes: The number of occurrences for each fault type is determined based on the fault data. The occurrence count of each fault is compared with the corresponding fault count threshold. Fault types with a fault occurrence count greater than the fault count threshold are identified as the target fault types, thereby determining several target fault types.

[0030] This application further clarifies the mechanism for determining target fault types. Specifically, it includes counting the occurrence frequency of each fault type and comparing it with a preset fault frequency threshold to filter out target fault types that require special attention. Through quantitative analysis, it achieves objectivity and accuracy in fault identification, effectively reducing errors that may be introduced by subjective human judgment. By comparing the number of fault occurrences with the corresponding thresholds, the system can quickly identify frequently occurring fault types and subsequently adjust the collection frequency of various characteristic parameters to improve the targeting and efficiency of data collection.

[0031] In a preferred embodiment, fault data within the current data acquisition period is obtained from the lithium battery fault database of the energy storage power station to determine fault types exceeding a preset fault count threshold. Fault types may include thermal runaway, capacity decay, and battery short circuit. It should be noted that the fault data in the fault database includes fault type, fault occurrence time, battery number, and operating environment parameters. For example, the fault type may be explicitly recorded as: thermal runaway, excessively rapid capacity decay, short circuit, etc., and the operating environment parameters may include temperature, humidity, etc. Furthermore, based on the fault data, the same fault event recorded multiple times is considered as one fault of the same event. For example, multiple records of the same battery with the same fault type within a short period (e.g., within a few minutes) can be considered as the same fault event. This method removes duplicate data records of the same fault event. Then, the statistical count of each fault type is compared with the corresponding preset fault count threshold, and the fault types exceeding the preset fault count threshold are selected as target fault types.

[0032] For failure frequency thresholds, they can typically be set based on quantifiable factors such as the scale and operating history of the energy storage power station. For example, theoretically, larger energy storage power stations with longer operating histories are more prone to failures, which aligns with actual operating conditions; therefore, the failure frequency threshold can be set higher. Under this principle, if a failure type still exceeds the failure frequency threshold, it indicates that the failure rate of that type in the energy storage power station is higher than expected and requires closer attention. Alternatively, the threshold can be dynamically adjusted based on non-quantifiable factors. For instance, if a new battery management system is adopted, effectively reducing the occurrence of failures, then theoretically, failures should be less likely. If failures are frequent, it indicates a discrepancy with actual operating conditions; therefore, the failure frequency threshold can be appropriately lowered. Under this principle, if a failure type still exceeds the failure frequency threshold, it indicates that the failure rate of that type in the energy storage power station is also higher than expected and requires closer attention.

[0033] Furthermore, this application embodiment employs a Bayesian estimation method to dynamically estimate the failure rate based on the operating data of each cycle, thereby determining the corresponding failure count threshold for the next cycle.

[0034] In one possible implementation, when calculating the fault count threshold for any fault type based on the operating data of the target energy storage power station in the previous data acquisition cycle, the calculation of the respective fault count thresholds based on the operating data of the target energy storage power station in the previous data acquisition cycle includes: Obtain the operating data of the target energy storage power station in the previous data acquisition cycle. The operating data includes the number of faults of the fault type and the lithium battery operating energy data. If there is no prior distribution of the failure rate of the fault type under unit operating energy in the current data acquisition period, then the prior modeling of the failure rate of the fault type under unit operating energy is performed based on the historical data of the target energy storage power station to obtain the prior distribution. Based on the operational data and the prior distribution, Bayesian conjugate inference is performed to determine the posterior distribution in the Bayesian estimation method; Based on the posterior distribution, the posterior number of fault occurrences of the fault type in the current data collection period is predicted using a Bayesian estimation method, and the posterior number of fault occurrences is used as the fault count threshold for the fault type. The posterior distribution is used as the prior distribution of the failure rate of the fault type under unit operating energy in the next data acquisition cycle.

[0035] This application provides a method for determining a fault count threshold. At the end of the current data acquisition cycle, the prior distribution of the current data acquisition cycle is obtained. Then, combined with operational data, Bayesian conjugate inference is used to obtain the posterior distribution of the current data acquisition cycle. Finally, the number of faults in the current data acquisition cycle is predicted based on the posterior distribution, and this predicted value is used as the fault count threshold for the current data acquisition cycle. This application introduces Bayesian inference theory from statistics, making the setting of the fault threshold more scientific and reasonable, and effectively addressing uncertainties in the operating environment. Furthermore, this embodiment also uses the posterior distribution of each cycle as the prior distribution of the next data acquisition cycle, realizing iterative updates of Bayesian parameters. This enables Bayesian conjugate inference to have self-learning and continuous optimization capabilities, dynamically adjusting the prediction model based on continuously accumulated operational data, thereby improving the adaptive capability and robustness of the data acquisition system and providing solid data support for the long-term reliable operation of energy storage power stations. Furthermore, this application embodiment also considers the scenario of performing Bayesian estimation for the first time. In this case, the prior distribution cannot be obtained directly. Therefore, based on the historical data of the target energy storage power station, the initial prior distribution is obtained through prior modeling, which ensures the smooth progress of Bayesian parameter iteration and update.

[0036] Furthermore, based on historical data from the target energy storage power station, prior modeling is performed on the failure rate of the fault type per unit operating energy to obtain the prior distribution, including: Based on the length of the data collection cycle, the historical operating data is divided into several sub-historical operating data. A priori model is performed on the failure rate of the aforementioned failure type under unit operating energy to construct the initial prior distribution in the Bayesian estimation method; Based on the historical data of each sub-sub-distribution, the method of moments is used to estimate the parameters of the initial prior distribution, determine the parameter values ​​of the initial prior distribution, and then obtain the prior distribution.

[0037] This application provides a modeling process for a prior distribution, which involves dividing historical operational data into several subsets, constructing an initial prior distribution, and estimating parameters using the method of moments to obtain the final prior distribution. This embodiment, through refined processing of historical data, ensures the accuracy and representativeness of the prior distribution, laying a reliable foundation for subsequent Bayesian inference. This makes the calculation of the fault threshold more robust and reliable, and improves the targeting and efficiency of subsequent data collection.

[0038] In a preferred embodiment, taking a certain type of fault as an example, when initially implementing the Bayesian estimation method, the failure rate of this type of fault is first modeled a priori, and the occurrence rate λ of this type of fault under unit operating energy is modeled as a Gamma distribution: Where λ represents the failure rate per unit of operating energy. and These are the shape parameter and the scale parameter, respectively. As the value increases, the distribution tends to be symmetrical and the kurtosis decreases, making it closer to a normal distribution; with... As the value increases, the variance of the distribution increases, and the shape of the distribution becomes wider and flatter.

[0039] Calculate using the method of moment estimation and First, based on the number of times n of this type of fault occurs. i and the total operating energy E during this period i Calculate the historical failure rate λ for each time period. i : We obtain a sample set {λ1, λ2, ..., λ} with respect to λ. N Then, parameter estimation is performed based on the sample set.

[0040] The sample moments of the gamma distribution are: in, The number of samples; The mean of the sample; This represents the sample variance.

[0041] The theoretical moments of the gamma distribution are: in, This is the theoretical mean; Let the theoretical variance be the sample moments of the gamma distribution. Then, let the sample moments of the gamma distribution equal the theoretical moments, and solve for the parameter estimates. and : Then, based on the prior distribution, the failure rate distribution parameters are updated posteriorly. In the current data acquisition period, the number of times this type of failure occurs is n, and the total operating volume of the power station in this period is: Where T is the duration of the preset period. Based on Bayesian conjugate inference, the posterior distribution of the failure rate can be obtained as follows: It should be noted that the posterior distribution of the current data collection period will be used as the prior distribution of the next data collection period, thereby realizing the iterative update of the distribution parameters.

[0042] Based on the parameter update completed using the posterior distribution, the fault threshold for the next data acquisition cycle is predicted. Let the operational power in the next cycle be P', and the preset cycle be T', then the total operational power in the next cycle is: According to the Bayesian prediction formula, the number of such failures X in the next period follows a negative binomial distribution: The negative binomial distribution describes the probability distribution of the number of failures X over multiple future periods when the failure rate parameter λ is uncertain. The q% quantile of this distribution is used as the failure count threshold. For example, when q=90, it means that after combining historical data and the latest data, the actual number of failures in the next data collection cycle exceeds the predicted failure number threshold. The probability of this is only 10%, making it a low-probability event. Therefore, in actual monitoring, if the number of faults exceeds the threshold, there is a high degree of confidence that the occurrence rate of this type of fault in the power plant has significantly increased, and it should be identified as the target fault type.

[0043] Furthermore, the lithium battery data dynamic acquisition method also includes updating the posterior distribution and the fault count threshold if the target energy storage power station has undergone non-quantitative optimization in the current data acquisition cycle, specifically as follows: Based on the prior distribution, the number of first prior fault occurrences of the fault type in the current data acquisition period is predicted using a Bayesian estimation method. The parameters of the prior distribution and the posterior distribution are updated by a preset adjustment coefficient to obtain the corrected prior distribution and the corrected posterior distribution. Based on the modified prior distribution, the number of second prior fault occurrences of the fault type in the current data acquisition period is predicted using a Bayesian estimation method. Based on the modified posterior distribution, the first posterior fault occurrence count of the fault type in the current data acquisition period is predicted using a Bayesian estimation method. Based on the difference between the first prior fault occurrence count and the second prior fault occurrence count, the first posterior fault occurrence count is adjusted to obtain the second posterior fault occurrence count, and the second posterior fault occurrence count is used as the fault occurrence count threshold for the fault type. The modified posterior distribution is used as the prior distribution of the failure rate of the fault type under unit operating energy in the next data acquisition cycle.

[0044] This application further considers a threshold update mechanism for non-quantitative optimization scenarios. After the energy storage power station undergoes non-quantitative optimization, an adjustment coefficient is introduced to express the expected reduction in failure rate brought about by this technical improvement. Then, after updating the parameters based on the adjustment coefficient, the first posterior failure occurrence count is adjusted according to the difference between the first and second prior failure occurrence counts before and after the parameter update. This achieves a forward-looking threshold reduction for high-probability risks while maintaining the historical trend of the prior distribution, shifting and tightening the threshold towards a lower failure rate. This makes the corrected failure count threshold closer to the actual failure frequency of the target energy storage power station, ensuring that the data acquisition system remains efficient and sensitive and can dynamically adjust based on changes in the external environment. This further improves the targeting and efficiency of data acquisition, providing strong support for the long-term stable operation of the energy storage power station.

[0045] In a preferred embodiment, after the energy storage power station undergoes non-quantitative optimization (such as battery management system updates, thermal control system optimizations, etc.), an adjustment coefficient is introduced. This is used to express the expected reduction in the failure rate parameter λ resulting from this technological improvement. Subsequently, the original prior distribution parameters are simultaneously scaled to: This allows for the proactive lowering of thresholds for reducing high-probability risks while maintaining the historical trend of the prior distribution, thus shifting and tightening the overall threshold towards lower failure rates. It should be noted that in actual operation, there may be multiple versions of threshold settings, for example: The threshold for the number of failures calculated under the initial prior conditions is X. AThe first prior fault occurred; the threshold obtained after non-quantization optimization based on scaling correction is X. B This refers to the number of times the second prior fault occurs; after further data collection, a threshold of X is obtained through Bayesian inference. C This refers to the number of first posterior fault occurrences. At this point, the fault occurrence threshold X is calculated. A With the fault count threshold X B The difference between them, and then, based on the difference, the failure number threshold X. C Compensation is performed with a compensation coefficient greater than 0 and less than 1, so that the number of second posterior fault occurrences after compensation is less than the fault occurrence threshold X. C The larger the difference, the greater the improvement to the energy storage power station. Therefore, the number of first posterior fault occurrences needs to be reduced; that is, the larger the difference, the smaller the compensation coefficient. Finally, the number of second posterior fault occurrences is used as the standard to determine the preset fault count threshold for the energy storage power station.

[0046] In one possible implementation, in step S3, the first compensation is performed on the basic acquisition level of each characteristic parameter of the target energy storage power station based on each of the target fault types and a preset parameter-fault correlation matrix, to determine the first acquisition level of each of the characteristic parameters, including: Iterate through each of the feature parameters and determine the first compensation coefficient for each of the feature parameters based on the parameter-fault correlation matrix and the preset weights of each of the target fault types. During the traversal process, for any feature parameter, the correlation degree between the feature parameter and each target fault type is determined according to the parameter-fault correlation degree matrix. The corresponding correlation degrees are weighted and summed according to the preset weights of each target fault type to obtain the first compensation coefficient of the feature parameter. The first acquisition level of each feature parameter is determined by performing a first compensation on the basic acquisition level of each feature parameter based on each of the first compensation coefficients.

[0047] This application provides a method for first compensation of the basic acquisition level. By traversing each characteristic parameter, calculating the weighted sum of correlation degrees, and determining the first compensation coefficient, the first acquisition level of each characteristic parameter is finally determined. This application dynamically adjusts the acquisition level by quantitatively analyzing the correlation between each parameter and the fault, thereby optimizing the allocation of data resources, improving the targeting and efficiency of data acquisition, and providing strong support for the long-term stable operation of energy storage power stations.

[0048] In a preferred embodiment, during the first compensation process, a characteristic parameter acquisition level table for the energy storage power station is first obtained. This table contains characteristic parameters and their corresponding basic acquisition levels. For example, characteristic parameters include battery voltage, current, and temperature, which reflect the battery's operating state and performance changes. The basic acquisition level directly determines the initial sampling frequency of each parameter; the higher the level value, the higher the corresponding data acquisition frequency. According to industry practice and research data, there are significant differences in the basic acquisition level settings for different parameters. Parameters that change rapidly, such as voltage and current, usually require a higher basic level to meet the needs of dynamic response and rapid fault diagnosis. Temperature, as a key indicator for thermal safety management, generally uses a medium acquisition level. Parameters that change slowly, such as internal resistance and state of equilibrium (SOH), can be set to a relatively lower acquisition level to optimize system resource allocation.

[0049] Then, a search is performed in the feature parameter acquisition level table to obtain the basic acquisition level of each feature parameter in the feature parameter set. When multiple fault types occur simultaneously, the feature parameters that are related to them are determined, and the basic acquisition level of each feature parameter is compensated according to the correlation between the feature parameter and the target fault type. Specifically, suppose the system defines 𝑚 fault types, constituting a fault set. and a set of n feature parameters In the first compensation mechanism, to improve the monitoring efficiency of key parameters, the system enhances the acquisition level by quantifying the correlation between characteristic parameters and various faults. First, a parameter-fault correlation matrix is ​​constructed, where each element... ,in, Indicates parameters With fault The correlation between them is calculated, with higher values ​​indicating stronger correlations. Additionally, weights are assigned to each fault type. ,satisfy This reflects the importance of different faults.

[0050] Based on this, the first compensation coefficient for the i-th feature parameter is defined as follows: In this formula, the constant "1" is used to maintain the basic acquisition level, and the summation term characterizes the comprehensive correlation strength between the parameters and all faults. The formula for calculating the compensated first acquisition level is: This compensation mechanism ensures that parameters strongly correlated with multiple faults are acquired at a higher level, significantly improving the system's ability to detect composite and associated faults, and providing effective data support for the safe operation of energy storage power stations.

[0051] In one possible implementation, in step S4, when a second acquisition level of a feature parameter is determined by performing a second compensation on the first acquisition level of the feature parameter based on the change magnitude of each feature parameter within the corresponding preset time window, the step of performing a second compensation on the first acquisition level of each feature parameter based on the change magnitude of each feature parameter within the corresponding preset time window, and determining the second acquisition level of each feature parameter, includes: Obtain several sampled values ​​of the feature parameter within the corresponding preset time window, and further determine the initial value, final value, and average value of the feature parameter within the preset time window; The rate of change index of the feature parameter is calculated based on the initial value, the final value, and the length of the preset time window; The fluctuation range index of the feature parameter is calculated based on the sampled values ​​and the average value. If the rate of change index is greater than the corresponding preset rate of change threshold or the fluctuation amplitude index is greater than the corresponding preset fluctuation amplitude threshold, then the first acquisition level of the feature parameter is compensated for a second time according to the dynamic compensation coefficient to determine the second acquisition level of the feature parameter. The dynamic compensation coefficient is dynamically generated based on a first difference and / or a second difference. The first difference is the difference between the rate of change index and the preset rate of change threshold, and the second difference is the difference between the fluctuation amplitude index and the preset fluctuation amplitude threshold.

[0052] This application further proposes a second compensation mechanism based on the variation amplitude of characteristic parameters. By calculating the rate of change and fluctuation amplitude indicators, and comparing them with preset thresholds, a second compensation is applied to the first acquisition level, thereby determining the final second acquisition level of the characteristic parameters. This application embodiment statistically analyzes data changes within a time window to calculate the rate of change and fluctuation amplitude indicators. This allows for the sensitive capture of instantaneous anomalies and trend changes in parameters, enabling targeted adjustments to the acquisition level. This prevents insufficient acquisition or data distortion due to the dynamic characteristics of parameters, improving the targeting and efficiency of data acquisition. Furthermore, this application embodiment incorporates a dynamic compensation mechanism, adjusting the compensation coefficient based on the difference between the indicator value and the threshold. The greater the degree of exceeding the threshold, the stronger the compensation, ensuring minimum compensation when the parameter just exceeds the threshold. This avoids drastic changes in acquisition frequency due to small fluctuations, maintaining system stability. Simultaneously, it ensures timely increases in acquisition frequency during periods of drastic parameter fluctuations, thereby guaranteeing data integrity and timeliness.

[0053] In a preferred embodiment, by monitoring the instantaneous rate of change and fluctuation amplitude of characteristic parameters, a second acquisition level compensation is automatically triggered to ensure the timeliness and accuracy of data acquisition under abnormal operating conditions. Specifically, the time window is set as follows: Calculate parameters within this window. Rate of change indicators and fluctuation amplitude indicators: 1) Rate of change index To capture transient changes in parameters and identify sudden anomalies, the value of Δt should be matched with the characteristics of the parameter itself and its typical fault modes. For example, for voltage and current, Δt may need to be set smaller to capture rapid transients; for temperature and internal resistance, Δt can be appropriately increased.

[0054] 2) Volatility Indicator It is used to evaluate the stability of parameter fluctuations, detect persistent anomalies, and provide a basis for medium- and long-term performance evaluation and early fault diagnosis of the system.

[0055] When any indicator exceeds the threshold, dynamic compensation enhancement is triggered: In the formula, the threshold It can be a fixed value, or it can be dynamically adjusted based on historical battery data (such as the mean and standard deviation within a sliding time window) or current operating conditions (such as SOC, temperature, and charge / discharge rate). The adaptive adjustment method is as follows: in, It is the moving average of the historical rate of change / standard deviation; The standard deviation of the historical rate of change / standard deviation; This is the sensitivity coefficient.

[0056] The dynamic compensation coefficient γ is positively correlated with the extent to which the monitored value of the characteristic parameter exceeds the corresponding threshold; that is, the greater the degree of exceeding the threshold, the stronger the compensation. Its specific calculation uses the following nonlinear model: in, As a benchmark compensation coefficient, it can ensure that there is minimum compensation force when the parameter just exceeds the threshold, avoid drastic changes in the acquisition frequency due to small fluctuations, and maintain system stability; This is the proportional gain coefficient, used to adjust the amplification factor of the compensation intensity. Its value is positively correlated with the importance of the parameter and the level of safety risk. Critical parameters (such as voltage and current) are assigned higher values, while less important parameters are assigned lower values. This is the actual value; Threshold; Nonlinear adjustment index This ensures that there is minimum compensation when the parameter just exceeds the threshold, avoiding drastic changes in the acquisition frequency due to small fluctuations and maintaining system stability.

[0057] In the process of compensating for the acquisition level, the acquisition level can be converted into an acquisition score. The compensation coefficient is then multiplied by the acquisition score to obtain the compensated acquisition score. The compensated acquisition score is then converted into a compensated acquisition level. Finally, the actual acquisition level of each feature parameter is the sum of the basic acquisition level and the compensated acquisition level. A mapping table between acquisition levels and acquisition scores can be set to facilitate the conversion between the two. The compensation coefficient is greater than 1, and its value can be set by those skilled in the art.

[0058] In one possible implementation, the lithium battery data dynamic acquisition method further includes: sequentially performing noise removal and missing point interpolation operations on each feature parameter data obtained through the real-time data acquisition to obtain each effective feature parameter data after data processing.

[0059] This application's embodiments incorporate post-processing steps such as noise removal and missing point interpolation after data acquisition, further improving the quality of the original data. By effectively eliminating environmental interference and filling in data gaps, the accuracy and usability of the data are significantly improved. Furthermore, this method can provide a cleaner and more complete dataset for subsequent data analysis and fault diagnosis, offering strong support for the long-term stable operation of energy storage power stations.

[0060] In a preferred embodiment, noise interference is removed from the feature parameter data. For example, a Kalman filter or low-pass filter algorithm is used. Missing data points are extracted from the feature parameter data. Since each feature parameter data point has a collection time point, if the feature parameter data for that collection time point is empty, it indicates that the feature parameter data for that collection time is missing. Interpolation algorithms are used to compensate for the missing data points. For example, linear interpolation or spline interpolation is used.

[0061] Furthermore, the lithium battery data dynamic acquisition method also includes: performing quality assessment on each of the effective feature parameter data according to preset evaluation rules to obtain corresponding data quality assessment results, wherein the quality assessment includes data consistency assessment and data integrity assessment.

[0062] This application further proposes a mechanism for quality assessment of valid characteristic parameter data, including data consistency assessment and integrity assessment. By systematically assessing the quality of the data, data problems can be identified and corrected in a timely manner, thereby ensuring that the collected data meets the preset standards and providing strong support for the long-term stable operation of energy storage power stations.

[0063] In a preferred embodiment, the valid feature parameter data is evaluated for quality according to preset evaluation rules. For example, the consistency and completeness of the data can be evaluated. For instance, completeness mainly checks for missing data and assesses the rationality of the compensated data. If the number of missing data exceeds a preset threshold, the quality is poor. Alternatively, if the compensated feature parameter data exhibits fixed values ​​that do not conform to physical laws, this may mean that the compensation algorithm has not correctly recovered the missing data, also indicating poor quality. For example, consistency refers to the degree of conformity between different feature parameters or the same feature parameter under different acquisition devices or conditions. For lithium batteries, there is a certain relationship between current and voltage (according to Ohm's law, etc.). If, during charging, the theoretical current calculated based on voltage and internal resistance differs significantly from the actual acquired current (e.g., the theoretical current is 2A, but the actual acquired current exceeds 3A or is less than 1A), it indicates poor data consistency and thus poor quality. Finally, the quality evaluation results of the lithium battery feature parameter data for the energy storage power station are generated and fed back to the client.

[0064] Furthermore, common characteristic parameters that are simultaneously associated with two or more target fault types can be marked in the quality assessment results, thereby increasing the weight of common characteristic parameters in safety analysis.

[0065] In summary, the embodiments of this application have at least the following beneficial effects: 1. It enables the effective collection and filtering of battery data, providing valuable data support for the safety assessment of lithium batteries. This helps improve the safety, reliability, and economy of battery systems.

[0066] 2. By setting and updating the fault count threshold based on the operating data of each cycle and Bayesian conjugate inference, the fault count threshold can be dynamically changed based on the actual situation of the energy storage power station, rather than being set manually. This allows for timely and accurate screening of fault types with a higher probability of potential danger, thereby enabling the selection of the set of characteristic parameters that need to be focused on in the energy storage power station.

[0067] 3. Determine the set of characteristic parameters related to the selected fault types, and select several characteristic parameters corresponding to each target fault type. These characteristic parameters have a higher correlation with the failure rate of lithium batteries and should be given priority. Furthermore, the collection frequency for parameters with rapid operational changes should be higher than that for parameters with slow operational changes, and these should also be given priority. Based on this, the collection levels for different characteristic parameters can be determined, and the characteristic parameters that require key attention in this energy storage power station can be selected.

[0068] It should be noted that allowing those characteristic parameters that are more critical to reflecting fault conditions and whose dynamic changes are more obvious to be focused on at a more appropriate acquisition level improves the effectiveness and relevance of the acquired data. Adjusting the acquisition frequency based on the acquisition level of each characteristic parameter after the update ensures that key parameter values ​​are acquired at appropriate time intervals. This avoids both excessively frequent acquisitions that waste resources and the loss of important data changes due to excessively long acquisition intervals, thus helping to grasp the status of lithium batteries more efficiently and accurately.

[0069] 4. By removing noise interference and extracting and compensating for missing data points in the feature parameter data, the continuity and integrity of the data are ensured, making the data foundation for subsequent analysis more reliable. Furthermore, the quality of the effective feature parameter data is assessed from multiple aspects such as consistency and completeness, enabling accurate judgment of the data quality.

[0070] Example 2: like Figure 2 As shown, Embodiment 2 provides a lithium battery data dynamic acquisition system for an energy storage power station, including an acquisition module 10, a fault type determination module 20, a first compensation module 30, a second compensation module 40, an acquisition frequency determination module 50, and a data acquisition module 60. The acquisition module 10 is used to acquire fault data of the target energy storage power station in the current data acquisition cycle at the end of the current data acquisition cycle. The fault type determination module 20 is used to determine several target fault types based on the fault data and various fault frequency thresholds. The various fault frequency thresholds are calculated based on the operating data of the target energy storage power station in the previous data acquisition cycle. The first compensation module 30 is used to perform a first compensation on the basic acquisition level of each characteristic parameter of the target energy storage power station according to each of the target fault types and the preset parameter-fault correlation matrix, and to determine the first acquisition level of each of the characteristic parameters. The second compensation module 40 is used to perform a second compensation on the first acquisition level of each feature parameter according to the change range of each feature parameter within the corresponding preset time window, and to determine the second acquisition level of each feature parameter; The acquisition frequency determination module 50 is used to determine the acquisition frequency of each feature parameter in the next data acquisition cycle according to each of the second acquisition levels. The data acquisition module 60 is used to collect data in real time on each of the characteristic parameters of the target energy storage power station according to each acquisition frequency in the next data acquisition cycle.

[0071] Furthermore, the fault type determination module 20 determines several target fault types based on the fault data and various fault frequency thresholds, including: The number of occurrences for each fault type is determined based on the fault data. The occurrence count of each fault is compared with the corresponding fault count threshold. Fault types with a fault occurrence count greater than the fault count threshold are identified as the target fault types, thereby determining several target fault types.

[0072] In one possible implementation, when calculating the fault count threshold for any fault type based on the operating data of the target energy storage power station in the previous data acquisition cycle, the calculation of the respective fault count thresholds based on the operating data of the target energy storage power station in the previous data acquisition cycle includes: Obtain the operating data of the target energy storage power station in the previous data acquisition cycle. The operating data includes the number of faults of the fault type and the lithium battery operating energy data. If there is no prior distribution of the failure rate of the fault type under unit operating energy in the current data acquisition period, then the prior modeling of the failure rate of the fault type under unit operating energy is performed based on the historical data of the target energy storage power station to obtain the prior distribution. Based on the operational data and the prior distribution, Bayesian conjugate inference is performed to determine the posterior distribution in the Bayesian estimation method; Based on the posterior distribution, the posterior number of fault occurrences of the fault type in the current data collection period is predicted using a Bayesian estimation method, and the posterior number of fault occurrences is used as the fault count threshold for the fault type. The posterior distribution is used as the prior distribution of the failure rate of the fault type under unit operating energy in the next data acquisition cycle.

[0073] Furthermore, based on historical data from the target energy storage power station, prior modeling is performed on the failure rate of the fault type per unit operating energy to obtain the prior distribution, including: Based on the length of the data collection cycle, the historical operating data is divided into several sub-historical operating data. A priori model is performed on the failure rate of the aforementioned failure type under unit operating energy to construct the initial prior distribution in the Bayesian estimation method; Based on the historical data of each sub-sub-distribution, the method of moments is used to estimate the parameters of the initial prior distribution, determine the parameter values ​​of the initial prior distribution, and then obtain the prior distribution.

[0074] Furthermore, the lithium battery data dynamic acquisition method also includes updating the posterior distribution and the fault count threshold if the target energy storage power station has undergone non-quantitative optimization in the current data acquisition cycle, specifically as follows: Based on the prior distribution, the number of first prior fault occurrences of the fault type in the current data acquisition period is predicted using a Bayesian estimation method. The parameters of the prior distribution and the posterior distribution are updated by a preset adjustment coefficient to obtain the corrected prior distribution and the corrected posterior distribution. Based on the modified prior distribution, the number of second prior fault occurrences of the fault type in the current data acquisition period is predicted using a Bayesian estimation method. Based on the modified posterior distribution, the first posterior fault occurrence count of the fault type in the current data acquisition period is predicted using a Bayesian estimation method. Based on the difference between the first prior fault occurrence count and the second prior fault occurrence count, the first posterior fault occurrence count is adjusted to obtain the second posterior fault occurrence count, and the second posterior fault occurrence count is used as the fault occurrence count threshold for the fault type. The modified posterior distribution is used as the prior distribution of the failure rate of the fault type under unit operating energy in the next data acquisition cycle.

[0075] In one possible implementation, the first compensation module 30 performs a first compensation on the basic acquisition level of each characteristic parameter of the target energy storage power station based on each of the target fault types and a preset parameter-fault correlation matrix, determining the first acquisition level of each of the characteristic parameters, including: Iterate through each of the feature parameters and determine the first compensation coefficient for each of the feature parameters based on the parameter-fault correlation matrix and the preset weights of each of the target fault types. During the traversal process, for any feature parameter, the correlation degree between the feature parameter and each target fault type is determined according to the parameter-fault correlation degree matrix. The corresponding correlation degrees are weighted and summed according to the preset weights of each target fault type to obtain the first compensation coefficient of the feature parameter. The first acquisition level of each feature parameter is determined by performing a first compensation on the basic acquisition level of each feature parameter based on each of the first compensation coefficients.

[0076] In one possible implementation, when a second acquisition level of a feature parameter is determined by performing a second compensation on the first acquisition level of the feature parameter based on the change magnitude of each feature parameter within the corresponding preset time window, the second compensation module 40 performs a second compensation on the first acquisition level of each feature parameter based on the change magnitude of each feature parameter within the corresponding preset time window, and determines the second acquisition level of each feature parameter, including: Obtain several sampled values ​​of the feature parameter within the corresponding preset time window, and further determine the initial value, final value, and average value of the feature parameter within the preset time window; The rate of change index of the feature parameter is calculated based on the initial value, the final value, and the length of the preset time window; The fluctuation range index of the feature parameter is calculated based on the sampled values ​​and the average value. If the rate of change index is greater than the corresponding preset rate of change threshold or the fluctuation amplitude index is greater than the corresponding preset fluctuation amplitude threshold, then the first acquisition level of the feature parameter is compensated for a second time according to the dynamic compensation coefficient to determine the second acquisition level of the feature parameter. The dynamic compensation coefficient is dynamically generated based on a first difference and / or a second difference. The first difference is the difference between the rate of change index and the preset rate of change threshold, and the second difference is the difference between the fluctuation amplitude index and the preset fluctuation amplitude threshold.

[0077] In one possible implementation, the lithium battery data dynamic acquisition system further includes a data processing module, which is used to sequentially perform noise removal and missing point interpolation operations on each feature parameter data obtained through the real-time data acquisition to obtain each effective feature parameter data after data processing.

[0078] In one possible implementation, the lithium battery data dynamic acquisition system further includes a data evaluation module, which is used to perform quality evaluation on each of the effective feature parameter data according to preset evaluation rules to obtain the corresponding data quality evaluation results. The quality evaluation includes data consistency evaluation and data integrity evaluation.

[0079] This application provides a dynamic data acquisition system for lithium batteries in an energy storage power station. First, at the end of the current data acquisition cycle, fault data of the target energy storage power station is acquired. Then, based on historical operating data, Bayesian inference is used to set thresholds for the occurrence frequency of various faults, thereby identifying high-frequency target fault types. At the end of each cycle, the actual fault data is compared with the predicted data from Bayesian conjugate inference. Based on the comparison results, the acquisition level and frequency of characteristic parameters are dynamically adjusted. Furthermore, this application uses a preset parameter-fault correlation matrix to initially compensate the basic acquisition level, forming a first acquisition level. Then, a second compensation is performed based on the variation amplitude of characteristic parameters within a preset time window, forming a second acquisition level. Finally, the acquisition frequency for the next cycle is determined based on the second acquisition level, and real-time data acquisition is performed. Therefore, this embodiment possesses a high degree of adaptability and intelligence, dynamically adjusting the acquisition strategy according to the actual operating status of the energy storage power station, thereby more effectively capturing potential fault information, significantly improving the targeting and efficiency of data acquisition, overcoming the resource waste or key data omission problems common in traditional fixed-frequency acquisition modes, and providing strong support for the long-term stable operation of energy storage power stations.

[0080] For a more detailed explanation of the working principle and procedures of this embodiment, please refer to the relevant description in Embodiment 1.

[0081] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above descriptions are merely specific embodiments of this application and are not intended to limit the scope of protection of this application. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application for those skilled in the art.

Claims

1. A lithium battery data dynamic acquisition method for energy storage power stations, characterized in that, The method comprises the following steps: At the end of the current data collection period, obtaining fault data of the target energy storage power station in the current data collection period; According to the fault data and each fault frequency threshold, determining a plurality of target fault types, wherein the fault frequency threshold is calculated according to the operation data of the target energy storage power station in the last data collection period; According to each target fault type and a preset parameter-fault correlation matrix, performing first compensation on the basic collection level of each characteristic parameter of the target energy storage power station to determine the first collection level of each characteristic parameter; According to the variation amplitude of each characteristic parameter within a corresponding preset time window, performing second compensation on the first collection level of each characteristic parameter to determine the second collection level of each characteristic parameter; According to each second collection level, determining the collection frequency of each characteristic parameter in the next data collection period; In the next data collection period, collecting real-time data of each characteristic parameter of the target energy storage power station according to the collection frequency.

2. The method of claim 1, wherein the method further comprises: The method comprises the following steps: According to the fault data, determining the fault occurrence frequency of each fault type; Comparing each fault occurrence frequency with a corresponding fault frequency threshold, and determining the fault type with a fault occurrence frequency greater than the fault frequency threshold as the target fault type, thereby determining a plurality of target fault types.

3. The method of claim 1, wherein the method further comprises: When calculating the fault frequency threshold of any fault type according to the operation data of the target energy storage power station in the last data collection period, the method comprises the following steps: Obtaining the operation data of the target energy storage power station in the last data collection period, wherein the operation data comprises the fault occurrence frequency of the fault type and lithium battery operation energy data; If there is no prior distribution of the fault rate of the fault type under unit operation energy in the current data collection period, then according to the historical data of the target energy storage power station, the fault rate of the fault type under unit operation energy is modeled to obtain the prior distribution; According to the operation data and the prior distribution, performing Bayesian conjugate inference to determine the posterior distribution in the Bayesian estimation method; According to the posterior distribution, the posterior fault occurrence frequency of the fault type in the current data collection period is predicted by the Bayesian estimation method, and the posterior fault occurrence frequency is taken as the fault frequency threshold of the fault type; The posterior distribution is taken as the prior distribution of the fault rate of the fault type under unit operation energy in the next data collection period.

4. The method of claim 3, wherein the method further comprises: The method comprises the following steps: According to the period length of the data collection period, the historical operation data is divided into a plurality of sub-historical operation data; The fault rate of the fault type under unit operation energy is modeled to construct the initial prior distribution in the Bayesian estimation method; According to each of the sub-historical operation data, a method of matrix estimation is used to perform parameter estimation on the initial prior distribution, to determine each parameter value in the initial prior distribution, and to further obtain the prior distribution.

5. The method of claim 3, wherein the method further comprises: The lithium battery data dynamic acquisition method further includes updating the posterior distribution and the fault frequency threshold if the target energy storage power station has performed non-quantitative optimization in the current data acquisition period, specifically: According to the prior distribution, a first prior fault occurrence frequency of the fault type in the current data acquisition period is predicted by a Bayesian estimation method; Parameters of the prior distribution and the posterior distribution are updated by a preset adjustment coefficient to obtain a corrected prior distribution and a corrected posterior distribution; According to the corrected prior distribution, a second prior fault occurrence frequency of the fault type in the current data acquisition period is predicted by a Bayesian estimation method; According to the corrected posterior distribution, a first posterior fault occurrence frequency of the fault type in the current data acquisition period is predicted by a Bayesian estimation method; According to the difference between the first prior fault occurrence frequency and the second prior fault occurrence frequency, the first posterior fault occurrence frequency is adjusted to obtain a second posterior fault occurrence frequency, and the second posterior fault occurrence frequency is taken as the fault frequency threshold of the fault type; The corrected posterior distribution is taken as the prior distribution of the fault rate of the fault type under unit operating energy in the next data acquisition period.

6. The method of claim 1, wherein the method further comprises: The first compensation of the basic acquisition level of each feature parameter of the target energy storage power station according to each of the target fault types and a preset parameter-fault correlation matrix includes: Each of the feature parameters is traversed, and a first compensation coefficient of each of the feature parameters is determined according to the parameter-fault correlation matrix and a preset weight of each of the target fault types; In the traversal process, for any of the feature parameters, the correlation degree of the feature parameter and each of the target fault types is determined according to the parameter-fault correlation matrix, and the first compensation coefficient of the feature parameter is obtained by weighted summing of each of the correlation degrees corresponding to each of the target fault types according to the preset weight of each of the target fault types; The first compensation of the basic acquisition level of each of the feature parameters according to each of the first compensation coefficients determines the first acquisition level of each of the feature parameters.

7. The method for dynamic acquisition of lithium battery data in an energy storage power station as described in claim 1, characterized in that, When the first acquisition level of any of the feature parameters is secondly compensated according to the variation amplitude of the feature parameter in the corresponding preset time window to determine the second acquisition level of the feature parameter, the second compensation of the first acquisition level of each of the feature parameters according to the variation amplitude of each of the feature parameters in the corresponding preset time window to determine the second acquisition level of each of the feature parameters includes: A plurality of sampling values of the feature parameter in the corresponding preset time window are obtained, and an initial value, a final value and an average value of the feature parameter in the preset time window are further determined; According to the initial value, the final value and the length of the preset time window, a change rate index of the characteristic parameter is calculated; According to the several sampling values and the average value, a fluctuation amplitude index of the characteristic parameter is calculated; If the change rate index is greater than a corresponding preset change rate threshold or the fluctuation amplitude index is greater than a corresponding preset fluctuation amplitude threshold, a first collection level of the characteristic parameter is secondly compensated according to a dynamic compensation coefficient to determine a second collection level of the characteristic parameter, the dynamic compensation coefficient is dynamically generated according to a first difference value or / and a second difference value, the first difference value is a difference value between the change rate index and the preset change rate threshold, and the second difference value is a difference value between the fluctuation amplitude index and the preset fluctuation amplitude threshold.

8. The method of claim 1-7, wherein, The lithium battery data dynamic collection method further comprises: sequentially performing noise removal and missing point interpolation operations on each characteristic parameter data obtained through the real-time data collection to obtain each effective characteristic parameter data after data processing.

9. The method of claim 8, wherein the method further comprises: The lithium battery data dynamic collection method further comprises: performing quality evaluation on each of the effective characteristic parameter data according to a preset evaluation rule to obtain a corresponding data quality evaluation result, and the quality evaluation includes data consistency evaluation and data integrity evaluation.

10. A lithium battery data dynamic acquisition system of an energy storage power station, characterized in that, The method comprises an acquisition module, a fault type determination module, a first compensation module, a second compensation module, a collection frequency determination module and a data collection module. The acquisition module is configured to acquire fault data of the target energy storage power station in a current data collection period when the current data collection period ends. The fault type determination module is configured to determine a plurality of target fault types according to the fault data and a plurality of fault frequency threshold values, the plurality of fault frequency threshold values being obtained according to operation data of the target energy storage power station in a previous data collection period. The first compensation module is configured to perform first compensation on a basic collection level of each characteristic parameter of the target energy storage power station according to each target fault type and a preset parameter-fault correlation matrix to determine a first collection level of each characteristic parameter. The second compensation module is configured to perform second compensation on the first collection level of each characteristic parameter according to a change amplitude of each characteristic parameter in a corresponding preset time window to determine a second collection level of each characteristic parameter. The collection frequency determination module is configured to determine a collection frequency of each characteristic parameter in a next data collection period according to each second collection level. The data collection module is configured to perform real-time data collection on each characteristic parameter of the target energy storage power station according to each collection frequency in the next data collection period.

Images