Online detection method and detection system for battery performance of energy storage power station

Abstract

The invention relates to the technical field of energy storage, in particular to an energy storage power station battery performance online detection method and system. The method comprises the steps of obtaining an original data set of historical operation of a battery in an energy storage power station, extracting a first type of key time sequence characteristics representing the health state of the battery to establish a health state evaluation model, and extracting a second type of key time sequence characteristics representing the service life of the battery to establish a residual life prediction model, and third-class key time sequence characteristics representing battery operation faults are extracted to establish fault mode recognition models, time sequence data sets acquired by real-time acquisition of the batteries are input into the models respectively, and a health state result, a residual life result and a fault recognition result of battery operation in the energy storage power station are obtained. According to the invention, a high-precision health state real-time evaluation result, a residual life prediction result reflecting the long-term attenuation trend of the battery and a specific identification result for potential or already-occurring faults can be synchronously output, and timely early warning and risk prevention and control capabilities are provided for safe and stable operation of a power station.

Description

Technical Field

[0001] This invention relates to the field of energy storage technology, and in particular to an online testing method and system for the performance of batteries in energy storage power stations. Background Technology

[0002] With the rapid development of renewable energy, energy storage technology is playing an increasingly prominent role in power systems. Lithium-ion batteries, due to their high energy density and long cycle life, have become the mainstream choice for independent energy storage power stations. However, the performance of lithium-ion batteries degrades with time, temperature, and charge-discharge cycles, and the failure of a single battery module or cell can affect the safe and stable operation of the entire energy storage system.

[0003] Existing battery performance testing methods are mostly offline, requiring the shutdown of the energy storage power station, which is inefficient and cannot reflect the battery's health status in real time. Some online testing methods also suffer from insufficient testing accuracy, inability to comprehensively assess battery performance, or significant disturbance to the operating energy storage system. Therefore, there is an urgent need for a method that can perform high-precision, non-disruptive, real-time online testing of the battery performance of independent lithium-ion energy storage power stations. Summary of the Invention

[0004] The technical problem to be solved by the embodiments of the present invention is to provide an online detection method and system for battery performance in energy storage power stations, so as to solve the problem that the existing technology cannot accurately detect the battery performance in lithium-ion battery energy storage power stations.

[0005] This invention discloses an online performance testing method for batteries in an energy storage power station, comprising: Obtain the raw dataset of the historical operation of batteries in the energy storage power station, and preprocess the raw dataset; Extract the first type of key time-series features characterizing the battery health status from the preprocessed original dataset, and establish a health status assessment model based on the first type of key time-series features. A second type of key time-series features characterizing battery life are extracted from the preprocessed original dataset, and a remaining life prediction model is established based on the second type of key time-series features. A third type of key time-series feature characterizing battery operation failure is extracted from the preprocessed original dataset, and a failure mode recognition model is established based on the third type of key time-series feature. The time-series dataset of multi-dimensional operation of the battery in the energy storage power station is acquired in real time. The time-series dataset is then input into the health status assessment model, the remaining life prediction model, and the fault mode recognition model to obtain the health status result, remaining life result, and fault recognition result of the battery operation in the energy storage power station.

[0006] Optionally, obtaining the raw dataset of historical battery operation in the energy storage power station and preprocessing the raw dataset includes: Based on the operating mode and usage frequency of the batteries in the energy storage power station, the original dataset including multiple battery operating data is selected from the historical database of the energy storage power station. All time-series data in the original dataset are converted into a unified format, and the unified time-series data are processed for missing values ​​and outliers are removed to obtain the first preprocessed data. The first preprocessed data is standardized, and the standardized first preprocessed data is divided into windows of fixed length to obtain the second preprocessed data.

[0007] Optionally, establishing a health status assessment model based on the first type of key time-series features includes: Based on the battery health status assessment criteria, key characteristic parameters associated with battery health status are determined; Based on the determined key characteristic parameters, the first type of key time-series features are extracted from the second preprocessed data using time-series analysis. The first type of key time-series features include internal resistance change rate and capacity decay rate. Based on all the extracted first-class key time-series features, a machine learning algorithm is used to build the health status assessment model.

[0008] Optionally, the step of establishing a remaining lifetime prediction model based on the second type of key time-series features includes: The second type of key time-series features are extracted from the second preprocessed data based on time-series analysis. The second type of key time-series features include voltage plateau changes and self-discharge rate. Define the end of the battery's lifespan and label each of the second type of key timing features with a lifespan tag such as the number of remaining cycles; Based on all the second-class key temporal features with lifetime labels, a remaining lifetime prediction model is established using a long short-term memory network.

[0009] Optionally, establishing the fault mode recognition model based on the third type of key time-series features includes: The third type of key time-series features are extracted from the second preprocessed data based on time-series analysis. The third type of key time-series features include temperature rise characteristics and abnormal voltage fluctuations. Based on all the extracted third-category key temporal features, a classification algorithm is used to establish the fault mode recognition model.

[0010] Optionally, the online battery performance monitoring method for energy storage power stations further includes a method for inputting the time-series dataset into each model for early warning, including: The time-series dataset is acquired in real time by a multi-dimensional sensing module deployed on the battery cells in the energy storage power station. The multi-dimensional sensing module includes an impedance tester, a charge-discharge cycle recorder, a temperature sensor, a current sensor, and a voltage sensor. In response to the real-time acquisition of the time series dataset, the time series dataset is sent to the preprocessing unit via 5G wireless transmission for preprocessing of the time series dataset, and real-time time series features are extracted from the preprocessed time series dataset. In response to the extraction of real-time time-series features, the health status assessment model, the remaining life prediction model, and the fault mode recognition model are invoked respectively within a preset time period to identify the real-time time-series features; If the output results of the health status assessment model and the remaining life prediction model are less than the preset standard threshold, a first warning message is generated; if the fault mode recognition model outputs that there is a fault in battery operation, a second warning message is generated. In response to the generation of the output results of the health status assessment model, the remaining life prediction model, and the fault mode recognition model, as well as the first warning information and the second warning information, the output results and warning information are transmitted wirelessly to a remote human-machine interface terminal via 5G.

[0011] Optionally, the online battery performance testing method for the energy storage power station further includes a battery management method, comprising: Based on the battery health status, remaining lifespan, and fault type displayed on the human-machine interface, optimization commands are sent to the BMS / EMS.

[0012] This invention also discloses a testing system that employs the above-described online performance testing method for energy storage power station batteries. The testing system includes: The historical data acquisition module is used to acquire the raw dataset of the historical operation of batteries in the energy storage power station and to preprocess the raw dataset. The health status assessment model building module is used to extract the first type of key time-series features representing the battery health status from the preprocessed original dataset, and to build a health status assessment model based on the first type of key time-series features. The remaining life prediction model building module is used to extract a second type of key time-series features characterizing battery life from the preprocessed original dataset, and to build a remaining life prediction model based on the second type of key time-series features. The fault mode recognition model building module is used to extract the third type of key time series features characterizing battery operation faults from the preprocessed original dataset, and to build a fault mode recognition model based on the third type of key time series features. The battery performance detection module is used to acquire time-series datasets of multi-dimensional operation of batteries in the energy storage power station in real time. The time-series datasets are then input into the health status assessment model, the remaining life prediction model, and the fault mode recognition model to obtain the health status results, remaining life results, and fault recognition results of the batteries in the energy storage power station.

[0013] The present invention also discloses a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-described online battery performance testing method for energy storage power stations.

[0014] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described online battery performance testing method for energy storage power stations.

[0015] Compared with the prior art, the online battery performance testing method and system for energy storage power stations provided in this invention have the following advantages: By collaboratively constructing an evaluation model using three types of key time-series features, a comprehensive and non-disruptive online monitoring system for lithium-ion battery energy storage power stations is achieved. Specifically, feature extraction and model calculation are performed based on real-time collected multi-dimensional operational data, ensuring continuous online operation of battery health status assessment, lifetime prediction, and fault identification, significantly improving detection efficiency and avoiding operational interruptions. Secondly, addressing the insufficient accuracy of existing online methods, this system innovatively separates and extracts time-series features characterizing different performance dimensions: the first type of key time-series feature accurately depicts health indicators such as battery capacity decay and internal resistance changes, driving the health status assessment model to output quantifiable values ​​of battery health; the second type of key time-series feature focuses on the correlation between cyclic aging trajectory and temperature rise, constructing a remaining lifetime prediction model that integrates physical failure mechanisms, enabling rolling interval prediction of remaining lifetime; the third type of key time-series feature captures abnormal modes such as voltage platform distortion and self-discharge mutations, allowing the fault diagnosis model to provide early warnings of high-risk events such as internal short circuits and thermal runaway. While ensuring the continuous and stable operation of the energy storage system, this system effectively extends battery life, reduces operation and maintenance costs, and significantly improves the safety level of grid-side energy storage. Attached Figure Description

[0016] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 A schematic block diagram illustrating the overall steps of the online battery performance testing method for energy storage power stations provided in this embodiment of the invention; Figure 2 A schematic flowchart illustrating the online performance testing method for energy storage power station batteries provided in this embodiment of the invention. Detailed Implementation

[0017] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0018] This invention discloses an online performance testing method for batteries in energy storage power stations, such as... Figures 1-2 As shown, including: S1. Obtain the raw dataset of the historical operation of batteries in the energy storage power station and preprocess the raw dataset; S2. Extract the first type of key time-series features representing the battery health status from the preprocessed raw dataset, and establish a health status assessment model based on the first type of key time-series features. S3. Extract the second type of key time-series features characterizing battery life from the preprocessed original dataset, and establish a remaining life prediction model based on the second type of key time-series features. S4. Extract the third type of key time-series features that characterize battery operation failures from the preprocessed original dataset, and establish a fault mode recognition model based on the third type of key time-series features. S5. Real-time acquisition of time-series datasets of multi-dimensional battery operation in energy storage power stations. Input the time-series datasets into the health status assessment model, remaining life prediction model, and fault mode recognition model respectively to obtain the health status results, remaining life results, and fault recognition results of battery operation in energy storage power stations.

[0019] Through the implementation of the above-described online detection method, the system systematically processes the raw dataset of historical battery operation in energy storage power stations, extracts and distinguishes three types of key time-series features, and constructs specialized models for battery health status assessment, remaining life prediction, and operational fault identification, forming a multi-level, multi-dimensional dynamic analysis system. In real-time applications, by acquiring and inputting multi-dimensional time-series data of battery operation into the above three models in real time, high-precision real-time health status assessment results, remaining life prediction results reflecting the long-term degradation trend of the battery, and specific identification results for potential or existing faults can be output simultaneously. This effectively solves the problems of existing offline detection methods, such as requiring the power station to be shut down, low efficiency, and inability to reflect real-time status. It also overcomes the major shortcomings of other online detection methods, such as insufficient accuracy, biased assessment, or potential disturbance to the operating system itself during detection. Therefore, comprehensive and real-time monitoring of battery performance can be completed without interrupting the normal operation of the power station, ensuring the absolute non-disruptiveness of the monitoring process. By modeling and analyzing multi-dimensional time-series features, the immediate accuracy of health status diagnosis can be significantly improved, the long-term reliability of remaining life prediction can be enhanced, and the sensitivity and specificity of fault mode identification can be significantly improved, enabling it to more effectively capture early, subtle performance degradation signals and potential fault symptoms. Therefore, the method of this invention constructs a highly integrated and practical online battery performance monitoring system for independent lithium-ion battery energy storage power stations, providing timely status warnings and risk prevention capabilities for the safe and stable operation of the power station.

[0020] Furthermore, the raw dataset of historical battery operation data in the energy storage power station is obtained, and the raw dataset is preprocessed, including: Based on the operating mode and usage frequency of the batteries in the energy storage power station, the original dataset including multiple battery operating data is selected from the historical database of the energy storage power station. All time-series data in the original dataset are converted into a unified format, and missing data and outlier removal are performed on the unified time-series data to obtain the first preprocessed data; The first preprocessed data is standardized, and the standardized first preprocessed data is divided into fixed-length windows to obtain the second preprocessed data.

[0021] Through the implementation of the above online detection method, historical data was filtered based on the actual battery operating mode and usage frequency, ensuring the high representativeness and relevance of the original dataset. Unifying the format of the time-series data and performing missing value processing and outlier removal effectively eliminated noise interference and data inconsistency, significantly improving data integrity and reliability, forming high-quality first-stage preprocessed data. Subsequent standardization and data window segmentation brought all feature values ​​to a unified dimension and formed structured time-series segments, directly generating second-stage preprocessed data that the model could recognize and process. This directly solved the problem that the original operating data, due to its complex origin, inconsistent format, missing data, or errors, could not be directly and effectively used for model training. This ensured the accuracy and efficiency of key time-series feature extraction, providing high-quality, standardized, and easy-to-model input data for the subsequent construction of high-performance health status assessment models, remaining life prediction models, and fault mode recognition models, thereby improving the accuracy and robustness of the output results of the three core models.

[0022] Furthermore, a health status assessment model is established based on the first category of key time-series characteristics, including: Based on the battery health status assessment criteria, key characteristic parameters associated with battery health status are determined; Based on the determined key characteristic parameters, the first type of key time series features are extracted from the second preprocessed data using time series analysis. The first type of key time series features include internal resistance change rate and capacity decay rate. Based on all the extracted first-class key time-series features, a health status assessment model is built using machine learning algorithms.

[0023] Through the implementation of the above-described online detection method, key characteristic parameters were identified based on battery health status assessment standards, and time-sensitive indicators such as internal resistance change rate and capacity decay rate, which directly characterize battery health, were accurately located. These features were extracted using time-series analysis, ensuring that the obtained first type of key time-series features accurately reflect the continuous degradation trend of the battery during operation. Subsequently, a machine learning algorithm was used to build a model, enabling the model to learn and mine the complex nonlinear relationship between time-series features and health status, thereby achieving real-time and refined assessment of battery health status. This differs from traditional assessment methods that rely on static voltage and current parameters, effectively solving the problem of insufficient ability of existing technologies to capture the dynamic degradation process inside the battery. It can directly output health status assessment results reflecting the current actual performance of the battery, providing accurate status data for proactive operation and maintenance and risk management of energy storage power stations.

[0024] The health status assessment model employs Kalman filtering, extended Kalman filtering, particle filtering, and neural network algorithms, combined with data such as voltage, current, temperature, and impedance, to establish a battery state of health (SOH) assessment model. The model learns the aging characteristics of the battery under different operating conditions to achieve online SOH assessment.

[0025] Furthermore, a remaining lifetime prediction model is established based on the second type of key time-series features, including: Based on time series analysis, a second type of key time series features were extracted from the second preprocessed data. The second type of key time series features include voltage plateau changes and self-discharge rate. Define the end point of battery life and label each second-class key timing feature with a life label such as the number of remaining cycles; Based on all second-class key temporal features with lifetime labels, a remaining lifetime prediction model is established using a long short-term memory network.

[0026] Through the implementation of the above-described online detection method, two key time-series features—voltage plateau changes and self-discharge rate—are extracted based on time-series analysis, directly capturing core indicators reflecting battery internal structural degradation and active material loss. By clearly defining the End of Life (EOL) and labeling each feature with the remaining cycle count (e.g., the end of life can be preferably the battery capacity decaying to 80% of the rated value or the internal resistance increasing to 150%), supervised learning samples strictly corresponding to the battery's actual lifespan are constructed. A prediction model is built using a Long Short-Term Memory (LSM) network, effectively addressing the shortcomings of traditional methods in modeling long-term time-series dependencies. This allows the model to accurately learn the complex mapping from early operating characteristics to the remaining lifespan end based on the gradual trend of voltage plateau changes and the cumulative effect of self-discharge rate. This significantly improves the reliability and early prediction capability of remaining cycle count prediction, overcoming the problems of large prediction deviations and short effective periods caused by neglecting detailed changes in voltage plateau or the evolution of self-discharge rate in existing technologies. This provides a scientific basis for the lifespan management, asset replacement decisions, and preventative maintenance planning of energy storage power stations.

[0027] Among them, the remaining life prediction model is based on the SOH decay trend and historical data, combined with support vector machine and empirical aging model to predict the remaining life of the battery.

[0028] Furthermore, a fault mode recognition model is established based on the third type of key temporal features, including: Based on time series analysis, a third type of key time series features were extracted from the second preprocessed data. The third type of key time series features include temperature rise characteristics and abnormal voltage fluctuations. Based on all the extracted third-class key time-series features, a fault mode recognition model is established using a classification algorithm.

[0029] Through the implementation of the above-described online detection method, key time-series features directly reflecting battery faults, such as temperature rise characteristics and abnormal voltage fluctuations, are extracted based on time-series analysis. The core characterization quantities of abnormal behaviors, such as localized overheating, precursors to thermal runaway, or internal short circuits, are identified. A classification algorithm is used to build a model based on the extracted feature set, enabling the model to effectively learn the mapping rules between fault features and specific fault modes, accurately distinguishing different fault types. This significantly improves the timeliness and accuracy of fault identification, especially addressing the insufficient detection capability of traditional methods for transient or weak signals such as abnormal temperature rise rates and voltage ripple caused by micro-short circuits. It can quickly identify typical fault modes such as overcharging, over-discharging, internal short circuits, and poor contact, providing a reliable basis for timely alarms, accurate location, and proactive intervention, effectively ensuring the safe and stable operation of the energy storage system.

[0030] The fault mode identification model employs a statistically based 3σ criterion and K-means clustering analysis anomaly detection algorithm to monitor various parameters in real time and detect abnormal behaviors of battery modules or cells (increased voltage inconsistency, sudden changes in internal resistance, and abnormal temperature increases). By establishing a fault knowledge base, it compares abnormal features with known fault modes (open circuit, short circuit, thermal runaway risk, and sudden capacity drop) to identify fault modes.

[0031] Furthermore, the online performance monitoring method for energy storage power station batteries also includes methods for inputting time-series datasets into various models for early warning, including: The time-series dataset is acquired in real time by a multi-dimensional sensing module deployed on the battery cells in the energy storage power station. The multi-dimensional sensing module includes an impedance tester, a charge-discharge cycle recorder, a temperature sensor, a current sensor, and a voltage sensor. In response to the real-time acquisition of the time series dataset, the time series dataset is sent to the preprocessing unit via 5G wireless transmission for preprocessing of the time series dataset, and real-time time series features are extracted from the preprocessed time series dataset. In response to the extraction of real-time time series features, the health status assessment model, remaining life prediction model, and failure mode recognition model are invoked respectively within a preset time period to identify the real-time time series features. If the output of the health status assessment model and the remaining life prediction model is less than the preset standard threshold, a first warning message is generated; if the fault mode recognition model outputs that there is a fault in battery operation, a second warning message is generated. In response to the output results of the health status assessment model, remaining life prediction model, and failure mode recognition model, as well as the generation of the first and second early warning information, the output results and early warning information are transmitted wirelessly to a remote human-machine interface via 5G.

[0032] Through the implementation of the above-described online detection method, a multi-dimensional sensing module consisting of an impedance tester, a charge-discharge cycle recorder, and temperature, current, and voltage sensors is directly deployed in the battery cells of the energy storage power station to comprehensively capture core parameters affecting battery health, lifespan, and safety, forming a highly timely time-series dataset. 5G wireless transmission technology is used to instantly send the raw data to the preprocessing unit for real-time cleaning and feature extraction, solving the problem of complex wiring that may interfere with operation in traditional wired transmission, while ensuring low latency in data transmission and the immediacy of the preprocessing process. Within a preset time frame, the established health status assessment model, remaining lifespan prediction model, and fault mode recognition model are invoked in parallel to synchronously analyze the extracted real-time time-series features, significantly compressing the processing cycle from data input to output, achieving millisecond-level concurrent diagnosis of the battery's current state, degradation trend, and abnormal risks. A first warning message is generated by accurately identifying quantitative deviations in health status and remaining lifespan using preset standard thresholds, while a second warning message is triggered based on the direct output of the fault recognition model. This mechanism enables intelligent classification and targeted alarming for two types of risks: gradual performance degradation and sudden operational failures. Finally, the precise diagnostic results and graded early warning information of the three models are pushed to the remote human-machine interface in real time via the 5G channel, realizing a complete technical chain of "full-dimensional data collection - wireless uninterrupted transmission - health assessment, life prediction and fault identification - graded precise alarm - remote instant push". This architecture not only completely avoids operation interruption or additional disturbance caused by detection shutdown, but also can proactively identify and report potential risks in a graded manner with a response speed of seconds. This enables operation and maintenance personnel to take intervention measures before the performance critical point or failure occurs based on real-time, complete and accurate performance evaluation results and precise early warning information, which greatly improves the reliability, safety and management efficiency of independent energy storage power station operation. It is the core supporting link for the online detection solution to achieve the goal of high precision, real-time and uninterrupted operation.

[0033] It can set up a multi-level early warning mechanism (mild warning, moderate warning, severe warning) and provide real-time warnings through audible and visual alarms, SMS notifications, and data uploads.

[0034] As mentioned above: Voltage sensor: Used to collect voltage data of battery clusters, battery modules, and even individual battery cells in real time.

[0035] Current sensor: Used to collect the total charging and discharging current of the battery cluster in real time.

[0036] Temperature sensor: Used to collect real-time data on the internal and external temperatures of the battery module and the ambient temperature.

[0037] Charge-discharge cycle recorder: used to record the capacity, depth, and number of cycles for each charge-discharge cycle.

[0038] Impedance tester: Used to measure the impedance of a battery online under specific operating conditions using methods such as AC impedance spectrum or DC internal resistance.

[0039] Furthermore, the online performance testing method for energy storage power station batteries also includes battery management methods, including: Based on the battery health status, remaining lifespan, and fault type displayed on the human-machine interface, optimization commands are sent to the BMS / EMS.

[0040] By implementing the above-described online detection method, and using the precise battery health status score, quantified remaining life prediction value, and specific fault type identification results displayed in real time on the human-machine interface, targeted optimization instructions are directly sent to the Battery Management System (BMS) or Energy Management System (EMS). Examples include: dynamically adjusting the charge / discharge rate based on the health status to prevent over-stress damage; optimizing the charge / discharge depth based on the remaining life prediction value to extend the overall service life; or immediately triggering corresponding protection mechanisms (such as isolating the faulty module or activating the cooling system) based on the identified specific fault type. This method transforms the high-precision diagnostic conclusions output by the online detection model into executable operational strategy adjustments and proactive protection actions, solving the problem of disconnect between detection results and control decisions in traditional models. It not only significantly improves the safety margin and system reliability of energy storage power station operation but also maximizes the benefits of full lifecycle management of battery assets by extending the effective battery life and reducing unplanned downtime losses.

[0041] Once a battery abnormality or fault is detected, optimization suggestions (balanced management, temperature control adjustment, and charge / discharge strategy adjustment) can be provided to the EMS or operators based on the SOH, EOL, and fault diagnosis results, combined with the operating conditions of the energy storage power station, and maintenance reports or battery replacement suggestions can be generated.

[0042] This invention also discloses a testing system that employs the above-mentioned online performance testing method for energy storage power station batteries. The testing system includes: The historical data acquisition module is used to acquire the raw dataset of the historical operation of batteries in the energy storage power station and to preprocess the raw dataset. The health status assessment model building module is used to extract the first type of key time-series features representing the battery health status from the preprocessed raw dataset, and to build a health status assessment model based on the first type of key time-series features. The remaining lifetime prediction model building module is used to extract the second type of key time-series features characterizing battery life from the preprocessed raw dataset, and to build the remaining lifetime prediction model based on the second type of key time-series features. The fault mode recognition model building module is used to extract the third type of key time series features that characterize battery operation faults from the preprocessed raw dataset, and to build a fault mode recognition model based on the third type of key time series features. The battery performance testing module is used to acquire time-series datasets of multi-dimensional battery operation in the energy storage power station in real time. The time-series datasets are input into the health status assessment model, the remaining life prediction model, and the fault mode recognition model, respectively, to obtain the health status results, remaining life results, and fault recognition results of the battery operation in the energy storage power station.

[0043] The present invention also discloses a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for online detection of battery performance in an energy storage power station.

[0044] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-mentioned online battery performance testing method for energy storage power stations.

[0045] This invention is described based on flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to specific embodiments. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, generate instructions for implementing the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0046] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0047] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1The steps of the function specified in one or more boxes.

[0048] It should be understood that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Those skilled in the art can modify the technical solutions described in the above embodiments, or make equivalent substitutions for some of the technical features; and all such modifications and substitutions should fall within the protection scope of the present invention.

Claims

1. A method for online performance testing of batteries in an energy storage power station, characterized in that, The online performance testing method for energy storage power station batteries includes: Obtain the raw dataset of the historical operation of batteries in the energy storage power station, and preprocess the raw dataset; Extract the first type of key time-series features characterizing the battery health status from the preprocessed original dataset, and establish a health status assessment model based on the first type of key time-series features. A second type of key time-series features characterizing battery life are extracted from the preprocessed original dataset, and a remaining life prediction model is established based on the second type of key time-series features. A third type of key time-series feature characterizing battery operation failure is extracted from the preprocessed original dataset, and a failure mode recognition model is established based on the third type of key time-series feature. The time-series dataset of multi-dimensional operation of the battery in the energy storage power station is acquired in real time. The time-series dataset is then input into the health status assessment model, the remaining life prediction model, and the fault mode recognition model to obtain the health status result, remaining life result, and fault recognition result of the battery operation in the energy storage power station.

2. The online performance testing method for energy storage power station batteries according to claim 1, characterized in that, The process of acquiring the raw dataset of historical battery operation data in the energy storage power station and preprocessing the raw dataset includes: Based on the operating mode and usage frequency of the batteries in the energy storage power station, the original dataset including multiple battery operating data is selected from the historical database of the energy storage power station. All time-series data in the original dataset are converted into a unified format, and the unified time-series data are processed for missing values ​​and outliers are removed to obtain the first preprocessed data. The first preprocessed data is standardized, and the standardized first preprocessed data is divided into windows of fixed length to obtain the second preprocessed data.

3. The online performance testing method for energy storage power station batteries according to claim 2, characterized in that, The step of establishing a health status assessment model based on the first type of key time-series features includes: Based on the battery health status assessment criteria, key characteristic parameters associated with battery health status are determined; Based on the determined key characteristic parameters, the first type of key time-series features are extracted from the second preprocessed data using time-series analysis. The first type of key time-series features include internal resistance change rate and capacity decay rate. Based on all the extracted first-class key time-series features, a machine learning algorithm is used to build the health status assessment model.

4. The online performance testing method for energy storage power station batteries according to claim 2, characterized in that, The step of establishing a remaining lifetime prediction model based on the second type of key time-series features includes: The second type of key time-series features are extracted from the second preprocessed data based on time-series analysis. The second type of key time-series features include voltage plateau changes and self-discharge rate. Define the end of the battery's lifespan and label each of the second type of key timing features with a lifespan tag such as the number of remaining cycles; Based on all the second-class key temporal features with lifetime labels, a remaining lifetime prediction model is established using a long short-term memory network.

5. The online performance testing method for energy storage power station batteries according to claim 2, characterized in that, The step of establishing a fault mode recognition model based on the third type of key time-series features includes: The third type of key time-series features are extracted from the second preprocessed data based on time-series analysis. The third type of key time-series features include temperature rise characteristics and abnormal voltage fluctuations. Based on all the extracted third-category key temporal features, a classification algorithm is used to establish the fault mode recognition model.

6. The online performance testing method for energy storage power station batteries according to claim 1, characterized in that, The online performance monitoring method for energy storage power station batteries also includes a method for inputting the time-series dataset into each model for early warning, including: The time-series dataset is acquired in real time by a multi-dimensional sensing module deployed on the battery cells in the energy storage power station. The multi-dimensional sensing module includes an impedance tester, a charge-discharge cycle recorder, a temperature sensor, a current sensor, and a voltage sensor. In response to the real-time acquisition of the time series dataset, the time series dataset is sent to the preprocessing unit via 5G wireless transmission for preprocessing of the time series dataset, and real-time time series features are extracted from the preprocessed time series dataset. In response to the extraction of real-time time-series features, the health status assessment model, the remaining life prediction model, and the fault mode recognition model are invoked respectively within a preset time period to identify the real-time time-series features; If the output results of the health status assessment model and the remaining life prediction model are less than the preset standard threshold, a first warning message is generated; if the fault mode recognition model outputs that there is a fault in battery operation, a second warning message is generated. In response to the generation of the output results of the health status assessment model, the remaining life prediction model, and the fault mode recognition model, as well as the first warning information and the second warning information, the output results and warning information are transmitted wirelessly to a remote human-machine interface terminal via 5G.

7. The online performance testing method for energy storage power station batteries according to claim 6, characterized in that, The online battery performance testing method for the energy storage power station also includes battery management methods, including: Based on the battery health status, remaining lifespan, and fault type displayed on the human-machine interface, optimization commands are sent to the BMS / EMS.

8. A testing system, employing the online performance testing method for energy storage power station batteries according to any one of claims 1-7, characterized in that, The detection system includes: The historical data acquisition module is used to acquire the raw dataset of the historical operation of batteries in the energy storage power station and to preprocess the raw dataset. The health status assessment model building module is used to extract the first type of key time-series features representing the battery health status from the preprocessed original dataset, and to build a health status assessment model based on the first type of key time-series features. The remaining life prediction model building module is used to extract a second type of key time-series features characterizing battery life from the preprocessed original dataset, and to build a remaining life prediction model based on the second type of key time-series features. The fault mode recognition model building module is used to extract the third type of key time series features characterizing battery operation faults from the preprocessed original dataset, and to build a fault mode recognition model based on the third type of key time series features. The battery performance detection module is used to acquire time-series datasets of multi-dimensional operation of batteries in the energy storage power station in real time. The time-series datasets are then input into the health status assessment model, the remaining life prediction model, and the fault mode recognition model to obtain the health status results, remaining life results, and fault recognition results of the batteries in the energy storage power station.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the online battery performance testing method for energy storage power stations as described in any one of claims 1-7.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the online battery performance testing method for energy storage power stations as described in any one of claims 1-7.

Images