Battery intelligent fault diagnosis method and system suitable for power grid energy storage system

Abstract

The invention relates to the technical field of power grid energy storage systems, in particular to a battery intelligent fault diagnosis method and system suitable for a power grid energy storage system. According to the method, multi-source monitoring data including battery electrical parameters, power grid features, environmental conditions and historical operation records are obtained, unified time step alignment and feature extraction processing are carried out on the data, and statistical features, time sequence features and frequency domain features are obtained. A double-layer diagnosis model structure is adopted, a random forest model is used for diagnosing main fault modes such as over-charging and over-discharging, and a convolutional neural network model is used for diagnosing complex fault modes such as short circuit and temperature out-of-control. The system has a dynamic updating mechanism, and online optimization is realized by calculating incremental updating values of model parameters; meanwhile, a model evaluation mechanism is established, and when an evaluation index is lower than a preset standard, a retraining process is triggered. According to the invention, high-precision and real-time identification and early warning of the battery fault are realized, and the operation reliability and safety of the energy storage system are significantly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid energy storage systems, and in particular to a battery intelligent fault diagnosis method and system applicable to power grid energy storage systems. Background Art

[0002] Among the many energy storage technologies, electrochemical energy storage systems (such as lithium-ion batteries, lead-acid batteries, etc.) have become the mainstream choice for grid energy storage due to their high efficiency, rapid response, modularity and scalability. This type of battery energy storage system forms a battery pack by connecting multiple battery cells in series and parallel to meet the power and capacity requirements of the grid. However, as the core component of the energy storage system, the performance and life of the battery are affected by many factors, including the number of charge and discharge cycles, operating temperature, depth of discharge, load fluctuations, and operating environment. When the battery has problems such as capacity decay, increased internal resistance, and abnormal temperature, it will not only affect the output quality and availability of the energy storage system, but may even cause equipment damage and safety risks. In addition, with the continuous increase in the scale of the power grid and energy storage capacity, the number of batteries is huge and the distribution is complex. The traditional operation and maintenance model can no longer meet the requirements of large-scale battery health management.

[0003] At present, battery fault diagnosis technology is not yet mature, and the main problems are reflected in the following aspects:

[0004] Traditional battery management systems (BMS) rely primarily on basic parameters such as voltage, current, and temperature for status monitoring, making it difficult to promptly detect early signs of potential faults, such as increased internal resistance and capacity degradation. By the time a fault becomes apparent, it often has already adversely affected the operation of the energy storage system, impacting the stability and reliability of the power grid. In actual grid energy storage operations, once a battery failure occurs, traditional methods often require manual intervention, resulting in lengthy repair and recovery times, increased O&M costs, and reduced grid reliability and flexibility.

[0005] Battery fault diagnosis is independent of the grid dispatch mechanism and lacks immediate linkage. It is impossible to quickly adjust the grid load or activate backup energy storage equipment according to changes in battery status, which weakens the energy storage system's ability to respond to grid peak and frequency regulation and emergency power supply.

[0006] Current research mostly relies on empirical models and simple algorithms, lacks in-depth application of big data and intelligent algorithms, has insufficient prediction accuracy, weak model adaptability, and limited intelligence, and cannot meet the urgent needs of large-scale energy storage systems for high precision, real-time and intelligent management. Summary of the Invention

[0007] In view of the problems existing in the prior art, the present invention is proposed.

[0008] Therefore, the problem to be solved by the present invention is how to solve the problems of delayed battery fault detection, long recovery time after fault, and disconnection between battery fault detection and grid scheduling in traditional battery management systems.

[0009] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0010] In a first aspect, an embodiment of the present invention provides a battery intelligent fault diagnosis method applicable to a power grid energy storage system, comprising: obtaining multi-source monitoring data of a target energy storage system, and fusing the multi-source monitoring data to form first monitoring data;

[0011] Presetting a first fault diagnosis model, and inputting the first monitoring data into the first fault diagnosis model;

[0012] Perform intelligent battery fault diagnosis based on the output of the first fault diagnosis model.

[0013] As a preferred solution of the battery intelligent fault diagnosis method applicable to the power grid energy storage system of the present invention, the multi-source monitoring data includes:

[0014] Battery electrical parameter data, including voltage, current, and temperature parameters;

[0015] Grid characteristic data, including load power and dispatch plan indicators;

[0016] Environmental condition data, including ambient temperature and humidity;

[0017] Historical operation record data, including the accumulated number of cycles and charge and discharge curves.

[0018] As a preferred solution of the battery intelligent fault diagnosis method applicable to the power grid energy storage system of the present invention, the data fusion to form the first monitoring data includes:

[0019] Performing unified time step alignment on the multi-source monitoring data;

[0020] Extracting statistical features of the multi-source monitoring data, wherein the statistical features include mean, variance, skewness and kurtosis;

[0021] Extracting time series features of the multi-source monitoring data, wherein the time series features include trend features and autocorrelation coefficients;

[0022] Extract frequency domain features of the multi-source monitoring data, where the frequency domain features include power spectral density.

[0023] As a preferred solution of the battery intelligent fault diagnosis method applicable to the power grid energy storage system of the present invention, the first fault diagnosis model includes:

[0024] Random forest model, used to diagnose major fault modes such as overcharging, over-discharging, overheating, increased internal resistance, and capacity attenuation;

[0025] Convolutional neural network models for diagnosing complex failure modes such as short circuits, damage, temperature runaway, charge imbalance, and cluster failures;

[0026] The loss function of the first fault diagnosis model includes positioning loss, confidence loss and classification loss.

[0027] As a preferred solution of the battery intelligent fault diagnosis method applicable to the power grid energy storage system of the present invention, the battery fault intelligent diagnosis based on the output of the first fault diagnosis model includes:

[0028] Calculate the posterior probability of each fault type;

[0029] When the posterior probability of a certain fault type is greater than the preset threshold, it is determined to be that type of fault;

[0030] An early warning signal is sent to the battery management system based on the fault judgment result.

[0031] As a preferred solution of the battery intelligent fault diagnosis method applicable to the power grid energy storage system described in the present invention, it also includes a dynamic update mechanism:

[0032] Calculate the incremental update values ​​of the model parameters based on the newly acquired monitoring data and model prediction errors;

[0033] performing online updating on parameters of the first fault diagnosis model;

[0034] Redo the fault diagnosis based on the updated model.

[0035] As a preferred solution of the battery intelligent fault diagnosis method applicable to the power grid energy storage system described in the present invention, it also includes a model evaluation mechanism:

[0036] Calculate the model's precision, recall, and F1 score on the validation set;

[0037] Generate confusion matrix to analyze the diagnostic accuracy of various faults;

[0038] When the model evaluation index is lower than the preset standard, the model retraining process is triggered.

[0039] In a second aspect, an embodiment of the present invention provides a battery intelligent fault diagnosis system applicable to a power grid energy storage system, which includes a data acquisition module for acquiring multi-source monitoring data of a target energy storage system and performing data fusion on the multi-source monitoring data to form first monitoring data;

[0040] a preset module, which presets a first fault diagnosis model and inputs the first monitoring data into the first fault diagnosis model;

[0041] An output module performs intelligent battery fault diagnosis based on the output of the first fault diagnosis model.

[0042] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the battery intelligent fault diagnosis method applicable to the grid energy storage system as described in the first aspect of the present invention are implemented.

[0043] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the battery intelligent fault diagnosis method applicable to a grid energy storage system as described in the first aspect of the present invention are implemented.

[0044] The beneficial effects of the present invention are: compared with traditional fault diagnosis methods, such as simple threshold methods and traditional algorithms, the method of the present invention shows higher accuracy and faster response time in multiple application scenarios. Through a unique feature extraction strategy, combined with the classification models of Random Forest and Convolutional Neural Network (CNN) to process the main fault modes and complex fault modes respectively, this method can accurately identify the multi-dimensional characteristics of battery faults. The feature extraction process not only includes basic statistical features and time series features, but also incorporates frequency domain feature analysis to capture subtle changes and complex fault modes of the battery system. The multi-dimensional feature extraction strategy greatly improves the model's ability to identify different types of faults, and effectively reduces the occurrence of false alarms and missed reports. In addition, this method ensures real-time data processing and efficient model reasoning.

[0045] The method of the present invention is outstanding in improving the stability and safety of power grid energy storage systems, especially in its application in early warning of faults. Through real-time monitoring and precise diagnosis, the system can promptly identify and issue early warning signals to prevent further spread and deterioration of faults. This mechanism effectively improves the overall operational stability of the power grid energy storage system, ensuring that the system can maintain balance and stability in the face of fluctuations in renewable energy. In addition, identifying potential battery failures in advance reduces safety hazards caused by battery failures, such as the risk of major accidents such as fire and explosion, ensures the safe operation of the energy storage system and peripheral equipment, and improves the safety of the entire power grid system.

[0046] After implementing the intelligent fault diagnosis method of the present invention, the power grid energy storage system has shown significant improvements in multiple economic indicators. Accurate fault diagnosis reduces unnecessary repair and replacement costs caused by false alarms and missed alarms. At the same time, it extends the service life of the battery and reduces the frequency of battery replacement, thereby significantly reducing operating and maintenance costs. Through real-time monitoring and automated fault diagnosis, operation and maintenance personnel can perform maintenance and management more efficiently, reduce the frequency and workload of manual inspections, and improve overall maintenance efficiency. Rapid and accurate fault diagnosis and timely early warning mechanisms reduce the time the system is down due to faults, improve the availability of the energy storage system and the overall operating efficiency of the power grid, and not only reduce the economic losses caused by downtime, but also improve the service quality of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0048] Figure 1 A flow chart of a battery intelligent fault diagnosis method applicable to a grid energy storage system;

[0049] Figure 2 A computer device diagram for a battery intelligent fault diagnosis method applicable to a grid energy storage system;

[0050] Figure 3 The system architecture diagram of the battery intelligent fault diagnosis method suitable for grid energy storage systems. DETAILED DESCRIPTION

[0051] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0052] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0053] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it designate a separate or selective embodiment that is mutually exclusive with other embodiments.

[0054] Example 1

[0055] Reference Figures 1 and 2 , which is the first embodiment of the present invention, provides a battery intelligent fault diagnosis method applicable to a grid energy storage system, comprising:

[0056] S100: Acquire multi-source monitoring data of a target energy storage system, and perform data fusion on the multi-source monitoring data to generate first monitoring data;

[0057] In the embodiments of this application, multi-source monitoring data primarily comes from a variety of sensors and monitoring devices deployed within the energy storage system. Specifically, these devices include voltage sensors, current sensors, and temperature sensors located at key locations in the battery pack, as well as environmental monitors and grid status monitors. These sensors utilize industrial-grade standard designs, offering high precision and stability, ensuring accurate data collection.

[0058] In an optional embodiment, when fusing multi-source monitoring data, the sampling frequencies of different data sources can be flexibly configured. For example, the sampling frequency of voltage and current data can be set to 100Hz to capture rapidly changing electrical characteristics; the sampling frequency of temperature data can be set to 1Hz, as temperature changes relatively slowly; and the sampling frequency of environmental data can be set to 0.1Hz to meet environmental monitoring requirements. This differentiated sampling strategy ensures the monitoring accuracy of key parameters while avoiding the waste of data storage and processing resources.

[0059] It should be noted that the data acquisition system utilizes a layered architecture, consisting of field, control, and management layers. The field layer is responsible for basic data collection, utilizing industrial communication protocols such as CAN bus and RS485. The control layer performs preliminary data processing and local storage, utilizing a PLC or industrial computer. The management layer is responsible for comprehensive data analysis and storage, utilizing industrial servers. This layered design improves system reliability and scalability.

[0060] S101: Multi-source monitoring data includes:

[0061] Battery electrical parameter data, including voltage, current and temperature parameters;

[0062] Grid characteristic data, including load power and dispatch plan indicators;

[0063] Environmental condition data, including ambient temperature and humidity;

[0064] Historical operation record data, including the cumulative number of cycles and charge and discharge curves.

[0065] In the embodiment of the present application, for the collection of electrical parameter data: the voltage sensor collects the battery cell voltage and group voltage, with a measurement range of 0-5V (cell) and 0-1000V (group voltage), and an accuracy of ±0.1%; the current sensor adopts a Hall sensor solution with a measurement range of ±500A and an accuracy of ±0.5%; the temperature sensor adopts a PT100 platinum resistance with a measurement range of -40℃ to 125℃ and an accuracy of ±0.1℃.

[0066] In an optional embodiment, the collection of electrical parameter data can also include internal resistance measurement. By injecting a small signal excitation during the battery's charge and discharge processes, the battery's internal resistance is measured using the AC impedance method. The internal resistance measurement system has a measurement range of 0.1mΩ to 100mΩ, an accuracy of ±1%, and an adjustable measurement frequency from 0.1Hz to 1kHz. This method can detect signs of battery performance degradation earlier.

[0067] It should be noted that electrical parameter collection utilizes a redundant design strategy. Dual sensors are deployed at key measurement points. These sensors operate simultaneously but employ different measurement principles, such as voltage division and differential measurement for voltage, and thermocouples and platinum resistance thermometers for temperature. If one sensor fails or detects abnormal data, the system automatically switches to the other sensor, ensuring continuous and reliable data collection.

[0068] In an embodiment of the present application, for the collection of grid characteristic data: load power is monitored in real time by an electric energy analyzer with a sampling rate of 10kHz, which can measure parameters such as active power, reactive power, and power factor; scheduling plan indicators are obtained in real time from the grid scheduling system, including target power curves, backup capacity requirements, etc.

[0069] In an optional embodiment, grid characteristic data can also include grid frequency and voltage quality parameters. Frequency measurement accuracy is ±0.01 Hz, and voltage harmonics can be measured up to the 50th harmonic. These parameters help analyze the grid-connected operation status of the energy storage system and identify potential problems in a timely manner.

[0070] It should be noted that the system has established a hierarchical caching mechanism for grid characteristic data. Based on data importance and usage frequency, the data is divided into three levels: real-time, minute-level, and hourly. These levels are stored in an in-memory database, a relational database, and a time-series database, respectively. This ensures efficient data queries while saving storage space.

[0071] In this embodiment, environmental condition data is monitored using high-precision temperature and humidity sensors with a temperature measurement range of -40°C to 85°C and an accuracy of ±0.3°C; and a humidity measurement range of 0.100% RH with an accuracy of ±2% RH. These sensors are located at the energy storage system's air inlet and outlet, as well as around key equipment.

[0072] In an optional embodiment, environmental monitoring can also include parameters such as dust concentration and combustible gas concentration. For example, a laser dust sensor can be used to measure PM2.5 and PM10 concentrations, and an electrochemical sensor can be used to monitor H2 concentration. These parameters are crucial for evaluating the operating environment of the energy storage system.

[0073] It should be noted that the environmental monitoring system utilizes a zoning management approach. The energy storage system is divided into functional areas, such as the battery compartment, PCS compartment, and power distribution compartment, with each area equipped with an independent monitoring unit. In addition to being used for fault diagnosis, monitoring data is also used for closed-loop control of environmental control systems, such as ventilation and cooling.

[0074] In the embodiment of the present application, for the maintenance of historical operation record data: the system automatically records the start and end time, depth, power and other parameters of each charge and discharge cycle; calculates the cumulative number of cycles by integration with an accuracy of 0.1 times; and stores complete charge and discharge curve data, including the time series changes of voltage, current and temperature.

[0075] In an optional embodiment, historical records can also include maintenance records and alarm records. Maintenance records contain detailed information about operations such as regular inspections and component replacements; alarm records contain fault time, type, and corrective action. These records help establish a health profile for the device and improve diagnostic accuracy.

[0076] It should be noted that historical data utilizes a distributed storage architecture. Recent data (e.g., the past three months) is stored on local servers for quick access, while historical data is stored in cloud data centers to support big data analysis. This two-tier storage system maintains data consistency through an incremental synchronization mechanism.

[0077] S102: Data fusion to form first monitoring data includes:

[0078] Align multi-source monitoring data to a unified time step;

[0079] Extract statistical features of multi-source monitoring data, including mean, variance, skewness and kurtosis;

[0080] Extract time series features of multi-source monitoring data, including trend features and autocorrelation coefficients;

[0081] Extract frequency domain features of multi-source monitoring data, including power spectral density.

[0082] In the embodiment of this application, time step alignment uses a two-level synchronization mechanism. First, a GPS or Beidou timing module provides a unified clock source for all acquisition devices, with a time accuracy better than 1μs. Then, an interpolation algorithm is used to align data of different frequencies, supporting multiple methods such as linear interpolation and cubic spline interpolation.

[0083] In an optional embodiment, data alignment can also employ an adaptive step-size strategy. When the battery state changes dramatically (e.g., sudden load increases or decreases), the alignment step size is automatically reduced to 10ms to capture transient characteristics. During steady-state operation, the step size is appropriately increased to 1s to reduce data volume. This strategy ensures data accuracy at critical moments while improving system efficiency.

[0084] It should be noted that the system has established a data quality assessment mechanism. By checking the completeness, accuracy, and timeliness of the data, each data point is assigned a quality label. In subsequent feature extraction, the feature calculation method is dynamically adjusted based on data quality, such as using robust statistical methods for noisy data.

[0085] In this embodiment, statistical features are extracted using a sliding window mechanism. The window length is configurable, typically 5 minutes. A 50% window overlap ensures feature continuity. For each window, statistics such as mean, variance, skewness, and kurtosis are calculated to characterize the data distribution.

[0086] In an optional embodiment, additional statistical features can be calculated, such as quantiles, crest factor, and crest factor. These features describe the statistical regularity of the data from different perspectives, helping to improve the accuracy of fault diagnosis. In particular, crest factor plays an important role in detecting sudden faults such as battery short circuits.

[0087] It should be noted that the calculation of statistical features uses an incremental update algorithm. Whenever new data arrives, only the incremental portion related to the old and new data needs to be calculated, significantly reducing the amount of computation required. Furthermore, the system periodically performs full calculations to prevent the accumulation of errors in incremental calculations.

[0088] In the examples of this application, time series feature extraction focuses on the dynamic changes in the data. Trend features are obtained through least squares fitting, reflecting the rate of change of parameters. The calculation of the autocorrelation coefficient supports multiple delay values ​​(such as 1 minute, 5 minutes, and 15 minutes) to capture periodic changes at different time scales.

[0089] In an optional embodiment, the time series features may also include nonlinear features such as conditional entropy and Lyapunov exponent. These features can characterize the dynamic characteristics of the battery system, especially its instability under complex operating conditions, and provide an important basis for early warning of faults.

[0090] It should be noted that the seasonal and cyclical effects of data are taken into account during time series feature extraction. The system automatically identifies and compensates for regular changes such as daily and weekly cycles, thereby more accurately extracting abnormal change features. For long-term trends, a sliding baseline method is used to eliminate the effects of drift.

[0091] In the present embodiment, frequency domain feature extraction is implemented using a fast Fourier transform (FFT). A Hanning window function is used to reduce spectral leakage, and the number of transformation points is 8192, ensuring frequency resolution. The power spectral density is calculated to analyze the energy distribution of the signal in each frequency band.

[0092] In an optional embodiment, the frequency domain analysis may also use time-frequency analysis methods such as wavelet transform or Hilbert-Huang transform. These methods can provide time-frequency joint distribution information of the signal and are particularly suitable for analyzing transient processes and non-stationary signals in battery systems.

[0093] It should be noted that an automated frequency band partitioning strategy is employed during the frequency domain feature extraction process. The spectrum is divided into multiple monitoring bands based on the characteristic frequencies of different fault types. The range of each frequency band can be dynamically adjusted based on actual operational experience to improve the targeted nature of the features.

[0094] S200: Preset a first fault diagnosis model and input first monitoring data into the first fault diagnosis model;

[0095] In this embodiment, the first fault diagnosis model employs a two-layer structure. The first layer uses a random forest model to identify common faults, such as overcharging and over-discharging; the second layer uses a convolutional neural network to handle complex fault patterns. These two layers work together to ensure both rapid diagnosis and improved accuracy.

[0096] In an optional embodiment, an attention mechanism can be introduced into the model. By learning the importance weights of different features, it automatically focuses on the most critical indicators for the current fault diagnosis. This mechanism improves the interpretability of the model and improves diagnostic performance.

[0097] It should be noted that the model is trained using a hybrid online-offline approach. The base model is obtained through offline training using historical fault data and expert-labeled samples. After deployment, it is continuously optimized through incremental learning to adapt to dynamic system changes. Cross-validation is used during training to ensure the model's generalization capabilities.

[0098] S201: The first fault diagnosis model includes:

[0099] Random forest model, used to diagnose major fault modes such as overcharging, over-discharging, overheating, increased internal resistance, and capacity attenuation;

[0100] Convolutional neural network models for diagnosing complex failure modes such as short circuits, damage, temperature runaway, charge imbalance, and cluster failures;

[0101] The loss function of the first fault diagnosis model includes positioning loss, confidence loss and classification loss.

[0102] In the embodiment of the present application, the configuration of the random forest model is as follows: the number of decision trees is set to 100 to balance computational efficiency and model performance; the maximum depth of each tree is 10 layers to prevent overfitting; the feature selection adopts the Gini coefficient evaluation indicator; the sample sampling uses weighted random sampling, and the sampling weight is appropriately increased for rare fault types.

[0103] In an optional embodiment, the random forest model can also integrate multiple sub-models, each specialized for a specific fault type. For example, the overcharge and over-discharge sub-model focuses on voltage characteristics, the overheating sub-model focuses on temperature characteristics, and the internal resistance and capacity sub-models primarily analyze impedance spectrum characteristics. This specialized design improves the model's diagnostic accuracy.

[0104] It should be noted that the random forest model uses a distributed training framework. Training data is sharded by fault type, and sub-models are trained in parallel on multiple compute nodes. The results are then aggregated through a voting mechanism. The system supports hot updates, allowing model parameters to be updated without downtime.

[0105] In this embodiment, the convolutional neural network model uses the ResNet architecture, comprising 18 convolutional layers. The input layer receives multi-channel time series data, with each channel corresponding to a monitoring parameter. The intermediate layers use 1D convolution to extract time series features. The fully connected layers perform the final classification. The model is trained using the Adam optimizer, with an initial learning rate of 0.001 and a cosine annealing strategy.

[0106] In an optional embodiment, a convolutional neural network can adopt a multi-branch structure. The main branch processes the raw time series data, while the auxiliary branches process statistical features and frequency domain features. A feature fusion layer combines the outputs of multiple branches, leveraging the strengths of different feature types. This design improves the model's ability to identify complex faults.

[0107] It should be noted that to improve the robustness of the model, various data augmentation techniques were employed during training, including the addition of Gaussian noise, random scaling, and time shifting. Dropout and L2 regularization were also used to prevent overfitting. The output of the model's intermediate layers is visualized to facilitate understanding of the network's decision-making process.

[0108] In this embodiment, the loss function is designed as follows: the location loss uses mean squared error (MSE) to evaluate the accuracy of fault time prediction; the confidence loss uses cross entropy to measure the reliability of the predicted probability; and the classification loss uses focal loss to focus on difficult-to-classify samples. The total loss is obtained by weighted summation of these three losses.

[0109] In an optional embodiment, the loss function can also include a temporal consistency loss. This loss term penalizes drastic fluctuations in prediction results, making the diagnosis more stable. A knowledge distillation loss is also introduced, using the output of the expert system to guide model learning.

[0110] It should be noted that the weight coefficients of the loss function use an adaptive adjustment mechanism. In the early stages of training, increasing the weight of the classification loss accelerates convergence. As training progresses, gradually increasing the weights of the localization loss and confidence loss improves model accuracy. The system records the loss distribution of each batch for evaluating training effectiveness.

[0111] S300: Perform intelligent battery fault diagnosis based on the output of the first fault diagnosis model.

[0112] S301: Performing intelligent battery fault diagnosis according to the output of the first fault diagnosis model includes:

[0113] Calculate the posterior probability of each fault type;

[0114] When the posterior probability of a certain fault type is greater than the preset threshold, it is determined to be that type of fault;

[0115] An early warning signal is sent to the battery management system based on the fault judgment result.

[0116] In the embodiments of this application, the calculation of posterior probabilities uses a Bayesian framework. First, the probability distribution of the fault type output by the model is obtained. Then, the posterior probability is calculated in combination with prior knowledge (such as equipment age and historical faults). The historical records of system maintenance probability calculations are used to optimize the warning threshold.

[0117] In an optional embodiment, the posterior probability calculation can also take into account the influence of environmental factors and operating conditions. For example, the warning threshold for overheating faults can be appropriately lowered during hot weather, while the sensitivity for detecting internal resistance faults can be increased under high-load conditions. This context-aware diagnostic strategy improves the accuracy of warnings.

[0118] It should be noted that the system utilizes a multi-level warning threshold design. For different types of faults, thresholds are set at three levels: warning, alarm, and fault. When the posterior probability exceeds the corresponding threshold, different levels of response measures are triggered. The thresholds can be dynamically adjusted based on operational experience and system importance.

[0119] In an embodiment of the present application, the generation of the early warning signal includes multiple steps: first, determining the severity and urgency of the fault; then generating an early warning message in a standard format, including information such as the fault type, location, and time; finally, sending the early warning through multiple channels to ensure that the information is delivered in a timely manner.

[0120] In an optional embodiment, the early warning system can also automatically generate treatment suggestions. Based on historical fault handling experience, the system recommends appropriate treatment plans, including emergency measures and repair suggestions. These suggestions can help operations and maintenance personnel quickly respond to faults.

[0121] It should be noted that early warning information is delivered using a tiered push strategy. Different levels of detail are delivered to personnel in different positions. On-site operations and maintenance personnel receive detailed technical parameters, while management personnel receive a fault overview and impact assessment. The system supports early warning confirmation and tracking of handling progress.

[0122] S302: Also includes a dynamic update mechanism:

[0123] Calculate the incremental update values ​​of the model parameters based on the newly acquired monitoring data and model prediction errors;

[0124] performing online updating on parameters of the first fault diagnosis model;

[0125] Redo the fault diagnosis based on the updated model.

[0126] In this embodiment, incremental updates utilize a sliding window mechanism. The most recent 30 days of operational data are retained, and model parameters are updated every 24 hours. The update process utilizes a stochastic gradient descent algorithm, with the learning rate adaptively adjusted based on the prediction error.

[0127] In an optional embodiment, the update mechanism can also include dynamic adjustments to the model structure. For example, when a new failure mode is discovered, the model's output layer is automatically expanded; when the discriminative power of certain features decreases, the network structure is appropriately adjusted. This adaptive architecture improves the model's scalability.

[0128] It should be noted that model updates use a gradual strategy to avoid drastic parameter changes. Before each update, the effect of the new parameters is evaluated on the validation set, and updates are only applied if performance improves. The system saves the five most recent versions of model parameters and supports rapid rollback.

[0129] In this embodiment of the present application, trigger conditions for model updates include: prediction error exceeding a set threshold; accumulation of a certain number of new samples; discovery of a new fault type; or significant changes in system operating conditions. The update process is performed on the backup server and does not affect the operation of the primary system.

[0130] In an optional embodiment, the update process can also incorporate expert knowledge. Through a human-computer interface, experts can review and modify the model's diagnostic results. These modifications serve as new training samples for model optimization. This human-computer collaborative approach improves the reliability of the model.

[0131] It should be noted that the system has established a model performance tracking mechanism. This records evaluation metrics before and after each update, such as accuracy, recall, and F1 score. By analyzing trends in these metrics, the health of the model can be assessed, allowing performance degradation issues to be identified and addressed promptly.

[0132] S303: Also includes model evaluation mechanism:

[0133] Calculate the model's precision, recall, and F1 score on the validation set;

[0134] Generate confusion matrix to analyze the diagnostic accuracy of various faults;

[0135] When the model evaluation index is lower than the preset standard, the model retraining process is triggered.

[0136] In this embodiment, model evaluation utilizes a tiered validation mechanism. The validation set is sampled stratified by fault type to ensure sufficient samples for each fault type. Evaluation metrics include precision, recall, F1 score, receiver operating characteristic (ROC) curve, and area under the curve (AUC). The system automatically calculates these metrics every hour and generates an evaluation report.

[0137] In an optional embodiment, the evaluation can also include a timeliness analysis of the model. By comparing prediction accuracy across different time windows, the model's stability and timeliness can be assessed. For example, performance indicators can be calculated over sliding windows of 24 hours, 7 days, and 30 days to promptly identify degradation trends in model performance. This multi-scale evaluation approach provides a comprehensive understanding of the model's operational status.

[0138] It should be noted that the evaluation process uses a weighted calculation method. High-risk fault types (such as short circuits and thermal runaway) are given higher weights, while low-risk fault types (such as minor capacity fade) are given lower weights. This differentiated evaluation strategy better meets actual application needs.

[0139] In the present embodiment, the confusion matrix analysis uses an automated tool. The system generates a detailed confusion matrix report, including:

[0140] Accuracy, missed detection rate, and false detection rate for each type of fault; identification of easily confused fault types; confidence distribution of fault judgments; and characteristic analysis of misdiagnosis cases.

[0141] In an optional embodiment, confusion matrix analysis can also be combined with fault severity. Faults are categorized into multiple levels based on severity, and the diagnostic accuracy for each level is calculated. For misdiagnoses of high-severity faults, the system triggers a manual review mechanism. Furthermore, by analyzing the causes of misdiagnosis, fault characteristics and diagnostic rules are continuously optimized.

[0142] It should be noted that the system has established a comprehensive mechanism for tracking assessment results. All assessment data is stored in a dedicated performance database, supporting historical trend analysis and comparison. Assessment reports are automatically distributed to relevant personnel, triggering manual intervention when necessary.

[0143] In an embodiment of the present application, the trigger conditions for model retraining include: the overall accuracy rate is lower than 90%; the recall rate of a certain type of critical fault is lower than 85%; the false alarm rate exceeds 5%; the performance index declines for 72 consecutive hours; and the number of new fault samples exceeds 20% of the existing samples.

[0144] In an optional embodiment, the retraining process can also include data cleaning and feature reconstruction. Before starting training, historical data is quality assessed and cleaned to remove noisy samples and outdated data. Simultaneously, feature engineering solutions are optimized based on feature analysis of newly added samples. This preprocessing can improve the effectiveness of retraining.

[0145] It should be noted that the retraining adopts a step-by-step iterative strategy: first, use the latest data for small-scale training to evaluate the effect; if the effect is significantly improved, expand the training scale; perform final training on the complete dataset; maintain the model's online service capability during training, and adopt a dual-model switching mechanism.

[0146] In an embodiment of the present application, the model deployment process after retraining is completed includes: verifying the performance of the new model in a test environment; conducting A / B testing to compare the effects of the new and old models; gradually switching business traffic to the new model; continuously monitoring the operating status of the new model; and retaining a rollback mechanism to ensure system reliability.

[0147] In an optional embodiment, a model version management mechanism can also be established. Each version of the model has a complete configuration record, including: the feature distribution of the training dataset; the model structure and parameters; the evaluation results and deployment history; the version update instructions and the person in charge.

[0148] It should be noted that the system conducts a comprehensive evaluation of the retraining results, including: performance improvement analysis; resource consumption assessment; impact on various fault diagnosis; comparison with historical versions; and stability and reliability verification.

[0149] This complete evaluation and retraining mechanism ensures that the fault diagnosis model always remains in optimal condition, providing reliable protection for the safe operation of the energy storage system.

[0150] Furthermore, this embodiment also provides a battery intelligent fault diagnosis system applicable to a power grid energy storage system, comprising:

[0151] A data acquisition module, which acquires multi-source monitoring data of the target energy storage system and performs data fusion on the multi-source monitoring data to form first monitoring data;

[0152] A preset module, which presets a first fault diagnosis model and inputs the first monitoring data into the first fault diagnosis model;

[0153] The output module performs intelligent diagnosis of battery faults according to the output of the first fault diagnosis model.

[0154] In summary, this paper proposes a method and system for intelligent battery fault diagnosis in grid energy storage systems. By combining data-driven machine learning models with the physical mechanisms of batteries, it achieves high-precision, real-time identification and early warning of battery faults. Compared with traditional fault diagnosis methods, this method demonstrates significant advantages in accuracy and response speed, can more effectively identify multiple fault types, reduce false positives and missed positives, and significantly improve the overall operational reliability of the system.

[0155] One of the key innovations of this invention is its unique feature extraction strategy. By deeply exploring the multi-dimensional characteristics of battery and grid status, including statistical, temporal, and frequency domain features, and combining them with advanced classification models (random forest and convolutional neural networks) to process major and complex failure modes respectively, the fault diagnosis model can more accurately identify the battery's health status and failure mode. This feature extraction method can capture subtle changes in battery operation and potential faults, thereby improving diagnostic accuracy.

[0156] Furthermore, the fault warning method of the present invention continuously monitors the real-time operation of the battery system based on an adaptively updated classification model, and issues a timely warning when the fault risk exceeds a set threshold. By combining real-time and historical data, the system not only improves the accuracy of warnings but also dynamically adjusts fault diagnosis strategies based on changing operating conditions, ensuring that the grid energy storage system operates efficiently and safely.

[0157] This method offers excellent scalability and adaptability, making it suitable for grid energy storage systems of varying types and sizes, and capable of meeting the development needs of future smart grids and microgrids. Through its modular design and standardized interfaces, the method can be easily integrated into existing grid management systems, supporting system expansion and upgrades, and possesses broad application prospects and market potential.

[0158] In summary, the intelligent battery fault diagnosis method of the present invention not only significantly improves the safety and economic benefits of grid energy storage systems, but also provides solid technical support for intelligent, data-driven power system management. This method has important engineering value and promotional significance in practical applications, effectively promoting the development of more efficient and intelligent power systems.

[0159] Example 2

[0160] Reference Figure 2-Figure 3, which is the second embodiment of the present invention.

[0161] The operating status of batteries in grid energy storage systems is influenced by multiple factors: the battery's electrical characteristics (voltage, current, and temperature), grid load and dispatch information, environmental conditions (such as temperature and humidity), and the battery's historical operating history (charge and discharge strategy, number of cycles, capacity decay, etc.). This invention constructs a stable and rich feature space at the data level, ensuring that key fault signals can be extracted from multi-source, multi-dimensional data.

[0162] Unified representation and time series alignment of multi-source data

[0163] By acquiring multi-source data from batteries, power grids, and environmental monitoring, the data is aligned using a unified time step and noise is removed through Kalman filtering to ensure data stability and reliability.

[0164] Set the discrete time step to Δt, and obtain corresponding observation data from the energy storage system and the power grid monitoring platform at each time t (t = 0, Δt, 2Δt). The basic data includes:

[0165] Battery electrical parameters: voltage V(t), current I(t), temperature T(t);

[0166] Grid characteristic parameters: load power P load (t), scheduling plan index S schcd (t);

[0167] Environmental conditions: external sensor data of temperature and humidity, represented by E cnv (t) Feature set historical operation records and statistics: cumulative number of cycles C cycle (t), specific charge and discharge curve characteristics The above dimensional information constitutes the original data vector at time t:

[0168] X(t)=[V(t),I(t),T(t),P load (t),S schcd (t),E cnv (t),C cycle (t),]←

[0169] Since different sensors and data sources may have different sampling frequencies, this paper uses linear interpolation to align time series. Linear interpolation uses the observation values ​​x(t1) and x(t2) at two adjacent time points to calculate the interpolated value x(t′) at the interpolation point t′:

[0170]

[0171] Ensure that all observations have consistent resolution at a uniform time step Δt.

[0172] In view of the possible noise signals in the sensor data, the present invention uses Kalman filtering technology to denoise the data. Kalman filtering combines the prior estimate and the observed value through the prediction-update iterative process to obtain the denoised data value.

[0173]

[0174] Where K(t) is the Kalman gain, expressed as:

[0175]

[0176] P(t|t-1) is the prediction covariance, and R is the measurement noise covariance. Kalman filtering effectively smooths data noise, making it suitable for dynamic and complex power grid operating environments.

[0177] Feature extraction: To reduce redundant information and improve the data's value in identifying the model, this paper uses feature extraction and dimensionality reduction techniques to extract key features related to fault signals and lifespan degradation based on the original data vector X(t). The extracted features are as follows:

[0178] Statistical features: Calculate statistical indicators such as mean, variance, skewness, and kurtosis for each data channel. These statistical features help capture the global characteristics of battery and grid status. Specifically, they include:

[0179] Mean: represents the average level of a data channel within a certain time window, used to capture long-term trends.

[0180]

[0181] Variance: Measures the degree of data fluctuation and reflects the stability of the battery or power grid.

[0182]

[0183] Skewness: Indicates the asymmetry of the data distribution. For battery health monitoring, skewness helps identify abnormal changes.

[0184]

[0185] Kurtosis: measures the “sharpness” of the data distribution and helps identify fault characteristics of high-frequency fluctuations

[0186]

[0187] Maximum and minimum values: reflect extreme battery behavior during charging and discharging, such as the risk of overcharging or overdischarging.

[0188] Max(X)=max{X(t1),X(t2),X(t n )}

[0189] Min(X)=min{X(t1),X(t2),X(t n )}

[0190] (2) Timing characteristics

[0191] The trend characteristics and autocorrelation coefficients of each data channel are calculated using the sliding window method to further extract dynamic features related to the fault. Specifically,

[0192] Trend: Indicates the changing trend of data within a certain time window.

[0193]

[0194] Autocorrelation: This measures the correlation of data over time, capturing periodic or recurring patterns. It is important for understanding the charge and discharge behavior of batteries during cycling.

[0195]

[0196] (3) Frequency domain characteristics

[0197] A fast Fourier transform (FFT) is used to convert the signal from the time domain to the frequency domain and extract the power spectral density within the frequency domain. These frequency domain features help identify high-frequency or low-frequency signals in the battery system under different operating conditions. In particular, they can identify high-frequency noise or abnormal vibration patterns when monitoring the battery's health status.

[0198] Spectral analysis: Obtain frequency domain characteristics by calculating the power spectrum density of battery voltage or current

[0199]

[0200] Where P(f) is the spectrum power, f is the frequency, t i is the sampling point of the time series.

[0201] After extracting effective features, the present invention uses principal component analysis (PCA) to reduce the dimensionality of the extracted high-dimensional features to ensure data compactness and discriminability. Suppose the original high-dimensional feature matrix is ​​Z = [z1, z2, ..., z m ], where each z i Represents the extracted i-th feature. By calculating the oblique variance matrix C:

[0202]

[0203] Where n is the number of samples, is the transpose of the feature matrix. Then, by performing eigenvalue decomposition on the covariance matrix C, the principal component z′ corresponding to the largest eigenvalue is extracted:

[0204]

[0205] Where w is the principal component weight vector, which represents the projection of the feature in the principal component space.

[0206] After dimensionality reduction, to ensure model efficiency and accuracy, this paper further screens the most meaningful features for battery health diagnosis and fault prediction through information gain evaluation and classification contribution analysis. Information gain is used to assess the correlation between each feature and the fault category, thereby selecting the key features that maximize the performance of the classification model.

[0207] In addition, by analyzing the contribution of features to different failure modes, the feature set is further optimized, redundant features are removed, and features most closely related to health status and failure mode are retained. The resulting reduced-dimensional feature set Z′ will serve as the input to the diagnosis and prediction model.

[0208] Dynamic data update mechanism

[0209] In order to adapt to the dynamic operating environment of the grid energy storage system, the present invention introduces a sliding window update mechanism in the feature extraction process. For the window length τ, the sliding feature φ win Calculated as:

[0210] φ win (X(t))=φ(X(t-τ),X(t-τ+1),X(t)),

[0211] Where X(t) is the raw data vector at the current moment, and X(t-τ), X(t-τ+1), and X(t) represent the data from the past τ moments. Within the window, the characteristic function φ calculates the comprehensive characteristics of the data within that time period, ensuring that the extracted features always reflect the latest dynamic changes of the battery under the current operating conditions.

[0212] Through the above-mentioned data fusion and feature extraction strategy, the present invention realizes multi-source integration, noise processing and feature optimization of raw data, and provides a stable, reliable and engineering-valuable input for the intelligent fault diagnosis model.

[0213] Intelligent fault diagnosis mechanism

[0214] Based on a data-driven intelligent classification model, this paper proposes a mechanism for real-time diagnosis of battery faults in grid energy storage systems. This mechanism extracts and analyzes multi-source feature data, combined with a posterior probability judgment model, to accurately identify and provide early warnings for various battery fault types. This method is particularly suitable for grid energy storage systems, where the battery operating environment is complex and operating conditions are highly variable.

[0215] Fault type definition and model input

[0216] Battery failures may involve various types. To facilitate diagnosis, the following failure sets are defined:

[0217] F=

[0218] {Overcharge, over-discharge, overheating, increased internal resistance, capacity attenuation, short circuit, damage, temperature out of control, unbalanced charging, the group indicates the possible fault types and normal status.

[0219] This set includes common fault types (main fault modes) as well as more complex fault types (complex fault modes) that are difficult to identify using traditional machine learning methods. The specific fault types are as follows:

[0220] Main failure modes (identified by random forest):

[0221] Overcharge: The battery voltage exceeds the maximum safe voltage, which may cause fire or explosion.

[0222] Overdischarge: The battery voltage drops below the safety threshold, which may cause battery damage.

[0223] Overheating: The battery temperature is too high, resulting in thermal runaway or performance degradation.

[0224] Increased internal resistance: The internal resistance increases due to battery aging or damage, affecting the charging and discharging efficiency.

[0225] Capacity fade: The gradual reduction in battery capacity, resulting in a decrease in available energy.

[0226] Complex failure modes (identified by convolutional neural networks (CNN)):

[0227] Short circuit: A short circuit occurs within the battery or between cells, causing the battery to overheat rapidly.

[0228] Damage: Physical damage or cracks occur inside the battery, resulting in a decrease in battery performance.

[0229] Temperature runaway (thermal runaway): Abnormally high battery temperature may trigger a thermal runaway reaction.

[0230] Charge imbalance: The charge between multiple battery modules is unbalanced, resulting in overcharge or over-discharge of some modules.

[0231] Mass failure: The performance of the entire battery pack or system is degraded due to multiple single cell failures or system management problems.

[0232] At time t, the feature vector Z′(t) obtained after feature extraction and dimensionality reduction is input into the fault diagnosis classifier f diag , calculate the posterior probability of each fault type:

[0233] p(f|Z′(t))=f diag (Z′(t)),f∈F

[0234] The classifier constructs a posterior probability distribution based on the training data and uses different machine learning algorithms for classification learning. Specifically, a random forest is used to classify the main fault modes, and a convolutional neural network (CNN) is used to further analyze and diagnose complex fault modes.

[0235] Fault judgment and early warning mechanism

[0236] According to the posterior probability p(f|Z′(t)) output by the classifier, fault judgment is performed by maximizing the criterion:

[0237]

[0238] where f * (t) is the fault type with the highest probability in the diagnosis results. The conditions for fault judgment are:

[0239] 1. Determine the fault type f * (t)≠normal

[0240] 2. Corresponding posterior probability p(f * (t)|Z′(t))>η, where η is the threshold for fault judgment.

[0241] If both of the above conditions are met, the system will issue a fault warning signal. The warning results can be directly fed back to the battery management system (BMS) and the power grid dispatch center to support timely operation and maintenance response.

[0242] Adaptive fault model update mechanism

[0243] The present invention introduces an online update mechanism to adapt to the dynamic working conditions of the energy storage system. The parameter set of the classifier is defined as Θ diag , the classification model is optimized through incremental learning method. The update rule is:

[0244]

[0245] in The increment adjusted according to the latest data is calculated by combining the new observation data Z′(t) and the prediction error of the model:

[0246]

[0247] Where α is the learning rate, is the classification error loss function, y t is the real fault label (such as obtained through historical verification).

[0248] Model evaluation and validation

[0249] In order to ensure the effectiveness and reliability of the intelligent fault diagnosis mechanism proposed in the present invention, the present invention uses the following evaluation method to verify the classification model:

[0250] Evaluation Metrics:

[0251] (1) Precision

[0252] Precision is used to measure the proportion of samples predicted by the model as faulty that are actually faulty, focusing on controlling the false positive rate. The calculation formula is:

[0253]

[0254] Among them, TP is a true positive example and FP is a false positive example.

[0255] (2) Recall

[0256] Recall is used to measure the model's ability to identify actual fault samples, focusing on the control of missed detection rate. The calculation formula is:

[0257] Among them, TP is a true positive example and FN is a false negative example.

[0258] (3) F1 Score

[0259] The F1 score is the harmonic mean of precision and recall, taking into account both false positives and false negatives, and is suitable for scenarios with unbalanced data categories. The calculation formula is:

[0260]

[0261] (4) Confusion Matrix

[0262] The confusion matrix helps analyze the model's performance on different fault types by showing the model's prediction results under various fault modes. The confusion matrix consists of:

[0263] True Positives (TP): The number of samples that are predicted to be faulty and are actually faulty.

[0264] True Negatives (TN): The number of samples that are predicted to be normal and are actually normal.

[0265] False Positives (FP): The number of samples predicted to be faulty but actually normal (false positives).

[0266] False negatives (FN): The number of samples that are predicted to be normal but are actually faulty (false negatives).

[0267] The confusion matrix helps developers identify how the model performs under different fault types, making it easier to adjust and optimize.

[0268] The above evaluation indicators can comprehensively evaluate the accuracy, stability, and reliability of the intelligent fault diagnosis method of the present invention in practical applications, thereby ensuring that the model can effectively identify battery faults and reduce false positives and missed negatives.

[0269] Example 3

[0270] This embodiment also provides a computer device applicable to a battery intelligent fault diagnosis method applicable to a power grid energy storage system, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement a forced oscillation detection and positioning method for a distribution network as proposed in the above embodiment.

[0271] This embodiment further provides a storage medium storing a computer program, which, when executed by a processor, implements a forced oscillation detection and positioning method for a distribution network as proposed in the above embodiment.

[0272] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0273] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0274] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0275] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0276] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0277] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A battery intelligent fault diagnosis method applicable to a power grid energy storage system, characterized in that: The method includes acquiring multi-source monitoring data of a target energy storage system, and fusing the multi-source monitoring data to form first monitoring data; Presetting a first fault diagnosis model, and inputting the first monitoring data into the first fault diagnosis model; Intelligent battery fault diagnosis is performed based on the output of the first fault diagnosis model.

2. The battery intelligent fault diagnosis method applicable to the power grid energy storage system according to claim 1, characterized in that: The multi-source monitoring data includes: Battery electrical parameter data, including voltage, current and temperature parameters; Grid characteristic data, including load power and dispatch plan indicators; Environmental condition data, the environmental condition data including ambient temperature and humidity; Historical operation record data, wherein the historical operation record data includes the accumulated number of cycles and the charge and discharge curve.

3. The battery intelligent fault diagnosis method applicable to the power grid energy storage system according to claim 2, characterized in that: The data fusion to form the first monitoring data includes: Performing unified time step alignment on the multi-source monitoring data; Extracting statistical features of the multi-source monitoring data, wherein the statistical features include mean, variance, skewness and kurtosis; Extracting time series features of the multi-source monitoring data, wherein the time series features include trend features and autocorrelation coefficients; The frequency domain features of the multi-source monitoring data are extracted, wherein the frequency domain features include power spectral density.

4. The battery intelligent fault diagnosis method applicable to the power grid energy storage system according to claim 3, characterized in that: The first fault diagnosis model comprises: Random forest model, used to diagnose the main fault modes of overcharge, over-discharge, overheating, increased internal resistance and capacity attenuation; Convolutional neural network models for diagnosing complex failure modes such as short circuits, damage, temperature runaway, charge imbalance, and group failures; The loss function of the first fault diagnosis model includes positioning loss, confidence loss and classification loss.

5. The battery intelligent fault diagnosis method applicable to the power grid energy storage system according to claim 4, characterized in that: The intelligent diagnosis of battery faults according to the output of the first fault diagnosis model comprises: Calculate the posterior probability of each fault type; When the posterior probability of a certain fault type is greater than the preset threshold, it is determined to be this type of fault; A warning signal is sent to the battery management system based on the fault judgment result.

6. The battery intelligent fault diagnosis method applicable to the power grid energy storage system according to claim 5, characterized in that: Also includes a dynamic update mechanism: Calculate the incremental update values ​​of the model parameters based on the newly acquired monitoring data and the model prediction error; Updating the parameters of the first fault diagnosis model online; Redo troubleshooting based on the updated model.

7. The battery intelligent fault diagnosis method applicable to the power grid energy storage system according to claim 6, characterized in that: Also includes model evaluation mechanism: Calculate the precision, recall, and F1 score of the model on the validation set; Generate confusion matrix to analyze the diagnostic accuracy of various faults; When the model evaluation index is lower than the preset standard, the model retraining process is triggered.

8. A battery intelligent fault diagnosis system applicable to a power grid energy storage system, based on the battery intelligent fault diagnosis method applicable to a power grid energy storage system according to any one of claims 1 to 7, characterized in that: It also includes a data acquisition module, which acquires multi-source monitoring data of the target energy storage system and performs data fusion on the multi-source monitoring data to form first monitoring data; A preset module, which presets a first fault diagnosis model and inputs the first monitoring data into the first fault diagnosis model; An output module performs intelligent battery fault diagnosis according to the output of the first fault diagnosis model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the battery intelligent fault diagnosis method applicable to the power grid energy storage system according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the battery intelligent fault diagnosis method applicable to a power grid energy storage system according to any one of claims 1 to 7 are implemented.

Images