Accident reason determination method and device, storage medium and electronic equipment

Abstract

The embodiment of the invention provides an accident cause determination method and device, a storage medium and electronic equipment, and the method comprises the steps: extracting feature data in time series data corresponding to an accident event; inputting the feature data into a target anomaly detection model, so that the target anomaly detection model identifies abnormal data in the time series data based on the feature data; intercepting the time sequence data according to abnormal data to determine part of time sequence data; the partial time series data is analyzed to determine an accident cause of the accident event.

Description

Technical Field

[0001] The present application relates to the field of data analysis, and more specifically, to a method and device for determining the cause of an accident, a storage medium, and an electronic device. Background Art

[0002] In the operation and management of complex systems, it is crucial to accurately identify and analyze the causes of accidents. Traditional accident recall and analysis methods often rely on post-incident manual review and simple threshold-based alarm systems. This method has significant limitations when facing massive time series data, such as:

[0003] After an accident occurs, manually reviewing a large amount of time series data to trace the accident process and determine the cause is a time-consuming and error-prone process. Especially when multiple factors such as electrical parameters, environmental conditions and equipment status need to be considered comprehensively, the complexity and difficulty of analysis increases dramatically.

[0004] With respect to the problem of low efficiency in determining the cause of an accident in the prior art, no effective solution has been proposed so far.

[0005] Therefore, it is necessary to improve the related technology to overcome the above-mentioned defects in the related technology. Summary of the invention

[0006] The embodiments of the present application provide a method and device for determining the cause of an accident, a storage medium, and an electronic device, so as to at least solve the problem of low efficiency in determining the cause of an accident in the prior art.

[0007] According to one embodiment of the present application, a method for determining the cause of an accident is provided, including: extracting feature data from time series data corresponding to an accident event; inputting the feature data into a target anomaly detection model so that the target anomaly detection model identifies abnormal data in the time series data based on the feature data; intercepting the time series data according to the abnormal data to determine part of the time series data; and analyzing the part of the time series data to determine the cause of the accident event.

[0008] In an exemplary embodiment, characteristic data related to an accident event is extracted from time series data, including at least one of the following: extracting statistical characteristics of each variable in the time series data, wherein the statistical characteristics include: maximum value, minimum value, mean, standard deviation; calculating the rate of change of each variable in the time series data; identifying the correlation between any two variables in the time series data; determining the frequency domain characteristics of the time series data based on the Fourier exchange method, wherein the characteristic data includes at least one of the following: the statistical characteristics of each variable, the rate of change of each variable, the correlation between different variables, and the frequency domain characteristics of the time series data.

[0009] In an exemplary embodiment, identifying correlations between two variables in the time series data includes:

[0010] The correlation r between two variables in the time series data is identified by the following formula:

[0011] Among them, i represents the index variable, r represents the correlation, n represents the number of samples, and x i represents the i-th observation value of variable X, y i represents the i-th observation value of variable Y, represents the sample mean of variable X, represents the sample mean of variable Y.

[0012] In an exemplary embodiment, before inputting the feature data into a target anomaly detection model so that the target anomaly detection model identifies anomaly data in the time series data based on the feature data, the method further includes: dividing the data set into a training set and a test set, and training the anomaly detection model through the training set;

[0013] The reconstruction error of the anomaly detection model is evaluated using the test set through a reconstruction error distribution algorithm, wherein the reconstruction error distribution algorithm is: E recon represents the reconstruction error, x train represents the input data of the anomaly detection model, Reconstructed data representing the input data reconstructed by the anomaly detection model;

[0014] When the reconstruction error is less than a preset reconstruction error, it is determined that the training of the anomaly detection model is completed, and the trained anomaly detection model is determined as the target anomaly detection model.

[0015] In an exemplary embodiment, the anomaly detection model includes an encoder module and a decoder module, wherein the encoder module is expressed as follows:

[0016] z=f encode (x;θ encode ), where z represents the output data of the encoder, f encode represents the encoder function, which converts high-dimensional input data x into low-dimensional representation z, where x represents the input data and θ encode represents the set of trainable parameters of the encoder;

[0017] The expression of the decoder module is as follows:

[0018] in, represents the reconstructed data of the decoder;decode Represents the decoder function, which reconstructs the low-dimensional representation z back to high-dimensional data θ decode represents a set of trainable parameters of the decoder.

[0019] In an exemplary embodiment, the feature data is input into a target anomaly detection model so that the target anomaly detection model identifies the abnormal data in the time series data based on the feature data, including: determining a preset threshold value T by the following formula:

[0020] T=percentile(E recon (X), A), where percentile represents the percentile function in statistics, E recon (X) represents the combination of reconstruction errors on the feature data X, X represents the feature data, and A represents the Ath percentile; determine the target reconstruction error corresponding to each feature data, and determine the size relationship between the target reconstruction error and the preset threshold; when the size relationship indicates that any target reconstruction error is greater than the preset threshold, determine that the feature data corresponding to any target reconstruction error is the abnormal data.

[0021] In an exemplary embodiment, analyzing the partial time series data to determine the cause of the accident event includes: determining target feature data of the partial time series data, wherein the target feature data includes at least one of the following: data trend of each variable of the partial time series data, fluctuation information of each variable of the partial time series data, and correlation between two variables of the partial time series data; determining the cause of the accident event based on the target feature data.

[0022] According to another embodiment of the present application, a device for determining the cause of an accident is provided, including: an extraction module for extracting feature data from time series data corresponding to an accident event; an identification module for inputting the feature data into a target anomaly detection model so that the target anomaly detection model identifies abnormal data in the time series data based on the feature data; an interception module for intercepting the time series data according to the abnormal data to determine part of the time series data; and an analysis module for analyzing the part of the time series data to determine the cause of the accident event.

[0023] According to another embodiment of the present application, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when run.

[0024] According to another embodiment of the present application, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0025] According to another embodiment of the present application, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0026] Through the present application, feature data is extracted from the time series data corresponding to the accident event; the feature data is input into the target anomaly detection model so that the target anomaly detection model identifies the anomaly data in the time series data based on the feature data; the time series data is intercepted according to the anomaly data to determine part of the time series data; the part of the time series data is analyzed to determine the cause of the accident event. In the embodiment of the present application, the accuracy of determining the cause of the accident is improved through feature extraction, anomaly detection, abnormal data determination, and in-depth analysis, and the efficiency is significantly improved, thereby avoiding the waste of resources and time caused by manual analysis and analysis of a large amount of useless data in traditional methods. Therefore, the problem of low efficiency in determining the cause of the accident can be solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0028] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0029] Figure 1 It is a hardware structure block diagram of a computer device of a method for determining the cause of an accident in an embodiment of the present application;

[0030] Figure 2 is a flow chart of a method for determining the cause of an accident according to an embodiment of the present application;

[0031] Figure 3 It is a structural block diagram of a device for determining the cause of an accident according to an embodiment of the present application. DETAILED DESCRIPTION

[0032] The embodiments of the present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0033] It should be noted that the terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0034] The method embodiments provided in the embodiments of the present application can be executed in a computer device or a similar computing device. Taking running on a computer device as an example, Figure 1 1 is a hardware structure block diagram of a computer device for a method for determining the cause of an accident in an embodiment of the present application. Figure 1 As shown, the computer device may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the above-mentioned computer device may also include a transmission device 106 and an input and output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above-mentioned computer device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0035] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the method for determining the cause of the accident in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories can be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0036] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer device. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0037] In this embodiment, a method for determining the cause of an accident is provided. Figure 2 is a flow chart of a method for determining the cause of an accident according to an embodiment of the present application, such as Figure 2 As shown, the process includes the following steps:

[0038] Step S202, extracting feature data from the time series data corresponding to the accident event;

[0039] Extracting feature data related to accidents from the original time series data is equivalent to a preliminary screening and dimensionality reduction of the data, which reduces the complexity of data processing and makes subsequent analysis more focused and efficient. The extraction of feature data is based on the understanding of time series data. Statistical features such as maximum value, minimum value, mean, standard deviation, and dynamic features such as rate of change and correlation between variables can be selected to help locate abnormal signals more accurately.

[0040] Step S204, inputting the feature data into a target anomaly detection model, so that the target anomaly detection model identifies abnormal data in the time series data based on the feature data;

[0041] The extracted feature data is input into the pre-trained target anomaly detection model, and the model's recognition ability is used to automatically filter out abnormal data in the time series data. Target anomaly detection models, such as deep learning models such as autoencoders, can learn the patterns of normal operating data to accurately identify anomalies, avoiding the subjectivity and limitations of traditional threshold methods and improving the accuracy and efficiency of anomaly detection.

[0042] Step S206, intercepting the time series data according to the abnormal data to determine part of the time series data;

[0043] Once the abnormal data is identified, the time series data can be intercepted based on this to determine the specific time window for analysis. This step significantly narrows the scope of analysis, focusing attention on the period before and after the accident, avoiding wasteful analysis of data from irrelevant time periods, and greatly improving analysis efficiency.

[0044] Step S208: Analyze the partial time series data to determine the cause of the accident event.

[0045] By analyzing the intercepted part of the time series data, we can have a deeper understanding of the nature of the abnormal data, including the trend of change, the relationship with other variables, etc., so as to determine the root cause of the accident. We can make a comprehensive judgment from multiple angles such as electrical, environmental, and equipment status, avoiding the inefficiency caused by one-sided analysis.

[0046] Through the above steps, feature data in the time series data corresponding to the accident event is extracted; the feature data is input into the target anomaly detection model so that the target anomaly detection model identifies the anomaly data in the time series data based on the feature data; the time series data is intercepted according to the anomaly data to determine part of the time series data; the part of the time series data is analyzed to determine the cause of the accident event. In the embodiment of the present application, the accuracy of determining the cause of the accident is improved through feature extraction, anomaly detection, abnormal data determination, and in-depth analysis, and the efficiency is significantly improved, thereby avoiding the waste of resources caused by a large amount of useless data analysis in traditional methods. Therefore, the problem of low efficiency in determining the cause of the accident can be solved.

[0047] In an exemplary embodiment, characteristic data related to an accident event is extracted from time series data, including at least one of the following: extracting statistical characteristics of each variable in the time series data, wherein the statistical characteristics include: maximum value, minimum value, mean, standard deviation; calculating the rate of change of each variable in the time series data; identifying the correlation between any two variables in the time series data; determining the frequency domain characteristics of the time series data based on the Fourier exchange method, wherein the characteristic data includes at least one of the following: the statistical characteristics of each variable, the rate of change of each variable, the correlation between different variables, and the frequency domain characteristics of the time series data.

[0048] By calculating the maximum, minimum, mean, and standard deviation of each variable, you can quickly understand the distribution and fluctuation range of the data. These statistics can help identify abnormal highs or lows in the data, and whether the data is stable or has large fluctuations. In accident analysis, any abnormal value that deviates from the normal mean or standard deviation range may be a harbinger of an accident, while a sudden change in the maximum or minimum value indicates a possible fault point or an unstable state of the system.

[0049] Analyzing the rate of change of each variable in time series data over time can capture changes in data trends, especially rapid changes or mutations. In energy storage systems or other related systems, sudden changes in parameters are often a direct manifestation of equipment failure, operational errors, or environmental changes. Rate of change analysis can reveal the dynamic behavior of parameters and plays an important role in identifying accident precursors and diagnosing faults.

[0050] By identifying the correlation between two variables in time series data, complex interactions within the system can be discovered. In many cases, the occurrence of an accident is not the result of a single variable anomaly, but a chain reaction of multiple variables interacting with each other. For example, an increase in temperature and an increase in current may occur at the same time, which may indicate the risk of battery overheating or short circuit. Using methods such as the Pearson correlation coefficient can help analysts understand the relationship between variables and provide a basis for in-depth analysis of the cause of the accident.

[0051] Based on the Fourier transform method, time domain signals can be converted to the frequency domain for analysis, which helps to identify periodic patterns and frequency components in the signal. In power systems, abnormal changes in frequency components are often related to system oscillations, harmonic interference, or equipment failures. Frequency domain analysis can reveal patterns that are not easily perceived in the time domain, providing other perspectives for anomaly detection.

[0052] Through the above feature extraction method, a more comprehensive and in-depth anomaly detection model can be constructed, which can effectively identify abnormal patterns related to accident events. It not only improves the accuracy of anomaly detection, but also speeds up the speed of accident recall, making accident analysis more efficient in terms of time and computing resources, and providing strong support for timely corrective measures and prevention of future accidents.

[0053] In an exemplary embodiment, identifying correlations between two variables in the time series data includes:

[0054] The correlation r between two variables in the time series data is identified by the following formula:

[0055] Among them, i represents the index variable, r represents the correlation, n represents the number of samples, and x i represents the i-th observation value of variable X, y i represents the i-th observation value of variable Y, represents the sample mean of variable X, represents the sample mean of variable Y.

[0056] The above correlation can be understood as the Pearson correlation coefficient, which is a statistic that measures the degree of linear correlation between two variables. The value of the Pearson correlation coefficient ranges from -1 to +1, where:

[0057] +1 indicates a perfect positive correlation, meaning that when one variable increases, the other variable tends to increase in the same proportion.

[0058] 0 means there is no linear correlation, and there is no obvious linear relationship between the changes in the variables.

[0059] -1 indicates a perfect negative correlation, meaning that when one variable increases, the other tends to decrease in the same proportion.

[0060] In time series data analysis, by calculating the Pearson correlation coefficient between variables, the interdependence between variables can be identified. This is particularly important in energy storage system accident recall, because accidents are often caused by the interaction of multiple variables. For example, an abnormal increase in temperature may be associated with a sudden increase in current, and this correlation may indicate battery overheating or abnormal electrical activity in the system. By detecting this correlation, the context of the accident can be more fully understood, which helps to locate the root cause of the accident.

[0061] The formula works by normalizing the average of the products of the two variables across all samples by calculating the average deviation (the numerator) of the products, and then dividing by the square root of the product of the sum of the squares of their respective deviations (the denominator). This normalization process ensures that the value of the correlation coefficient is not affected by the dimension or magnitude of the variable, making correlations between different types of variables (such as voltage and temperature) directly comparable.

[0062] In practical applications, by calculating the Pearson correlation coefficient between all pairs of variables of interest in the time series data, a correlation matrix can be constructed to further analyze which pairs of variables have the strongest correlation, thus providing clues for accident analysis. For example, if it is found that the correlation between current and temperature suddenly increases around a certain time point, then this may be an abnormal pattern worthy of attention, indicating that an event related to current and temperature changes occurred near that time point, which may be a precursor or direct cause of the accident.

[0063] In the embodiment of the present application, by identifying the Pearson correlation coefficient between two variables in the time series data, not only can the linear relationship between the variables be discovered, but also the interaction mechanism within the system can be understood, which is of great value for abnormal detection and accident recall of complex systems such as energy storage systems. It can reveal the coordinated changes of variables under specific conditions, provide a scientific basis for preventing and handling accidents, and thus improve the safety and reliability of the system.

[0064] In an exemplary embodiment, before inputting the feature data into a target anomaly detection model so that the target anomaly detection model identifies anomaly data in the time series data based on the feature data, the method further includes: dividing the data set into a training set and a test set, and training the anomaly detection model through the training set;

[0065] The reconstruction error of the anomaly detection model is evaluated using the test set through a reconstruction error distribution algorithm, wherein the reconstruction error distribution algorithm is: E recon represents the reconstruction error, x train represents the input data of the anomaly detection model, Reconstructed data representing the input data reconstructed by the anomaly detection model;

[0066] When the reconstruction error is less than a preset reconstruction error, it is determined that the training of the anomaly detection model is completed, and the trained anomaly detection model is determined as the target anomaly detection model.

[0067] In the embodiment of the present application, the data set is divided into a training set and a test set (or validation set). The training set is used for model learning and parameter adjustment, while the test set or validation set is used to independently evaluate the performance of the model.

[0068] Use the training set data to train anomaly detection models, such as autoencoders. The training goal of the model is to minimize the difference between the input data and the reconstructed data. The autoencoder encodes the input data into a low-dimensional space through forward propagation, and then decodes the low-dimensional representation back to the high-dimensional space of the original data through backward propagation. The training process adjusts the parameters of the encoder and decoder to minimize the reconstruction error, which is the difference between the model output and the input.

[0069] The reconstruction error of the model is evaluated by using the test set or validation set data. The reconstruction error reflects the performance of the model on unseen data. If the reconstruction error of the model on the validation set is less than the preset threshold, that is, the model can reconstruct normal data with a sufficiently low error, then the model is considered trained and can be used for abnormal data detection. The preset reconstruction error threshold is set according to business needs and the specific goals of anomaly detection. It is a balance point to ensure that the model is neither too loose and regards all data as normal; nor too strict and regards all small fluctuations as abnormal.

[0070] Through the embodiments of the present application, not only a model that can distinguish between normal data and abnormal data is trained, but also the generalization ability of the model is ensured, that is, it can effectively identify anomalies on unseen data. This provides a solid foundation for anomaly detection in time series data analysis. In practical applications, the target anomaly detection model can be applied to new time series data in real time or regularly to automatically identify potential abnormal events, thereby providing early warnings and taking timely measures to prevent the occurrence or expansion of accidents.

[0071] In an exemplary embodiment, the anomaly detection model includes an encoder module and a decoder module, wherein the encoder module is expressed as follows:

[0072] z=f encode (x;θ encode ), where z represents the output data of the encoder, f encode represents the encoder function, which converts high-dimensional input data x into low-dimensional representation z, where x represents the input data and θ encode represents the set of trainable parameters of the encoder;

[0073] The expression of the decoder module is as follows:

[0074] in, represents the reconstructed data of the decoder; decode Represents the decoder function, which reconstructs the low-dimensional representation z back to high-dimensional data θ decpde represents a set of trainable parameters of the decoder.

[0075] The autoencoder consists of an encoder module and a decoder module. The encoder module converts the high-dimensional raw input data into a low-dimensional representation. The encoder module involves nonlinear transformations, the purpose of which is to learn the essential characteristics of the data while removing redundant information. The decoder module reconstructs the low-dimensional representation output by the encoder back into high-dimensional data to restore the information of the original data as much as possible.

[0076] In an exemplary embodiment, the feature data is input into a target anomaly detection model so that the target anomaly detection model identifies the abnormal data in the time series data based on the feature data, including: determining a preset threshold value T by the following formula:

[0077] T=percentile(E recon (X), A), where percentile represents the percentile function in statistics, E recon (X) represents the combination of reconstruction errors on the feature data X, X represents the feature data, and A represents the Ath percentile; determine the target reconstruction error corresponding to each feature data, and determine the size relationship between the target reconstruction error and the preset threshold; when the size relationship indicates that any target reconstruction error is greater than the preset threshold, determine that the feature data corresponding to any target reconstruction error is the abnormal data.

[0078] The preset threshold is determined by the formula, where:

[0079] E recon (X) represents the combination of reconstruction errors on feature data X, that is, the error set between all training data after reconstruction by the autoencoder and the original data.

[0080] X represents feature data, which is time series data after preprocessing and feature extraction.

[0081] A represents the Ath percentile, which is artificially set to distinguish the boundary between normal data and abnormal data. For example, choosing the 95th percentile as the threshold means that the reconstruction error of 95% of normal data will be less than T, while the error beyond this threshold may come from abnormal data.

[0082] Through the percentile function in statistics, a threshold of reconstruction error can be determined, which distinguishes the reconstruction error distribution of normal data and potential abnormal data. The selection of an appropriate percentile is based on the understanding of the normal operating range and the trade-off between the risk of false positives and false negatives.

[0083] For each feature data input to the autoencoder, the model will output a reconstruction data. By comparing the original feature data with the reconstructed data, the target reconstruction error can be calculated.

[0084] Compare the target reconstruction error with the preset threshold. If any target reconstruction error is greater than the preset threshold, the corresponding feature data is considered abnormal data. This means that the reconstruction performance of the data point in the feature space deviates from the normal data pattern, which may indicate the occurrence of an accident or abnormal event.

[0085] Through the above steps, the energy storage system accident recall method based on time series data can accurately and timely identify abnormal data, and then conduct further accident analysis and recall, which is of great significance for preventing accidents and improving the safety and reliability of the system.

[0086] In an exemplary embodiment, analyzing the partial time series data to determine the cause of the accident event includes: determining target feature data of the partial time series data, wherein the target feature data includes at least one of the following: data trend of each variable of the partial time series data, fluctuation information of each variable of the partial time series data, and correlation between two variables of the partial time series data; determining the cause of the accident event based on the target feature data.

[0087] Target characteristic data, including at least one of the following:

[0088] Data trend: Analyze the changing trend of each variable over time. In energy storage systems, abnormal trends in key parameters such as voltage, current, and temperature may be a direct indication of an accident. For example, a rapid rise in battery temperature may indicate thermal runaway, while a sudden drop in current may be associated with an open circuit or short circuit event.

[0089] Fluctuation information: Check the fluctuation of each variable, including the amplitude and frequency of the fluctuation. Abnormal fluctuation patterns, such as sudden high-frequency fluctuations or large fluctuations beyond the normal range, are usually related to specific faults or abnormal conditions. Analysis of fluctuation information can help identify the instability of system status and potential failures.

[0090] Correlation between variables: Identify the correlation between two variables in some time series data. In complex systems, the interactions between multiple variables are often more complex. By analyzing the correlation between variables, some patterns or associations that cannot be detected when directly observing a single variable can be found. For example, an abnormal positive correlation between battery temperature and current may point to an overheating problem, while a change in the correlation between voltage and device status may be related to device failure.

[0091] After extracting the target feature data, analysts or automatic systems will infer the cause of the accident based on this feature information. The specific steps include: observing data trends and fluctuations, identifying any sudden changes or unusual behaviors that may be related to a sudden deterioration of the system status or a specific operational error.

[0092] Analyze the changes in the correlation between variables. For example, under normal operating conditions, there are stable linear or nonlinear relationships between certain variables, but before and after an accident, these relationships may become unstable or abnormal, which helps to locate the root cause of the accident.

[0093] Combining domain knowledge and historical data, target feature data is compared with features of previous accidents to identify similar patterns, thereby narrowing down the possible causes of the accident.

[0094] It is worth noting that when analyzing partial time series data, it is important to focus on the data before and after the accident, because these data usually contain the direct triggering factors of the accident and the changes in system status affected by it. By comparing the characteristics of normal operation data and accident data, the pattern and cause of the accident can be more clearly identified.

[0095] In order to better understand the process of the above-mentioned method for determining the cause of the accident, the implementation method flow of the above-mentioned method for determining the cause of the accident is described below in combination with an optional embodiment, but it is not used to limit the technical solution of the embodiment of the present application.

[0096] In this embodiment, a method for determining the cause of an accident is provided, which comprises the following steps:

[0097] S1, collect a large amount of time series data through the monitoring components in the energy storage system;

[0098] Among them, time series data includes voltage, current, power, temperature, humidity, equipment status, and operation records;

[0099] In this example, the monitoring components include a power monitoring sensor module, an environment monitoring sensor module, and a state monitoring sensor module;

[0100] The power monitoring sensor module is used to measure the voltage at the battery pack, busbar, and grid interface, and the current, power input, and output in the charge and discharge circuit;

[0101] The environmental monitoring sensor module is used to monitor the ambient temperature and humidity of the energy storage system;

[0102] The state monitoring sensor module is used to monitor the state of circuit breakers and relay switch equipment and record the history of manual or automatic operations.

[0103] S2, preprocessing the collected time series data;

[0104] In this example, preprocessing includes data cleaning, missing value processing, and outlier detection;

[0105] Wherein, the data cleaning is used to remove or correct errors and inconsistencies in the data;

[0106] The missing value processing is used to handle the situation where some observations are not recorded in the time series data set after data cleaning;

[0107] The outlier detection is used to monitor data points that deviate significantly from the majority of data after missing value processing.

[0108] S3, extracting features related to the accident from the preprocessed time series data (equivalent to the feature data in the above embodiment);

[0109] Among them, relevant features include mutation points of electrical parameters, abnormal temperature rise, and abnormal changes in equipment status;

[0110] In this example, the specific steps for extracting features related to accidents are:

[0111] S3.1, extract statistical features from time series data;

[0112] Among them, statistical characteristics include maximum value, minimum value, mean value, and standard deviation;

[0113] S3.2. Calculate the rate of change of variables over time;

[0114] S3.3, identify the correlations between different variables;

[0115] In this example, the correlation is identified based on the Pearson correlation coefficient, which is specifically:

[0116] Where i represents the index variable; r represents the Pearson correlation coefficient; n represents the number of samples; x i represents the i-th observation value of variable X; y i represents the i-th observation value of variable Y; represents the sample mean of variable X; represents the sample mean of variable Y.

[0117] Specifically, the Pearson correlation coefficient, also known as the Pearson product-moment correlation coefficient (PPMCC or PCCs), is a statistic used to measure the linear correlation between two variables. Its value ranges from -1 to +1, where:

[0118] +1 indicates a perfect positive correlation: when one variable increases, the other tends to increase in the same proportion.

[0119] 0 means no linear correlation: there is no clear linear relationship between the changes in the variables.

[0120] -1 indicates a perfect negative correlation: when one variable increases, the other tends to decrease in the same proportion.

[0121] S3.4. Use the Fourier exchange method to analyze the frequency domain characteristics of the data.

[0122] S4. Establish an anomaly detection model based on the autoencoder algorithm to identify accident events in the data;

[0123] Specifically, Autoencoder (AE) is an unsupervised learning method, which is usually used for tasks such as dimensionality reduction, feature learning, and anomaly detection. It learns to compress (encode) and decompress (decode) data by trying to copy the input to the output.

[0124] In this example, the specific process of building an anomaly detection model based on the autoencoder algorithm is as follows:

[0125] S4.1. Divide the data set into training set and test set;

[0126] S4.2. Use the training set data to train the anomaly detection model;

[0127] In this example, the anomaly detection model includes an encoder module and a decoder module;

[0128] Among them, the encoder module:

[0129] z=f encode (x;θ encode ), where z represents the encoder output; f encode represents the encoder function, which converts the high-dimensional input x into a low-dimensional representation z; x represents the input data; θ encode represents the set of trainable parameters of the encoder;

[0130] Decoder module:

[0131] in, Represents the data reconstructed by the decoder; f decode Represents the decoder function, which reconstructs the low-dimensional representation z back to high-dimensional data θ decode Represents the set of trainable parameters of the decoder.

[0132] S4.3. Evaluate the anomaly detection model using the validation set by reconstructing the error distribution algorithm.

[0133] Among them, the reconstruction error distribution algorithm is specifically as follows:

[0134] Among them, E recon represents the reconstruction error;

[0135] Specifically, the reconstruction error distribution algorithm is often used to evaluate the performance of autoencoder models, especially in anomaly detection tasks. This algorithm identifies anomalies by calculating the difference between the input data and its reconstructed data.

[0136] S4.4. After the anomaly detection model is trained, detect anomalies in new data.

[0137] In this example, the specific steps for detecting outliers in new data are:

[0138] The learning goal of the autoencoder is to minimize the difference between the input data and the reconstructed data, using the mean squared error as the loss function:

[0139] Where L represents the loss function; x J represents the input data point; represents the reconstructed data point; m represents the number of data points; J represents the index of the data point;

[0140] Set a threshold T to distinguish normal data from abnormal data:

[0141] T=percentile(E recon (X), 95); where percentile represents the percentile function in statistics; E recon (X) represents the combination of reconstruction errors on the new dataset X; X represents the new dataset; 95 represents the 95th percentile;

[0142] When E recon (x)>T, then x is considered to be abnormal data, that is, an accident event has occurred.

[0143] S5. When an accident event is detected, a detailed analysis of the time series data before and after the accident is performed to recall the process of the accident;

[0144] In this example, the detailed analysis is as follows:

[0145] Observe the data trends around the abnormal points and analyze the causes of the abnormalities;

[0146] Check the data fluctuations before and after the abnormal point to determine whether there is an obvious trend of change;

[0147] Analyze the relationship with other related variables to see if there is a correlation.

[0148] S6. Present the results of accident recall and cause analysis in the form of visualization or report.

[0149] In this example, the specific steps for displaying and optimizing the results of accident recall and cause analysis are as follows:

[0150] S6.1. Display the time series data before and after the accident through the visualization module;

[0151] Among them, time series data includes original data and reconstructed data;

[0152] S6.2. According to the results of the accident analysis, the results of the analysis are fed back to the anomaly detection model through the result feedback module;

[0153] S6.3. Retrain the anomaly detection model to improve the accuracy of future detections.

[0154] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0155] 1. Real-time data acquisition through monitoring components helps to find problems in a timely manner. Feature extraction can reduce data dimensions and computing costs while retaining important information.

[0156] 2. By analyzing the data before and after the accident, the development process of the accident can be traced, which helps to understand the root cause of the accident. Detailed analysis of the data before and after the accident can examine the occurrence and development of the accident from multiple perspectives. The analysis results are presented in the form of charts and reports, making the results more intuitive and easy to understand.

[0157] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0158] In this embodiment, a device for determining the cause of an accident is also provided, and the device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made are not repeated here. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0159] Figure 3 is a structural block diagram of a device for determining the cause of an accident according to an embodiment of the present application, such as Figure 3 As shown, the device comprises:

[0160] An extraction module 32, used to extract characteristic data from the time series data corresponding to the accident event;

[0161] An identification module 34, configured to input the feature data into a target anomaly detection model, so that the target anomaly detection model identifies the abnormal data in the time series data based on the feature data;

[0162] An interception module 36, configured to intercept the time series data according to the abnormal data to determine part of the time series data;

[0163] The analysis module 38 is used to analyze the partial time series data to determine the cause of the accident event.

[0164] Through the above-mentioned device, feature data in the time series data corresponding to the accident event is extracted; the feature data is input into the target anomaly detection model so that the target anomaly detection model identifies the anomaly data in the time series data based on the feature data; the time series data is intercepted according to the anomaly data to determine part of the time series data; the part of the time series data is analyzed to determine the cause of the accident event. In the embodiment of the present application, the accuracy of determining the cause of the accident is improved through feature extraction, anomaly detection, abnormal data determination, and in-depth analysis, and the efficiency is significantly improved, thereby avoiding the waste of resources and time caused by manual analysis and analysis of a large amount of useless data in traditional methods. Therefore, the problem of low efficiency in determining the cause of the accident can be solved.

[0165] In an exemplary embodiment, the extraction module 32 is used to perform at least one of the following: extracting statistical features of each variable in the time series data, wherein the statistical features include: maximum value, minimum value, mean, standard deviation; calculating the rate of change of each variable in the time series data; identifying the correlation between any two variables in the time series data; determining the frequency domain characteristics of the time series data based on the Fourier exchange method, wherein the feature data includes at least one of the following: statistical features of each variable, the rate of change of each variable, the correlation between different variables, and the frequency domain characteristics of the time series data.

[0166] In an exemplary embodiment, the extraction module 32 is used to identify the correlation r between any two variables in the time series data by the following formula:

[0167] Among them, i represents the index variable, r represents the correlation, n represents the number of samples, and x i represents the i-th observation value of variable X, y i represents the i-th observation value of variable Y, represents the sample mean of variable X, represents the sample mean of variable Y.

[0168] In an exemplary embodiment, the apparatus further comprises: a training module, configured to divide the data set into a training set and a test set, and train the anomaly detection model using the training set;

[0169] The reconstruction error of the anomaly detection model is evaluated using the test set through a reconstruction error distribution algorithm, wherein the reconstruction error distribution algorithm is: E recon represents the reconstruction error, x train represents the input data of the anomaly detection model, Reconstructed data representing the input data reconstructed by the anomaly detection model;

[0170] When the reconstruction error is less than a preset reconstruction error, it is determined that the training of the anomaly detection model is completed, and the trained anomaly detection model is determined as the target anomaly detection model.

[0171] In an exemplary embodiment, the anomaly detection model includes an encoder module and a decoder module, wherein the encoder module is expressed as follows:

[0172] z=f encode (x;θ encode ), where z represents the output data of the encoder, f encode represents the encoder function, which converts high-dimensional input data x into low-dimensional representation z, where x represents the input data and θ encode represents the set of trainable parameters of the encoder;

[0173] The expression of the decoder module is as follows:

[0174] in, represents the reconstructed data of the decoder; decode Represents the decoder function, which reconstructs the low-dimensional representation z back to high-dimensional data θ decpde represents a set of trainable parameters of the decoder.

[0175] In an exemplary embodiment, the identification module 34 is used to determine the preset threshold value T by the following formula:

[0176] T=percentile(E recon (X), A), where percentile represents the percentile function in statistics, E recon (X) represents the combination of reconstruction errors on the feature data X, X represents the feature data, and A represents the Ath percentile; determine the target reconstruction error corresponding to each feature data, and determine the size relationship between the target reconstruction error and the preset threshold; when the size relationship indicates that any target reconstruction error is greater than the preset threshold, determine that the feature data corresponding to any target reconstruction error is the abnormal data.

[0177] In an exemplary embodiment, the analysis module 38 is used to determine the target characteristic data of the partial time series data, wherein the target characteristic data includes at least one of the following: the data trend of each variable of the partial time series data, the fluctuation information of each variable of the partial time series data, and the correlation between two variables of the partial time series data; and determine the cause of the accident event based on the target characteristic data.

[0178] It should be noted that the above modules can be implemented by software or hardware. For the latter, it can be implemented in the following ways, but not limited to: the above modules are all located in the same processor; or the above modules are located in different processors in any combination.

[0179] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0180] Optionally, in this embodiment, the storage medium may be configured to store program codes for executing the following steps:

[0181] S1, extracting feature data from the time series data corresponding to the accident event;

[0182] S2, inputting the feature data into a target anomaly detection model, so that the target anomaly detection model identifies abnormal data in the time series data based on the feature data;

[0183] S3, intercepting the time series data according to the abnormal data to determine part of the time series data;

[0184] S4, analyzing the partial time series data to determine the cause of the accident event.

[0185] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0186] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0187] In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0188] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0189] S1, extracting feature data from the time series data corresponding to the accident event;

[0190] S2, inputting the feature data into a target anomaly detection model, so that the target anomaly detection model identifies abnormal data in the time series data based on the feature data;

[0191] S3, intercepting the time series data according to the abnormal data to determine part of the time series data;

[0192] S4, analyzing the partial time series data to determine the cause of the accident event.

[0193] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps in any one of the above method embodiments are implemented.

[0194] An embodiment of the present application further provides another computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0195] An embodiment of the present application also provides a computer program, which includes computer instructions, which are stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps in any one of the above method embodiments.

[0196] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0197] S1, extracting feature data from the time series data corresponding to the accident event;

[0198] S2, inputting the feature data into a target anomaly detection model, so that the target anomaly detection model identifies abnormal data in the time series data based on the feature data;

[0199] S3, intercepting the time series data according to the abnormal data to determine part of the time series data;

[0200] S4, analyzing the partial time series data to determine the cause of the accident event.

[0201] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0202] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0203] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for determining the cause of an accident, characterized in that: include: Extracting feature data from the time series data corresponding to the accident event; Inputting the feature data into a target anomaly detection model so that the target anomaly detection model identifies abnormal data in the time series data based on the feature data; intercepting the time series data according to the abnormal data to determine part of the time series data; The partial time series data is analyzed to determine the cause of the accident event.

2. The method according to claim 1, characterized in that: Extracting feature data related to the accident event from the time series data includes at least one of the following: Extracting statistical features of each variable in the time series data, wherein the statistical features include: maximum value, minimum value, mean value, and standard deviation; Calculating the rate of change of each variable in the time series data; Identifying correlations between pairs of variables in the time series data; The frequency domain characteristics of the time series data are determined based on the Fourier exchange method, wherein the characteristic data includes at least one of the following: statistical characteristics of each variable, the rate of change of each variable, the correlation between different variables, and the frequency domain characteristics of the time series data.

3. The method according to claim 2, characterized in that Identify the correlation between two variables in the time series data, including: The correlation r between two variables in the time series data is identified by the following formula: Among them, i represents the index variable, r represents the correlation, n represents the number of samples, and x i represents the i-th observation value of variable X, y i represents the i-th observation value of variable Y, represents the sample mean of variable X, represents the sample mean of variable Y.

4. The method according to claim 1, characterized in that Before inputting the feature data into a target anomaly detection model so that the target anomaly detection model identifies abnormal data in the time series data based on the feature data, the method further includes: Dividing the data set into a training set and a test set, and training the anomaly detection model using the training set; The reconstruction error of the anomaly detection model is evaluated using the test set through a reconstruction error distribution algorithm, wherein the reconstruction error distribution algorithm is: E recon represents the reconstruction error, x train represents the input data of the anomaly detection model, Reconstructed data representing the input data reconstructed by the anomaly detection model; When the reconstruction error is less than a preset reconstruction error, it is determined that the training of the anomaly detection model is completed, and the trained anomaly detection model is determined as the target anomaly detection model.

5. The method according to claim 4, characterized in that The anomaly detection model includes an encoder module and a decoder module, wherein the expression of the encoder module is as follows: z=f encode (x;θ encode ), where z represents the output data of the encoder, f encode represents the encoder function, which converts high-dimensional input data x into low-dimensional representation z, where x represents the input data and θ encode represents the set of trainable parameters of the encoder; The expression of the decoder module is as follows: in, represents the reconstructed data of the decoder; decode Represents the decoder function, which reconstructs the low-dimensional representation z back to high-dimensional data θ decode represents a set of trainable parameters of the decoder.

6. The method according to claim 1, characterized in that Inputting the feature data into a target anomaly detection model so that the target anomaly detection model identifies abnormal data in the time series data based on the feature data, including: The preset threshold T is determined by the following formula: T=percentile(E recon (X), A), where percentile represents the percentile function in statistics, E recon (X) represents the combination of the reconstruction errors on the feature data X, X represents the feature data, and A represents the Ath percentile; Determine a target reconstruction error corresponding to each feature data, and determine a magnitude relationship between the target reconstruction error and the preset threshold; When the magnitude relationship indicates that any target reconstruction error is greater than the preset threshold, it is determined that the characteristic data corresponding to any target reconstruction error is the abnormal data.

7. The method according to claim 1, characterized in that Analyze the partial time series data to determine the cause of the accident event, including: Determine target feature data of the partial time series data, wherein the target feature data includes at least one of the following: data trend of each variable of the partial time series data, fluctuation information of each variable of the partial time series data, and correlation between two variables of the partial time series data; The cause of the accident event is determined according to the target characteristic data.

8. A device for determining the cause of an accident, characterized in that: include: An extraction module, used to extract characteristic data from the time series data corresponding to the accident event; an identification module, configured to input the feature data into a target anomaly detection model, so that the target anomaly detection model identifies the abnormal data in the time series data based on the feature data; An interception module, used for intercepting the time series data according to the abnormal data to determine part of the time series data; The analysis module is used to analyze the partial time series data to determine the cause of the accident event.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program executes the method according to any one of claims 1 to 7 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method according to any one of claims 1 to 7 through the computer program.

Images