Energy storage cabinet monitoring method and device based on data analysis

Abstract

The application discloses an energy storage cabinet monitoring method and device based on data analysis, relates to the technical field of data processing, and can effectively utilize various data of energy storage cabinet operation, improve data utilization efficiency by collecting multi-dimensional operation data of the energy storage cabinet, and preprocessing the multi-dimensional operation data to obtain preprocessed data; multi-level features corresponding to the preprocessed data are extracted and fused, a pre-trained fault classification model is used to identify first fused feature data obtained through fusion to obtain a fault prediction result, and finally, an energy storage cabinet monitoring report is generated according to the fault prediction result and the multi-dimensional operation data, and the energy storage cabinet monitoring report is pushed to a staff, weak fault precursor signals can be identified from massive data, early and accurate prediction of faults is realized, monitoring efficiency and accuracy are improved, and the problem of easy missed reports is effectively avoided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a method and device for monitoring energy storage cabinet based on data analysis. BACKGROUND

[0002] With the rapid development of new energy industry, energy storage system as a key link of power grid peak shaving, renewable energy consumption, emergency backup power supply, etc., its application is increasingly widespread. The energy storage cabinet as the core unit of the energy storage system, its safety and stability of operation is crucial. The energy storage cabinet is composed of a large number of battery cells, battery management system, temperature control system, fire control system and other complex components, and its running state is affected by the coupling of multiple factors such as electrochemistry, thermodynamics and electrical characteristics. At present, the monitoring of energy storage cabinet mainly relies on the battery management system to alarm the threshold of voltage, current, temperature and other key parameters. For example, when the voltage of a certain battery cell exceeds the preset upper or lower limit, the system will issue an alarm. However, this single threshold-based method is mostly a post-alarm, that is, it can only be triggered when the fault has occurred or the parameters are obviously abnormal, and it cannot provide early warning before the fault occurs, lacking predictive maintenance capability. The energy storage cabinet generates a large amount of high-dimensional time series data during operation, and the traditional threshold monitoring method only uses the instantaneous value of the data, ignoring the changing trend of the data in the time dimension, the internal correlation between parameters and deeper statistical and frequency domain features, resulting in waste of valuable data resources. Due to the complex working condition of the energy storage system, instantaneous working condition changes (such as high-power charging and discharging) may cause temporary parameter over-limit, leading to false alarm; while some slowly developing potential faults (such as capacity attenuation and internal resistance increase) may not trigger threshold alarm in the early stage, leading to missed alarm. SUMMARY

[0003] The present application provides a method and device for monitoring energy storage cabinet based on data analysis, aiming to solve the problems of low timeliness of existing technology alarm, low data utilization efficiency and easy to miss alarm.

[0004] The first aspect of the present application provides a method for monitoring energy storage cabinet based on data analysis, comprising:

[0005] Collecting multi-dimensional running data of the energy storage cabinet, and preprocessing the multi-dimensional running data to obtain preprocessed data;

[0006] Extracting multi-level features corresponding to the preprocessed data, fusing the multi-level features to obtain first fused feature data, and identifying the first fused feature data through a pre-trained fault classification model to obtain a fault prediction result;

[0007] According to the fault prediction result and the multi-dimensional operation data, a storage cabinet monitoring report is generated, and the storage cabinet monitoring report is pushed to a staff.

[0008] In a possible implementation, multi-dimensional operation data of a storage cabinet is collected, and the multi-dimensional operation data is preprocessed to obtain preprocessed data, including:

[0009] Based on a preset data sampling frequency, cell voltage, cell temperature, bus current, bus voltage, ambient temperature and humidity, and fire extinguishing system state signals of the storage cabinet are collected to obtain multi-dimensional operation data of the storage cabinet;

[0010] A first data matrix is constructed based on multi-dimensional operation data at a plurality of continuous sampling time points, and the first data matrix is normalized to obtain preprocessed data.

[0011] In a possible implementation, multi-level features corresponding to the preprocessed data are extracted, and the multi-level features are fused to obtain first fusion feature data, including:

[0012] Statistical features, time domain features, correlation features, and frequency domain features corresponding to the preprocessed data are obtained, and the statistical features, time domain features, correlation features, and frequency domain features are spliced into a second data matrix to obtain first fusion feature data.

[0013] In a possible implementation, the first fusion feature data is identified by a pre-trained fault classification model to obtain a fault prediction result, including:

[0014] The first fusion feature data is taken as an input of the pre-trained fault classification model, an output of the pre-trained fault classification model is obtained, and target output data is obtained.

[0015] The class with the maximum probability in the target output data is taken as the fault prediction result; wherein the fault prediction result includes a specific fault class or a normal operation state class.

[0016] In a possible implementation, the pre-trained fault classification model is obtained by:

[0017] The hyperparameters of the fault classification model are initialized to obtain hyperparameter encodings;

[0018] The fitness corresponding to each hyperparameter encoding is obtained, and the hyperparameter encoding with the maximum fitness is determined as the optimal hyperparameter encoding;

[0019] According to the optimal hyperparameter encoding, variable spiral exploration is performed on the hyperparameter encoding to obtain the hyperparameter encoding after variable spiral exploration;

[0020] The hyperparameter encoding after the fuzzy crossover exploration is performed adaptive parameter escape exploration to obtain hyperparameter encoding after adaptive parameter escape exploration;

[0021] The hyperparameter encoding after the fuzzy crossover exploration is performed adaptive parameter escape exploration to obtain hyperparameter encoding after adaptive parameter escape exploration;

[0022] Determine whether the training end condition is met, if yes, determine the target hyperparameter encoding according to the hyperparameter encoding after adaptive parameter escape exploration, otherwise return to the step of determining the optimal;

[0023] According to the target hyperparameter encoding, determine the final hyperparameter of the fault classification model, and obtain the pre-trained fault classification model.

[0024] In a possible implementation, the fitness corresponding to each hyperparameter encoding is obtained, and the hyperparameter encoding with the maximum fitness is determined as the optimal hyperparameter encoding, comprising:

[0025] Collect historical multi-dimensional running data of the energy storage cabinet and real fault labels corresponding to the historical multi-dimensional running data; wherein the real fault labels include various specific fault category labels and normal running state category labels;

[0026] Extract multi-level features corresponding to the historical multi-dimensional running data, and fuse the multi-level features to obtain second fusion feature data;

[0027] After applying the hyperparameters in the hyperparameter encoding to the fault classification model, the second fusion feature data is taken as the actual input of the fault classification model, the real fault label corresponding to the second fusion feature data is taken as the expected output, and the root mean square loss function value is obtained;

[0028] Add the root mean square loss function value to a non-zero normal number and take the reciprocal to obtain the fitness corresponding to the hyperparameter encoding;

[0029] All hyperparameter encodings are traversed to obtain the fitness corresponding to all hyperparameter encodings, and the hyperparameter encoding with the maximum fitness is determined as the optimal hyperparameter encoding.

[0030] In a possible implementation, the hyperparameter encoding is performed variable spiral exploration according to the optimal hyperparameter encoding to obtain hyperparameter encoding after variable spiral exploration, comprising:

[0031]

[0032]

[0033] wherein, t represents the number of training times, represents the encoding of the i-th hyperparameter in the j-th training process, t represents the encoding of the i-th hyperparameter in the j-th training process, k represents the encoding of the i-th hyperparameter in the j-th training process, represents the encoding of the i-th hyperparameter in the j-th training process k =1,2,...,NP, NP represents the total number of hyperparameter encodings, represents the optimal hyperparameter encoding, represents a natural constant, represents a variable spiral shape factor, represents a variable spiral direction factor between (-1, 1) randomly, represents a cosine function, represents a constant pi, s represents a control coefficient of the variable spiral shape factor, and is set to 5; T represents the maximum number of training times.

[0034] In a possible implementation, the hyperparameter encoding after the variable spiral exploration is subjected to fuzzy crossover exploration, and the hyperparameter encoding after the fuzzy crossover exploration is:

[0035]

[0036]

[0037]

[0038]

[0039]

[0040]

[0041]

[0042] wherein, represents the i-th hyperparameter in the j-th training process, t represents the i-th hyperparameter in the j-th training process, m represents the i-th hyperparameter in the j-th training process, d represents the i-th hyperparameter in the j-th training process, m =1,2,...,NP, d =1,2,...,D, D represents the total dimension of the hyperparameter, represents the i-th hyperparameter in the j-th training process, m represents the i-th hyperparameter in the j-th training process, d represents the i-th hyperparameter in the j-th training process, represents a first fuzzy exploration factor, represents a second fuzzy exploration factor, represents the i-th hyperparameter in the j-th training process, t ​The hyperparameters were randomly selected and encoded during the training process. d dimensional hyperparameters, Indicates the first t The optimal hyperparameter encoding during the training process is the first d dimensional hyperparameters, Represents the first random number between (0,1). express The corresponding decision factors, Represents the arctangent function. Indicates the influence coefficient of fuzzy exploration. This represents the second random number in the range (0,1). Indicates the first t During the training process, the first m Membership degree of hyperparameter encoding after a variable spiral exploration Indicates intermediate parameters. This represents the function that takes the maximum value. This represents the fitness of the optimal hyperparameter encoding. This represents the fitness of the worst-case hyperparameter encoding. Indicates the first t During the training process, the first m Fitness of hyperparameter encoding after a variable spiral exploration This represents the adjustment coefficient between [0.8, 1).

[0043] In one possible implementation, the hyperparameter encoding after the fuzzy cross-exploration is subjected to adaptive parameter breakout exploration, resulting in the hyperparameter encoding after adaptive parameter breakout exploration as follows:

[0044]

[0045]

[0046] in, Indicates the first t During the training process, the first n The hyperparameter encoding after the first fuzzy cross-exploration d dimensional hyperparameters, n =1,2,...,NP Indicates the first n The first hyperparameter encoding after the breakthrough exploration of adaptive parameters d dimensional hyperparameters, This represents a random breakout exploration control factor between (0,1). Indicated by As the mean, with A Gaussian distributed random number with variance . Indicates the first t During the training process, the firstn The hyperparameter encoding after the fuzzy cross exploration d The adaptive parameter burst exploration factor corresponding to the hyperparameter of the dimension.

[0047] The second aspect of the application provides a data analysis-based energy storage cabinet monitoring device, comprising a data acquisition module, a data prediction module, and a monitoring and pushing module.

[0048] The data acquisition module is configured to acquire multi-dimensional operation data of the energy storage cabinet, and preprocess the multi-dimensional operation data to obtain preprocessed data.

[0049] The data prediction module is configured to extract multi-level features corresponding to the preprocessed data, fuse the multi-level features to obtain first fused feature data, and identify the first fused feature data through a pre-trained fault classification model to obtain a fault prediction result.

[0050] The monitoring and pushing module is configured to generate an energy storage cabinet monitoring report based on the fault prediction result and the multi-dimensional operation data, and push the energy storage cabinet monitoring report to a staff member.

[0051] Advantages:

[0052] The application provides a data analysis-based energy storage cabinet monitoring method and device, which can effectively utilize various data of the energy storage cabinet operation by acquiring multi-dimensional operation data of the energy storage cabinet and preprocessing the multi-dimensional operation data to obtain preprocessed data, thereby improving data utilization efficiency; extract multi-level features corresponding to the preprocessed data, fuse the multi-level features, and identify first fused feature data obtained by fusion through a pre-trained fault classification model to obtain a fault prediction result; finally, generate an energy storage cabinet monitoring report based on the fault prediction result and the multi-dimensional operation data, and push the energy storage cabinet monitoring report to a staff member, which can identify weak fault precursor signals from massive data, realize early and accurate prediction of faults, improve monitoring efficiency and accuracy, and effectively avoid the problem of easy missed reports. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the description of the embodiments of the application. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0054] Figure 1 is a flowchart of a data analysis-based energy storage cabinet monitoring method according to an embodiment of the application.

[0055] Figure 2 is a structural schematic diagram of a data analysis-based energy storage cabinet monitoring device according to an embodiment of the present application.

[0056] Figure 3 is a structural schematic diagram of an electronic device according to an embodiment of the present application.

[0057] Legend: 201-data acquisition module, 202-data prediction module, 203-monitoring push module, 301-memory, 302-processor, 303-communication bus. DETAILED DESCRIPTION

[0058] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0059] As shown in Figure 1 , the present application provides a data analysis-based energy storage cabinet monitoring method, comprising:

[0060] S101, collecting multi-dimensional running data of an energy storage cabinet, and pre-processing the multi-dimensional running data to obtain pre-processed data.

[0061] In a possible implementation, collecting multi-dimensional running data of an energy storage cabinet, and pre-processing the multi-dimensional running data to obtain pre-processed data, comprises:

[0062] Based on a preset data sampling frequency, collecting cell voltage, cell temperature, bus current, bus voltage, environmental temperature and humidity, and fire extinguishing system state signal of the energy storage cabinet to obtain multi-dimensional running data of the energy storage cabinet; constructing a first data matrix based on multi-dimensional running data at consecutive multiple sampling time points, and performing normalization processing on the first data matrix to obtain pre-processed data. For example, a value M can be pre-set, and then a first data matrix is constructed based on multi-dimensional running data at consecutive M sampling time points.

[0063] The fire extinguishing system state signal can be data collected by some Internet of Things sensors, such as data collected by smoke sensors, flammable gas concentration sensors, and other Internet of Things sensors. By performing normalization processing on the first data matrix to obtain pre-processed data, the complexity of the data can be effectively reduced, and the data processing efficiency can be improved.

[0064] S102, extract the multi-level features corresponding to the preprocessed data, fuse the multi-level features to obtain first fused feature data, and identify the first fused feature data through a pre-trained fault classification model to obtain a fault prediction result.

[0065] In a possible implementation, the extracting the multi-level features corresponding to the preprocessed data and fusing the multi-level features to obtain first fused feature data comprises: obtaining statistical features, time domain features, correlation features and frequency domain features corresponding to the preprocessed data, and splicing the statistical features, the time domain features, the correlation features and the frequency domain features into a second data matrix to obtain the first fused feature data.

[0066] Optionally, the statistical features can include mean, variance, range, skewness and / or kurtosis of each parameter in a preset time window, for example, the mean and variance of all cell voltages in a preset time window (i.e., a plurality of continuous sampling time points) can be calculated; the cell with the highest voltage and the cell with the lowest voltage are found, and the difference (i.e., the range) between the two is obtained; the mean and variance of all cell temperatures are calculated.

[0067] The time domain features can include the rate of change, the fluctuation rate and / or the trend slope of each parameter, for example, the average rate of change of the bus current in the window can be calculated, and the trend slope (through linear fitting) of the temperature of a certain cell in the window can be calculated.

[0068] The correlation features can include the correlation coefficient between different parameters, for example, the Pearson correlation coefficient between the voltage of a certain cell and the temperature of the cell can be calculated, and the correlation coefficient between the bus current and the average temperature change rate in the energy storage cabinet can be calculated.

[0069] The frequency domain features can include performing fast Fourier transform on the time series signals of the cell voltage, the cell temperature, the bus current, the bus voltage, etc., to extract the frequency spectrum features.

[0070] The embodiments of the present application comprehensively depict the running state of the energy storage cabinet by extracting the multi-level features such as the statistical features, the time domain features, the correlation features and the frequency domain features of the multi-dimensional running data. In combination with the pre-trained fault classification model, weak fault precursor signals can be identified from massive data to realize early and accurate prediction of faults, change post-maintenance into pre-warning, effectively improve the utilization rate of data and the timeliness of fault warning, and avoid the problem of easy omission.

[0071] The fault prediction result output by the fault classification model is a specific fault category (such as cell consistency deterioration, thermal runaway risk, etc.) or a normal running state, rather than a general parameter over-limit alarm, which greatly helps the operation and maintenance personnel to quickly locate the problem source, shorten the fault troubleshooting time, and improve the operation and maintenance efficiency.

[0072] S103, generating a storage cabinet monitoring report according to the fault prediction result and the multi-dimensional operation data, and pushing the storage cabinet monitoring report to a staff.

[0073] The embodiment of the application provides a storage cabinet monitoring method and device based on data analysis. The multi-dimensional operation data of the storage cabinet is collected, and the multi-dimensional operation data is preprocessed to obtain preprocessed data. The method can effectively utilize various data of the storage cabinet operation, and improve data utilization efficiency. The multi-level features corresponding to the preprocessed data are extracted and fused, and the first fused feature data obtained through fusion is identified by using a pre-trained fault classification model to obtain a fault prediction result. Finally, a storage cabinet monitoring report is generated according to the fault prediction result and the multi-dimensional operation data, and the storage cabinet monitoring report is pushed to a staff. The method can identify weak fault precursor signals from massive data, realize early and accurate prediction of faults, improve monitoring efficiency and accuracy, and effectively avoid the problem of easy missed reports.

[0074] In a possible implementation, the first fused feature data is identified by using the pre-trained fault classification model to obtain a fault prediction result, which includes:

[0075] The first fused feature data is taken as an input of the pre-trained fault classification model, and an output of the pre-trained fault classification model is obtained to obtain target output data.

[0076] The class with the maximum probability in the target output data is taken as the fault prediction result. The fault prediction result includes a specific fault class or a normal operation state class.

[0077] In a possible implementation, the method for obtaining the pre-trained fault classification model includes:

[0078] The hyperparameters of the fault classification model are initialized to obtain hyperparameter codes.

[0079] For example, the hyperparameters of the fault classification model can be initialized by using a random initialization method or a chaotic mapping initialization method, and the initialized hyperparameters are coded into vectors to obtain the hyperparameter codes.

[0080] The fitness corresponding to each hyperparameter code is obtained, and the hyperparameter code with the maximum fitness is determined as an optimal hyperparameter code.

[0081] According to the optimal hyperparameter code, variable spiral exploration is performed on the hyperparameter codes to obtain the hyperparameter codes after variable spiral exploration.

[0082] The hyperparameter encoding after the fuzzy crossover exploration is subjected to adaptive parameter breakout exploration to obtain hyperparameter encoding after adaptive parameter breakout exploration.

[0083] The hyperparameter encoding after the fuzzy crossover exploration is subjected to adaptive parameter breakout exploration to obtain hyperparameter encoding after adaptive parameter breakout exploration.

[0084] It is determined whether the training end condition is met, and if so, the target hyperparameter encoding is determined according to the hyperparameter encoding after adaptive parameter breakout exploration, and otherwise, the step of determining the optimal is returned.

[0085] In the case where the total number of training times reaches a preset maximum number of training times, the training end condition can be determined to be met.

[0086] According to the target hyperparameter encoding, the final hyperparameters of the fault classification model are determined (i.e., the hyperparameters in the target hyperparameter encoding are used as the final hyperparameters of the fault classification model), and a pre-trained fault classification model is obtained.

[0087] The embodiments of the present application adopt an innovative hybrid intelligent optimization algorithm, which combines variable spiral, fuzzy crossover and adaptive breakout strategies to train the fault classification model, can efficiently and automatically find the optimal or near-optimal solution in a complex hyperparameter space, avoids the problems of low efficiency and easy to fall into local optimum of traditional grid exploration or random exploration, and ensures that the fault prediction model has the highest accuracy and generalization ability.

[0088] In one possible implementation, the fitness corresponding to each hyperparameter encoding is obtained, and the hyperparameter encoding with the maximum fitness is determined as the optimal hyperparameter encoding, including:

[0089] The historical multi-dimensional running data of the energy storage cabinet and the real fault labels corresponding to the historical multi-dimensional running data are collected; wherein the real fault labels include various specific fault category labels and normal running state category labels;

[0090] The multi-level features corresponding to the historical multi-dimensional running data are extracted, and the multi-level features are fused to obtain second fusion feature data;

[0091] After applying the hyperparameters in the hyperparameter encoding to the fault classification model, the second fusion feature data is used as the actual input of the fault classification model, the real fault label corresponding to the second fusion feature data is used as the expected output, and the root mean square loss function value is obtained;

[0092] The root mean square loss function value is added to a non-zero normal item and taken as an inverse to obtain the fitness corresponding to the hyperparameter encoding;

[0093] Iterate through all hyperparameter codes, obtain the fitness of each hyperparameter code, and determine the hyperparameter code with the highest fitness as the optimal hyperparameter code.

[0094] In one possible implementation, based on the optimal hyperparameter encoding, a variable spiral exploration is performed on the hyperparameter encoding to obtain the hyperparameter encoding after the variable spiral exploration:

[0095]

[0096]

[0097] in, t Indicates the number of training sessions. Indicates the first t During the training process, the first k Hyperparameter encoding, This represents the hyperparameter encoding after variable spiral exploration. , k =1,2,...,NP, where NP represents the total number of hyperparameter codes. This represents the optimal hyperparameter encoding. Represents the natural constant. Indicates the variable spiral shape factor. This represents a random variable spiral direction factor between (-1, 1). Represents the cosine function. Represents pi (π). s The variable spiral shape factor is represented by a control coefficient, set to 5; T represents the maximum number of training iterations.

[0098] The hyperparameter encoding is subjected to variable spiral exploration, which simulates the spiral exploration behavior of organisms in nature. Centered on the current optimal hyperparameter encoding, other individuals are guided to approach it while maintaining a certain degree of exploration. The variable spiral shape factor is larger in the early stage to enhance global exploration and smaller in the later stage to enhance local development. It can efficiently utilize the information of the optimal solution and accelerate population convergence.

[0099] In one possible implementation, the hyperparameter encoding after the variable spiral exploration is subjected to fuzzy cross-exploration, resulting in the following hyperparameter encoding after fuzzy cross-exploration:

[0100]

[0101]

[0102]

[0103]

[0104]

[0105]

[0106]

[0107] in, Indicates the first t During the training process, the first m The hyperparameter encoding after the first variable spiral exploration d dimensional hyperparameters, m =1,2,...,NP d =1,2,...,D, where D represents the total dimension of the hyperparameters. Indicates the first m The hyperparameter encoding after the first fuzzy cross-exploration d dimensional hyperparameters, Represents the first fuzzy exploration factor. This represents the second fuzzy exploration factor. Indicates the first t The hyperparameters were randomly selected and encoded during the training process. d dimensional hyperparameters, Indicates the first t The optimal hyperparameter encoding during the training process is the first d dimensional hyperparameters, Represents the first random number between (0,1). express Corresponding decision factors Represents the arctangent function. This represents the influence coefficient of fuzzy exploration. This represents the second random number between (0,1). Indicates the first t During the training process, the first m Membership degree of hyperparameter encoding after a variable spiral exploration Indicates intermediate parameters. This represents the function that takes the maximum value. This represents the fitness of the optimal hyperparameter encoding. This represents the fitness of the worst-case hyperparameter encoding. Indicates the first t During the training process, the first m Fitness of hyperparameter encoding after a variable spiral exploration This represents the adjustment coefficient between [0.8, 1).

[0108] Fuzzy crossover exploration is performed on the hyperparameter encoding after the variable spiral exploration. The crossover strength is determined according to a fuzzy membership function. Unlike the traditional fixed crossover rate, fuzzy crossover can be dynamically adjusted according to the population evolution state. When the population diversity decreases, the crossover strength is increased to produce more diverse offspring, thereby escaping the local optimum.

[0109] In one possible implementation, the hyperparameter encoding after the fuzzy cross-exploration is subjected to adaptive parameter breakout exploration, resulting in the hyperparameter encoding after adaptive parameter breakout exploration as follows:

[0110]

[0111]

[0112] in, Indicates the first t During the training process, the first n The hyperparameter encoding after the first fuzzy cross-exploration d dimensional hyperparameters, n =1,2,...,NP Indicates the first n The first hyperparameter encoding after the breakthrough exploration of adaptive parameters d dimensional hyperparameters, This represents a random breakout exploration control factor between (0,1). Indicated by As the mean, with A Gaussian distributed random number with variance . Indicates the first t During the training process, the first n The hyperparameter encoding after the first fuzzy cross-exploration d The adaptive parameter breakthrough exploration factor corresponding to the hyperparameter.

[0113] Adaptive parameter breakthrough exploration is performed on the hyperparameter encoding after the fuzzy cross-exploration. Compared with the traditional uniformly distributed random walk, it can have a higher sampling density near the optimal hyperparameter encoding, and generate a series of sequences of mutations around the population's optimal value. This allows the hyperparameter encoding to generate statistically regular perturbations in the neighborhood of the global optimal solution, which maintains population diversity and enables refined exploration through decay characteristics.

[0114] Optionally, after the adaptive parameter breakthrough exploration, simulated annealing can be used to accept probabilistically inferior solutions. This ensures both the algorithm's global exploration capability and training speed, guaranteeing that the algorithm can find the global optimum at the end of training. Furthermore, after each exploration of hyperparameter encoding, out-of-bounds handling can be performed on the hyperparameter encoding to ensure its validity.

[0115] The embodiment of the application trains a fault classification model through variable spiral, fuzzy intersection and adaptive breakthrough strategies, can efficiently and automatically find an optimal or near-optimal solution in a complex hyperparameter space, avoids the problems of low efficiency and easy falling into local optimum of traditional grid exploration or random exploration, and ensures that the fault prediction model has the highest accuracy and generalization ability.

[0116] As shown in Figure 2 Based on the same inventive concept, the embodiment of the application provides a data analysis-based energy storage cabinet monitoring device, which comprises a data acquisition module 201, a data prediction module 202 and a monitoring and pushing module 203.

[0117] The data acquisition module 201 is configured to acquire multi-dimensional operation data of an energy storage cabinet, and pre-process the multi-dimensional operation data to obtain pre-processed data.

[0118] The data prediction module 202 is configured to extract multi-level features corresponding to the pre-processed data, fuse the multi-level features to obtain first fused feature data, and identify the first fused feature data through a pre-trained fault classification model to obtain a fault prediction result.

[0119] The monitoring and pushing module 203 is configured to generate an energy storage cabinet monitoring report according to the fault prediction result and the multi-dimensional operation data, and push the energy storage cabinet monitoring report to a staff member.

[0120] The data analysis-based energy storage cabinet monitoring device provided by the embodiment of the application can execute the above-mentioned method technical solution, and has similar principles and beneficial effects, which will not be described here.

[0121] As shown in Figure 3 Based on the same inventive concept, another embodiment of the application provides an electronic device, which comprises a memory 301 and a processor 302. The memory 301 and the processor 302 complete mutual communication through a communication bus 303.

[0122] The memory 301 is configured to store code instructions.

[0123] The processor 302 is configured to run the code instructions, so that the electronic device executes the data analysis-based energy storage cabinet monitoring method provided by the embodiment of the application.

[0124] The communication bus 303 mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus 303 can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or only one type of bus. The communication interface is used for communication between the terminal and other devices. The memory 301 can include a Random Access Memory (RAM) and can also include a non-volatile memory, such as at least one disk memory. Optionally, the memory 301 can also be at least one storage device located away from the aforementioned processor 302.

[0125] The processor 302 mentioned above can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; can also be a Digital Signal Processing (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component.

[0126] In addition, to achieve the above-mentioned purpose, the embodiment of the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the energy storage cabinet monitoring method based on data analysis provided by the embodiment of the present application.

[0127] Each embodiment in the specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same and similar parts between each embodiment can be referred to each other.

[0128] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, apparatuses, electronic device and computer program products according to the embodiments of the present application. It is understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal devices to produce a machine, so that the instructions, which are executed via the processor of the computer or other programmable data processing terminal devices, generate a device implemented in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 one or more flows and / or blocks

[0129] These computer program instructions can also be stored in a computer readable memory capable of directing a computer or other programmable data processing terminal devices to work in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including instruction devices, which implement the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 one or more flows and / or blocks

[0130] These computer program instructions can also be loaded into a computer or other programmable data processing terminal device, so that a series of operation steps are performed on the computer or other programmable terminal device to produce a computer implemented process, so that the instructions executed on the computer or other programmable terminal device provide steps for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 one or more flows and / or blocks

[0131] Although the preferred embodiments of the present application have been described, those skilled in the art can make further changes and modifications to the embodiments once they know the basic inventive concept. Therefore, the appended claims are intended to be interpreted as including all the preferred embodiments and all the changes and modifications falling within the scope of the embodiments of the present application.

[0132] Finally, it is also need to be explained that, in the embodiments of the present application, the relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or terminal device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or terminal device. Without more limitation, the element defined by the statement "including a" does not exclude the existence of other same elements in the process, method, article or terminal device including the element.

[0133] The principles and implementation manners of the present application are described by applying specific examples in the embodiments of the present application, and the above embodiment descriptions are only used to help understand the method of the present application and its core idea; meanwhile, for the general technical personnel in the field, according to the idea of the present application, there will be changes in the specific implementation manners and application ranges, and in view of the above, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A data analysis based energy storage cabinet monitoring method, characterized in that, The application relates to a method for monitoring a storage energy cabinet, and the method comprises the following steps: Collecting multi-dimensional operation data of a storage energy cabinet, and preprocessing the multi-dimensional operation data to obtain preprocessed data; Extracting multi-level features corresponding to the preprocessed data, fusing the multi-level features, obtaining first fused feature data, and identifying the first fused feature data through a pre-trained fault classification model to obtain a fault prediction result; Generating a storage energy cabinet monitoring report according to the fault prediction result and the multi-dimensional operation data, and pushing the storage energy cabinet monitoring report to a staff member; The method for obtaining the pre-trained fault classification model comprises: Initializing hyperparameters of the fault classification model to obtain hyperparameter codes; Obtaining the fitness of each hyperparameter code, and determining the hyperparameter code with the maximum fitness as the optimal hyperparameter code; Performing variable spiral exploration on the hyperparameter codes according to the optimal hyperparameter code to obtain hyperparameter codes after variable spiral exploration; Performing fuzzy crossover exploration on the hyperparameter codes after the variable spiral exploration to obtain hyperparameter codes after fuzzy crossover exploration; Performing self-adaptive parameter breakthrough exploration on the hyperparameter codes after the fuzzy crossover exploration to obtain hyperparameter codes after self-adaptive parameter breakthrough exploration; Determining whether a training end condition is met, and if yes, determining a target hyperparameter code according to the hyperparameter codes after the self-adaptive parameter breakthrough exploration, and if not, returning to the step of determining the optimal hyperparameter code; Determining the final hyperparameters of the fault classification model according to the target hyperparameter code to obtain the pre-trained fault classification model.

2. The data analysis based energy storage tank monitoring method of claim 1, wherein, Collecting multi-dimensional operation data of a storage energy cabinet, and preprocessing the multi-dimensional operation data to obtain preprocessed data, which comprises the following steps: Collecting cell voltage, cell temperature, bus current, bus voltage, environmental temperature and humidity, and fire extinguishing system state signals of the storage energy cabinet at a preset data sampling frequency to obtain multi-dimensional operation data of the storage energy cabinet; Constructing a first data matrix from the multi-dimensional operation data at a plurality of continuous sampling time points, and performing normalization processing on the first data matrix to obtain preprocessed data.

3. The data analysis based energy storage cabinet monitoring method of claim 1, wherein, Extracting multi-level features corresponding to the preprocessed data, fusing the multi-level features, obtaining first fused feature data, which comprises the following steps: Obtaining statistical features, time domain features, correlation features and frequency domain features corresponding to the preprocessed data, and splicing the statistical features, time domain features, correlation features and frequency domain features into a second data matrix to obtain first fused feature data.

4. The data analysis based energy storage tank monitoring method of claim 3, wherein, Identifying the first fused feature data through a pre-trained fault classification model to obtain a fault prediction result, which comprises the following steps: Taking the first fused feature data as the input of the pre-trained fault classification model, obtaining the output of the pre-trained fault classification model, and obtaining target output data; Taking the class with the maximum probability in the target output data as the fault prediction result; wherein the fault prediction result comprises a specific fault class or a normal operation state class.

5. The data analytics based energy storage cabinet monitoring method of claim 1, wherein, Obtaining the fitness of each hyperparameter code, and determining the hyperparameter code with the maximum fitness as the optimal hyperparameter code, which comprises the following steps: Collect historical multi-dimensional running data of the energy storage cabinet and real fault labels corresponding to the historical multi-dimensional running data; wherein, the real fault labels include various specific fault category labels and a normal running state category label; Extract multi-level features corresponding to the historical multi-dimensional running data, and fuse the multi-level features to obtain second fused feature data; After applying the hyperparameters in the hyperparameter encoding to the fault classification model, take the second fused feature data as the actual input of the fault classification model, take the real fault labels corresponding to the second fused feature data as the expected output, and obtain a root mean square loss function value; Add the root mean square loss function value to a non-zero normal term and take the reciprocal to obtain the fitness corresponding to the hyperparameter encoding; Iterate through all hyperparameter encodings, obtain the fitness corresponding to all hyperparameter encodings, and determine the hyperparameter encoding with the maximum fitness as the optimal hyperparameter encoding.

6. The data analytics based energy storage cabinet monitoring method of claim 1, wherein, According to the optimal hyperparameter encoding, perform variable spiral exploration on the hyperparameter encoding to obtain the hyperparameter encoding after variable spiral exploration as: in, t Indicates the number of training sessions. Indicates the first t During the training process, the first k Hyperparameter encoding, This represents the hyperparameter encoding after variable spiral exploration. , k =1,2,...,NP, where NP represents the total number of hyperparameter codes. This represents the optimal hyperparameter encoding. Represents the natural constant. Indicates the variable spiral shape factor. This represents a random variable spiral direction factor between (-1, 1). Represents the cosine function. Represents pi (π). s The variable spiral shape factor is represented by a control coefficient, set to 5; T represents the maximum number of training iterations.

7. The data analysis based energy storage cabinet monitoring method of claim 6, wherein, Perform fuzzy crossover exploration on the hyperparameter encoding after the variable spiral exploration to obtain the hyperparameter encoding after the fuzzy crossover exploration as: in, Indicates the first t During the training process, the first m The hyperparameter encoding after the first variable spiral exploration d dimensional hyperparameters, m =1,2,...,NP d =1,2,...,D, where D represents the total dimension of the hyperparameters. Indicates the first m The hyperparameter encoding after the first fuzzy cross-exploration d dimensional hyperparameters, Represents the first fuzzy exploration factor. This represents the second fuzzy exploration factor. Indicates the first t The hyperparameters were randomly selected and encoded during the training process. d dimensional hyperparameters, Indicates the first t The optimal hyperparameter encoding during the training process is the first d dimensional hyperparameters, Represents the first random number between (0,1). express The corresponding decision factors, Represents the arctangent function. This represents the influence coefficient of fuzzy exploration. This represents the second random number between (0,1). Indicates the first t During the training process, the first m Membership degree of hyperparameter encoding after a variable spiral exploration Indicates intermediate parameters. This represents the function that takes the maximum value. This represents the fitness of the optimal hyperparameter encoding. This represents the fitness of the worst-case hyperparameter encoding. Indicates the first t During the training process, the first m Fitness of hyperparameter encoding after a variable spiral exploration This represents the adjustment coefficient between [0.8, 1).

8. The data analysis based energy storage cabinet monitoring method of claim 7, wherein, Perform adaptive parameter breakthrough exploration on the hyperparameter encoding after the fuzzy crossover exploration to obtain the hyperparameter encoding after the adaptive parameter breakthrough exploration as: wherein, represents the i-th hyperparameter of the hyperparameter encoding after the j-th fuzzy crossover exploration in the k-th training process, t n represents the i-th hyperparameter of the hyperparameter encoding after the j-th fuzzy crossover exploration in the k-th training process, d n =1,2,...,NP, represents the i-th hyperparameter of the hyperparameter encoding after the j-th fuzzy crossover exploration in the k-th training process, n d represents the i-th hyperparameter of the hyperparameter encoding after the j-th fuzzy crossover exploration in the k-th training process, represents a random jie exploration control factor between (0, 1), represents a Gaussian distribution random number with a mean of and a variance of represents the i-th hyperparameter of the hyperparameter encoding after the j-th fuzzy crossover exploration in the k-th training process, t represents the i-th hyperparameter of the hyperparameter encoding after the j-th fuzzy crossover exploration in the k-th training process, n d represents the adaptive parameter jie exploration factor corresponding to the i-th hyperparameter of the hyperparameter encoding after the j-th fuzzy crossover exploration in the k-th training process.​​​​​​ 9. A data analysis-based energy storage tank monitoring device capable of performing the data analysis-based energy storage tank monitoring method according to any one of claims 1 to 8, characterized by It comprises: a data collection module, a data prediction module, and a monitoring pushing module; The data collection module is configured to collect multi-dimensional running data of the energy storage cabinet, and to preprocess the multi-dimensional running data to obtain preprocessed data; The data prediction module is configured to extract multi-level features corresponding to the preprocessed data, fuse the multi-level features to obtain first fused feature data, and identify the first fused feature data through a pre-trained fault classification model to obtain a fault prediction result; The monitoring pushing module is configured to generate an energy storage cabinet monitoring report according to the fault prediction result and the multi-dimensional running data, and to push the energy storage cabinet monitoring report to a staff.

Images