State evaluation method of heterogeneous uninterruptible power supply cluster and related device

Abstract

The invention relates to the technical field of batteries, in particular to a state evaluation method of a heterogeneous uninterruptible power supply cluster and a related device. The method comprises the following steps: acquiring a multi-class operation parameter sequence of each energy storage unit in a heterogeneous uninterruptible power supply cluster within a preset sampling duration; for each type of operation parameter sequence of each energy storage unit, segmenting the corresponding operation parameter sequence based on a mutation point detection algorithm to obtain a plurality of parameter subsequences; calculating the similarity among the parameter subsequences based on a dynamic time warping algorithm, and generating a similarity image; inputting the similarity image into a battery state evaluation model to generate a state evaluation result of the heterogeneous uninterruptible power supply cluster; wherein the battery state evaluation model is a convolutional neural network. According to the method, the operation parameter sequences of various types in different energy storage units are subjected to segmentation processing, and the segmented operation parameters are subjected to multi-modal data fusion, so that the recognition precision of the operation state of the heterogeneous uninterruptible power supply cluster is greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of battery technology, and in particular to a state assessment method and related devices for a heterogeneous uninterruptible power supply cluster. Background Art

[0002] Uninterruptible power supply (UPS) systems, as critical infrastructure, are widely used in data centers and communication base stations. Energy storage units, as core components of UPS systems, have a direct impact on system reliability and availability. Energy storage unit types include, but are not limited to, lead-acid batteries, lithium-ion batteries, sodium-ion batteries, lithium metal batteries, semi-solid-state batteries, and solid-state batteries. Lithium-ion batteries include, but are not limited to, lithium iron phosphate batteries, ternary lithium batteries, and lithium titanate batteries.

[0003] When multiple uninterruptible power supply systems are equipped with different types of energy storage units, a heterogeneous uninterruptible power supply cluster with different types of batteries can be formed. The existing method is usually to independently analyze the single operating parameters such as voltage or current of each uninterruptible power supply system to evaluate the operating status of the entire heterogeneous uninterruptible power supply cluster. Taking into account the obvious differences in physical characteristics, aging mechanisms and applicable working conditions of various energy storage units, the above method fails to integrate and analyze the operating parameters across systems, resulting in low data utilization and difficulty in comprehensively and accurately evaluating the battery's state of health (State of Health, SOH) and remaining useful life (Remaining Useful Life, RUL) and other status characteristics. Therefore, it is necessary to provide a state assessment method and related devices for a heterogeneous uninterruptible power supply cluster. Summary of the Invention

[0004] The present invention provides a method and related device for evaluating the status of a heterogeneous uninterruptible power supply cluster, which improves the problem in the prior art of inaccurate status evaluation of a heterogeneous uninterruptible power supply cluster due to the presence of multiple types of energy storage units.

[0005] The present invention provides a method for assessing the status of a heterogeneous uninterruptible power supply cluster, comprising: obtaining multiple types of operating parameter sequences for each energy storage unit in the heterogeneous uninterruptible power supply cluster within a preset sampling time; for each type of operating parameter sequence of each energy storage unit, segmenting the corresponding operating parameter sequence based on a mutation point detection algorithm to obtain multiple parameter subsequences; calculating the similarity between the various parameter subsequences based on a dynamic time warping algorithm to generate a similarity image; and inputting the similarity image into a battery status assessment model to generate a status assessment result for the heterogeneous uninterruptible power supply cluster; wherein the battery status assessment model is a convolutional neural network.

[0006] In one embodiment of the present invention, obtaining multiple operating parameter sequences for each energy storage unit in a heterogeneous uninterruptible power supply cluster within a preset sampling period includes: obtaining multiple initial operating parameter sequences for each energy storage unit in the heterogeneous uninterruptible power supply cluster within the preset sampling period; and filtering each initial operating parameter sequence for each energy storage unit to obtain a filtered operating parameter sequence.

[0007] In one embodiment of the present invention, for each initial parameter sequence of each energy storage unit, the initial operating parameter sequence is filtered to obtain a filtered operating parameter sequence, including: starting from the starting position of the initial operating parameter sequence, extracting multiple operating parameters based on a preset sliding window length, calculating the mean of all extracted operating parameters to obtain a parameter mean; moving the sliding window according to a preset step size, repeating the above calculation until all operating parameters in the initial operating parameter sequence participate in at least one mean calculation; and arranging all parameter means in sequence to obtain a filtered operating parameter sequence.

[0008] In one embodiment of the present invention, a corresponding operating parameter sequence is segmented based on a mutation point detection algorithm to obtain multiple parameter subsequences, including: performing mutation point detection on each operating parameter in the operating parameter sequence based on the mutation point detection algorithm to obtain all mutation parameters in the operating parameter sequence; and segmenting the operating parameter sequence based on all mutation parameters to obtain multiple parameter subsequences.

[0009] In one embodiment of the present invention, the similarity between each parameter subsequence is calculated based on the dynamic time warping algorithm to generate a similarity image, including: numbering all parameter subsequences in sequence according to the type of operating parameters; calculating the similarity between each parameter subsequence based on the dynamic time warping algorithm, and using the numbers of the corresponding parameter subsequences as the row index number and column index number of the similarity; arranging all similarities in sequence according to the corresponding row index number and column index number to generate a similarity image.

[0010] In one embodiment of the present invention, the similarity between any two parameter subsequences is generated according to the following steps: calculating the parameter value difference between the two parameter subsequences and constructing a corresponding distance matrix; based on the dynamic time warping algorithm, obtaining the alignment distance between the two parameter subsequences according to the distance matrix; and normalizing the alignment distance to obtain the similarity between the two parameter subsequences.

[0011] In one embodiment of the present invention, the evaluation results are multi-classified, and the battery status evaluation model includes a feature extraction network and multiple expert networks with different network structures cascaded therewith. The similarity image is input into the battery status evaluation model to generate the evaluation results of the heterogeneous uninterruptible power supply cluster, including: inputting the similarity image into the feature extraction network of the battery status evaluation model to extract shared image features; inputting the shared image features into multiple expert networks of the battery status evaluation model respectively, each expert network extracting task features associated with its corresponding evaluation result type and generating the corresponding type of evaluation results.

[0012] In one embodiment of the present invention, a state assessment device for a heterogeneous uninterruptible power supply cluster is also provided. The device includes: a parameter acquisition module for obtaining multiple types of operating parameter sequences for each energy storage unit in the heterogeneous uninterruptible power supply cluster within a preset sampling period; a mutation point detection module for segmenting the corresponding operating parameter sequence for each type of operating parameter sequence of each energy storage unit based on a mutation point detection algorithm to obtain multiple parameter subsequences; an image generation module for calculating the similarity between any two parameter subsequences based on a dynamic time warping algorithm to generate a similarity image; and an assessment module for inputting the similarity image into a battery state assessment model to generate a state assessment result for the heterogeneous uninterruptible power supply cluster; wherein the battery state assessment model is a convolutional neural network.

[0013] In one embodiment of the present invention, an electronic device is also provided, including: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the electronic device implements any of the above-mentioned methods for assessing the status of a heterogeneous uninterruptible power supply cluster.

[0014] In one embodiment of the present invention, a computer-readable storage medium is further provided, on which a computer program is stored. When the computer program is executed by a processor of a computer, the computer executes any of the above-mentioned methods for assessing the status of a heterogeneous uninterruptible power supply cluster.

[0015] As described above, the present invention proposes a method and related device for assessing the status of a heterogeneous uninterruptible power supply cluster, which has the following beneficial effects: within a preset sampling time, multiple types of operating parameters of each energy storage unit in the heterogeneous uninterruptible power supply cluster are collected and parameter sequences of corresponding types are generated. These parameters are divided into parameter subsequences representing different operating stages using a mutation point detection algorithm, and the similarity of the parameter subsequences is calculated using a dynamic time warping algorithm. Multimodal data fusion is performed on the different types of operating parameters to obtain a similarity image. A battery status assessment model is then invoked to extract image features, thereby obtaining a status assessment result for the heterogeneous uninterruptible power supply cluster. The present invention separately segments the multiple types of operating parameter sequences in different energy storage units and performs multimodal data fusion on the segmented operating parameters to fully extract fine-grained features of the operating parameters and establish associations between different types of operating parameters. This significantly improves the accuracy of identifying the operating status of the heterogeneous uninterruptible power supply cluster, further enhances the assessment performance of the battery management system, and effectively ensures the stable operation of the heterogeneous uninterruptible power supply cluster. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic flow chart of a method for evaluating the status of a heterogeneous uninterruptible power supply cluster provided by an embodiment of the present invention;

[0017] Figure 2 Shown is a structural block diagram of a state assessment device for a heterogeneous uninterruptible power supply cluster provided by an embodiment of the present invention;

[0018] Figure 3 Shown is a structural schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following describes the embodiments of the present invention through specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other unless they conflict.

[0020] It should be noted that the illustrations provided in the following embodiments are merely schematic illustrations of the basic concept of the present invention. Although the illustrations only show components related to the present invention and are not drawn according to the number, shape, and size of components in actual implementation, the type, quantity, and proportion of each component in actual implementation may be changed arbitrarily, and the component layout may also be more complex.

[0021] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the embodiments of the present invention may be practiced without these specific details. In other embodiments, structures and devices are shown in block diagram form rather than in detail to avoid obscuring the embodiments of the present invention.

[0022] The inventors have found that in recent years, with the continuous development of sensor technology and artificial intelligence technology, multimodal data fusion methods have gradually become an important means to improve the performance of battery management systems. Although studies have attempted to apply multimodal data such as temperature, internal resistance, and charge and discharge curves to battery status assessment, the existing methods still have the following shortcomings: First, the depth and breadth of data fusion are limited, and the correlation characteristics between each modal data are not effectively mined, resulting in one-sided and low-precision assessment results; second, because heterogeneous uninterruptible power supply clusters contain various types of energy storage units (such as lithium iron phosphate batteries, ternary lithium batteries, etc.), the existing methods lack an efficient fusion strategy that can use other unit data to indirectly infer the state of the target unit. In addition, the existing multimodal data fusion methods are insufficient in terms of real-time and responsiveness of data processing, and it is difficult to meet the actual needs of uninterruptible power supply systems for rapid battery status assessment.

[0023] The present invention provides a state assessment method for a heterogeneous uninterruptible power supply cluster. Within a preset sampling time, multiple types of operating parameter sequences of each energy storage unit in the heterogeneous uninterruptible power supply cluster are collected. The sequences are divided into parameter subsequences representing different operating stages using a mutation point detection algorithm, and the similarity of the parameter subsequences is calculated using a dynamic time warping algorithm. Multimodal data fusion is performed on different types of operating parameters to obtain a similarity image. A battery state assessment model is called to extract image features to obtain a state assessment result of the heterogeneous uninterruptible power supply cluster. The present invention segments multiple types of operating parameter sequences in different energy storage units separately, and performs multimodal data fusion on the segmented operating parameters, so as to fully extract fine-grained features of the operating parameters and establish associations between different types of operating parameters. This greatly improves the recognition accuracy of the operating state of the heterogeneous uninterruptible power supply cluster, further enhances the assessment performance of the battery management system, and effectively ensures the stable operation of the heterogeneous uninterruptible power supply cluster.

[0024] See Figure 1 ,The state assessment method for a heterogeneous uninterruptible power supply cluster includes the following steps:

[0025] S11. Obtain multiple types of operating parameter sequences of each energy storage unit in the heterogeneous uninterruptible power supply cluster within a preset sampling time.

[0026] A heterogeneous uninterruptible power supply (UPS) cluster consists of multiple UPS systems, each equipped with different types of energy storage units. Energy storage units refer to electrochemical cell modules used to store and release energy within an UPS system. These types include, but are not limited to, battery modules, battery clusters, or single cells. For example, a heterogeneous UPS cluster may include various types of energy storage units, such as ternary lithium batteries, lithium iron phosphate batteries, and lead-acid batteries. Each energy storage unit is equipped with multiple sensors to collect various operating parameters during operation. These operating parameters reflect the actual operating conditions or electrical characteristics of the energy storage unit, including but not limited to voltage, current, temperature, and internal resistance. Those skilled in the art can select appropriate operating parameters based on their suitability for specific application scenarios. To accurately assess the overall operating status of a heterogeneous UPS cluster, a sequence of multiple preset operating parameters for each energy storage unit in the cluster can be synchronously collected within a preset sampling period.

[0027] In an optional embodiment of the present invention, step S12 includes the following process: obtaining multiple types of initial operating parameter sequences for each energy storage unit in the heterogeneous uninterruptible power supply cluster within a preset sampling period; for each initial operating parameter sequence of each energy storage unit, filtering the initial operating parameter sequence to obtain a filtered operating parameter sequence.

[0028] Specifically, within a preset sampling period, sensor data is collected to obtain various preset operating parameters corresponding to each energy storage unit in the heterogeneous uninterruptible power supply cluster. These collected raw data are then arranged according to the acquisition time to form an initial operating parameter sequence corresponding to each parameter type. Each initial parameter sequence obtained for each energy storage unit is filtered to remove random noise from the data and improve data reliability. Filtering methods include, but are not limited to, sliding average or weighted moving average, and the specific methods are not limited here. Furthermore, the filtered parameter sequence can be normalized to unify the scale range of the data.

[0029] Furthermore, considering the possibility of missing or abnormal data during data collection, another optional embodiment of the present invention further performs the following processing on each operating parameter sequence: missing value filling and outlier filtering are performed on each operating parameter in the operating parameter sequence to improve the accuracy of subsequent data processing. Missing value filling can be performed using interpolation or sliding average filling, and outlier filtering can be performed using interquartile range or standard deviation. Those skilled in the art can adaptably select the corresponding data cleaning method based on actual task needs, and this is not limited here.

[0030] In an optional embodiment of the present invention, for each initial operating parameter sequence of each energy storage unit, the initial operating parameter sequence is filtered to obtain a filtered operating parameter sequence, including the following processing steps: starting from the starting position of the initial operating parameter sequence, multiple operating parameters are extracted based on a preset sliding window length, and the mean of all extracted operating parameters is calculated to obtain a parameter mean; the sliding window is moved according to a preset step size, and the above calculation is repeated until all operating parameters in the initial operating parameter sequence participate in at least one mean calculation; all parameter means are arranged in sequence to obtain a filtered operating parameter sequence.

[0031] Specifically, starting from the first operating parameter of the initial operating parameter sequence, continuous operating parameters are intercepted with a preset sliding window length (such as 3 sampling points), and the average of all operating parameters in the window is calculated as the smoothed output of the current window. The sliding window is shifted right by S positions according to the preset step size S (such as a step size of 1), and operating parameters of the same length are intercepted again and the average is calculated. Repeat the above process until each operating parameter in the initial operating parameter sequence participates in at least one mean calculation. The obtained series of parameter means are arranged in chronological order to form a filtered operating parameter sequence. The operating parameters are subjected to sliding average processing to remove the noise therein and retain the overall change trend of the operating parameters of the energy storage unit to provide a basis for subsequent similarity analysis.

[0032] S13. For each type of operating parameter sequence of each energy storage unit, the corresponding operating parameter sequence is segmented based on a mutation point detection algorithm to obtain multiple parameter subsequences.

[0033] Taking into account the sampling time, the operating parameter sequence may span a variety of different operating states, for example, it may include multiple charging and discharging processes, and even the load conditions corresponding to each charging and discharging process are different. If the operating parameter sequence is analyzed as a whole, the phased change characteristics within the sequence will be ignored, resulting in inaccurate subsequent analysis results. In order to improve the above problems, the present invention uses a mutation point detection algorithm to analyze each operating parameter sequence of each energy storage unit separately, identify the position where the statistical characteristics or distribution (such as mean, variance or trend, etc.) change significantly, and divide the entire operating parameter sequence into several shorter parameter subsequences. Each parameter subsequence has relatively consistent characteristic performance. Among them, the mutation point detection algorithm includes but is not limited to CUSUM, Bayesian change point detection method, rolling mean method, etc. Those skilled in the art can adaptably select the corresponding mutation point detection algorithm based on the actual data distribution characteristics, which will not be described in detail here.

[0034] In an optional embodiment of the present invention, step S13 includes the following processing:

[0035] Firstly, based on the mutation point detection algorithm, mutation point detection is performed on each operating parameter in the operating parameter sequence to obtain all mutation parameters in the operating parameter sequence.

[0036] Specifically, for each operating parameter sequence, the mean of the sequence is calculated, and the positive cumulative sum and negative cumulative sum of each operating parameter in the sequence are calculated using the CUSUM algorithm, as shown in formulas (1) and (2) respectively:

[0037]

[0038] in, are the positive cumulative sums of the t-th and t-1-th operating parameters, are the negative cumulative sums of the t-th and t-1-th operating parameters, respectively, x t is the t-th operating parameter in the parameter sequence, μ is the mean of the parameter sequence, k is the preset sensitivity factor, and the initial forward cumulative sum Negative cumulative sum Set to zero value. When a certain running parameter x f When the positive cumulative sum or negative cumulative sum of (f∈[1,N], N is the length of the corresponding parameter sequence) exceeds the preset threshold, the operating parameter x f By executing the above process in the entire operating parameter sequence, all mutation parameters can be detected.

[0039] Then, the running parameter sequence is segmented according to all mutation parameters to obtain multiple parameter subsequences.

[0040] Specifically, the entire operating parameter sequence is divided into several parameter subsequences, using all mutation parameters as demarcation points. The operating parameters within each parameter subsequence have relatively consistent statistical characteristics, accurately reflecting the operating status of the energy storage unit at the current stage. It is understood that when dividing the operating parameter sequence, mutation parameters can be classified as either the previous or the next parameter subsequence, without specific limitation.

[0041] S14. Calculate the similarity between each parameter subsequence based on the dynamic time warping algorithm and generate a similarity image.

[0042] Considering that different parameter subsequences may have different lengths, the Dynamic Time Warping (DTW) algorithm is used to calculate the similarity between any two parameter subsequences. The DTW algorithm calculates the minimum alignment distance between two parameter subsequences using a nonlinear alignment method, even when the lengths of the subsequences are inconsistent, thereby determining the similarity between the two parameter subsequences. After calculating the similarity between all parameter subsequences pairwise, a similarity graph is generated.

[0043] In an optional embodiment of the present invention, step S14 includes the following processing:

[0044] First, all parameter subsequences are numbered in sequence according to the type of the operating parameters.

[0045] Specifically, all parameter subsequences can be grouped according to the type of operating parameters to form multiple parameter subsequence sets. The following process is performed for each parameter subsequence set: based on the timestamp of the first operating parameter in each parameter subsequence, all parameter subsequences in the parameter subsequence set are arranged in ascending order according to chronological order. After all parameter subsequence sets have been processed according to the above process, all parameter subsequence sets are traversed according to the preset parameter type order (such as the order of current, voltage, and temperature), and starting from number 1, each parameter subsequence therein is assigned a unique number in a continuously increasing order. It should be noted that in order to subsequently construct a similarity image, the numbering is a global continuous numbering across types, such as the current subsequences are numbered 1-7, the voltage subsequences are numbered 8-15, and the temperature subsequences are numbered 16-24.

[0046] Then, the similarity between each parameter subsequence is calculated based on the dynamic time warping algorithm, and the numbers of the corresponding parameter subsequences are used as the row index and column index of the similarity.

[0047] For any two parameter subsequences, a distance matrix is ​​constructed based on the differences in the corresponding parameter values ​​at each time point in their sequences. The minimum path cost of the two parameter subsequences is solved using the DTW algorithm based on the distance matrix, and the similarity between the two parameter subsequences is determined accordingly.

[0048] Finally, all similarities are arranged in order according to the corresponding row index numbers and column index numbers to generate a similarity image.

[0049] The numbers of any two subsequences are used as row index numbers and column index numbers respectively, and their corresponding similarities are written into the corresponding positions in the matrix. After completing the similarity calculation between all subsequences and filling in the matrix, a complete similarity image can be obtained.

[0050] Furthermore, in an optional embodiment of the present invention, the similarity between any two parameter subsequences is generated according to the following steps:

[0051] First, the parameter value differences between the two parameter subsequences are calculated and the corresponding distance matrix is ​​constructed.

[0052] For any two parameter subsequences, compare the i-th parameter x of the first parameter subsequence point by point i and the jth parameter y of the second parameter subsequencej The difference between the two can be obtained by calculating the distance matrix, where the difference can be x i and y j The absolute value of the difference between the two can also be the square of the difference between the two, or other methods can be used, which are not limited here. According to the above method, all the running parameters of the two parameter subsequences are traversed to obtain the distance matrix D, where D(i,j) represents the i-th parameter x of the first parameter subsequence. i and the jth parameter y of the second parameter subsequence j degree of difference.

[0053] Then, based on the dynamic time warping algorithm, the alignment distance between the two parameter subsequences is obtained according to the distance matrix.

[0054] The dynamic programming strategy based on the DTW algorithm starts from the starting position of the distance matrix and gradually calculates the cumulative cost of each position. The cumulative cost of any position D(i,j) in the distance matrix is ​​determined by the sum of the difference value of the current position and the minimum cumulative cost of the previous position. After the cumulative cost of the distance matrix is ​​calculated, the DTW algorithm backtracks through the distance matrix to obtain a path with the minimum total cost. The cumulative cost at the end of the path is used as the alignment distance between the two parameter subsequences.

[0055] Finally, the alignment distance is normalized to obtain the similarity between the two parameter subsequences.

[0056] In order to enhance the comparability of alignment distances between subsequences with different parameters, the alignment distances are normalized to obtain standardized similarity.

[0057] S15. Input the similarity image into a battery status assessment model to generate a status assessment result of the heterogeneous uninterruptible power supply cluster; wherein the battery status assessment model is a convolutional neural network.

[0058] The similarity image is input into a battery status assessment model. Deep features in the image are extracted through multiple layers of convolution and pooling operations. A fully connected layer is used to generate a status assessment result for the heterogeneous uninterruptible power supply cluster. The assessment results include, but are not limited to, battery health status, remaining service life prediction, or fault type diagnosis. The battery status assessment model in the present invention can be based on, but is not limited to, lightweight convolutional neural networks such as MobileNet and SqueezeNet.

[0059] It should be noted that the battery status assessment model can generate assessment results for a single task, and can also generate assessment results for multiple different tasks. In a single-task scenario, the output dimension of the fully connected network can be modified to adapt to different tasks, such as battery health status estimation and remaining service life prediction as regression tasks (output dimension modified to 1), and fault diagnosis as a classification task (output dimension modified to the number of fault types). In a multi-task scenario, the battery status assessment model can use multiple single-task networks with different network structures to generate corresponding assessment results respectively, and can also generate assessment results for multiple different task types through a unified multi-task model, without specific limitations.

[0060] In an optional embodiment of the present invention, the evaluation results are multi-class, and the battery state assessment model includes a feature extraction network and a plurality of expert networks with different network structures cascaded therewith. Step S15 includes the following process:

[0061] First, the similarity image is input into the feature extraction network of the battery state assessment model to extract the shared image features.

[0062] When the evaluation results are multi-class, the battery state assessment model includes a feature extraction network and multiple expert networks, each corresponding to a different task type. The similarity image is input into the feature extraction network, where features are extracted from the image through multiple convolutional and pooling layers to obtain shared image features. Both the feature extraction network and the expert network are convolutional neural networks, and the specific network structure can be adaptively selected based on the actual task requirements.

[0063] The shared image features are then input into multiple expert networks of the battery state assessment model respectively. Each expert network extracts task features associated with its corresponding evaluation result type and generates an evaluation result of the corresponding type.

[0064] The shared image features are fed into multiple pre-defined expert networks within the battery state assessment model. Each expert network corresponds to a specific state assessment task type, such as battery health status, remaining useful life prediction, or fault type diagnosis. After receiving the shared image features, each expert network extracts task features associated with its corresponding assessment result type from the shared features. Based on these extracted task features, the network generates an assessment result of the corresponding type, enabling multi-dimensional state assessment of heterogeneous uninterruptible power supply clusters.

[0065] See Figure 2The status assessment device 200 for the heterogeneous uninterruptible power supply cluster includes: a parameter acquisition module 210, a mutation point detection module 220, an image generation module 230 and an assessment module 240. The above-mentioned parameter acquisition module 210 is used to obtain multiple types of operating parameter sequences of each energy storage unit in the heterogeneous uninterruptible power supply cluster within a preset sampling time. The mutation point detection module 220 is used to segment the corresponding operating parameter sequence of each type of operating parameter sequence of each energy storage unit based on the mutation point detection algorithm to obtain multiple parameter subsequences. The image generation module 230 is used to calculate the similarity between any two parameter subsequences based on the dynamic time warping algorithm to generate a similarity image. The assessment module 240 is used to input the similarity image into the battery status assessment model to generate the status assessment result of the heterogeneous uninterruptible power supply cluster; wherein, the battery status assessment model is a convolutional neural network.

[0066] The specific limitations of the heterogeneous uninterruptible power supply cluster status assessment device can be found in the limitations of the heterogeneous uninterruptible power supply cluster status assessment method described above and will not be further elaborated here. Each module in the heterogeneous uninterruptible power supply cluster status assessment device described above can be implemented in whole or in part via software, hardware, or a combination thereof. Each of these modules can be embedded in or independent of a processor in a computer device in hardware format, or stored in a computer device memory in software format, allowing the processor to invoke operations corresponding to each of these modules.

[0067] It should be noted that, in order to highlight the innovative part of the present invention, this embodiment does not introduce modules that are not closely related to solving the technical problem proposed by the present invention, but this does not mean that there are no other modules in this embodiment.

[0068] See Figure 3 The electronic device 3 may include a memory 32, a processor 33 and a bus, and may also include a computer program stored in the memory 32 and executable on the processor 33, such as a status assessment program for a heterogeneous uninterruptible power supply cluster.

[0069] The memory 32 includes at least one type of readable storage medium, including flash memory, a mobile hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 32 may be an internal storage unit of the electronic device 3, such as a mobile hard disk of the electronic device 3. In other embodiments, the memory 32 may also be an external storage device of the electronic device 3, such as a plug-in mobile hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 3. Furthermore, the memory 32 may include both an internal storage unit of the electronic device 3 and an external storage device. The memory 32 can be used not only to store application software installed on the electronic device 3 and various types of data, such as code for status assessment of a heterogeneous uninterruptible power supply cluster, but also to temporarily store data that has been output or is about to be output.

[0070] In some embodiments, the processor 33 may be comprised of an integrated circuit, such as a single packaged integrated circuit or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips. The processor 33 is the control core (Control Unit) of the electronic device 3. It utilizes various interfaces and circuits to connect the various components of the entire electronic device 3. It executes or runs programs or modules stored in the memory 32 (e.g., a status assessment program for a heterogeneous uninterruptible power supply cluster) and accesses data stored in the memory 32 to perform various functions of the electronic device 3 and process data.

[0071] The processor 33 executes the operating system and various installed application programs of the electronic device 3. The processor 33 executes the application programs to implement the steps in the above-mentioned method for assessing the status of a heterogeneous uninterruptible power supply cluster.

[0072] Exemplarily, the computer program may be divided into one or more modules, one or more of which are stored in the memory 32 and executed by the processor 33 to complete the present application. One or more modules may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device 3. For example, the computer program may be divided into a parameter acquisition module 210, a mutation point detection module 220, an image generation module 230, and an evaluation module 240.

[0073] The above-mentioned integrated unit implemented in the form of a software functional module can be stored in a computer-readable storage medium, which can be either non-volatile or volatile. The above-mentioned software functional module is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, computer device, or network device, etc.) or a processor to perform part of the functions of the heterogeneous uninterruptible power supply cluster status assessment method of each embodiment of the present application.

[0074] In summary, the present invention discloses a method and related device for assessing the status of a heterogeneous uninterruptible power supply (UPS) cluster. Within a preset sampling period, the method collects multiple operating parameter sequences from each energy storage unit in the UPS cluster. Using a mutation point detection algorithm, the sequences are divided into parameter subsequences representing different operating stages. The similarity of these parameter subsequences is calculated using a dynamic time warping algorithm. Multimodal data fusion is then performed on the different types of operating parameters to generate a similarity image. A battery status assessment model is then used to extract image features, resulting in a UPS cluster status assessment result. The present invention segments the multiple types of operating parameter sequences from different UPS units and then performs multimodal data fusion on the segmented operating parameters to fully extract fine-grained features and establish correlations between different types of operating parameters. This significantly improves the accuracy of identifying the operating status of the UPS cluster, further enhances the assessment performance of the battery management system, and effectively ensures the stable operation of the UPS system. Therefore, the present invention effectively overcomes the shortcomings of the prior art and has high industrial application value.

[0075] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the present invention. Anyone skilled in the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by one of ordinary skill in the art without departing from the spirit and technical principles disclosed herein are intended to be covered by the claims of the present invention.

Claims

1. A method for evaluating the status of a heterogeneous uninterruptible power supply cluster, characterized in that: The method comprises: Obtain multiple types of operating parameter sequences of each energy storage unit in a heterogeneous uninterruptible power supply cluster within a preset sampling time; For each type of operating parameter sequence of each energy storage unit, the corresponding operating parameter sequence is segmented based on the mutation point detection algorithm to obtain multiple parameter subsequences; Calculate the similarity between each parameter subsequence based on the dynamic time warping algorithm and generate a similarity image; The similarity image is input into a battery status assessment model to generate a status assessment result of the heterogeneous uninterruptible power supply cluster; wherein the battery status assessment model is a convolutional neural network.

2. The method for evaluating the status of a heterogeneous uninterruptible power supply cluster according to claim 1, wherein: The method of obtaining a plurality of operating parameter sequences of each energy storage unit in the heterogeneous uninterruptible power supply cluster within a preset sampling time period includes: Obtain multiple types of initial operating parameter sequences of each energy storage unit in a heterogeneous uninterruptible power supply cluster within a preset sampling time; For each initial operating parameter sequence of each energy storage unit, filtering is performed on the initial operating parameter sequence to obtain a filtered operating parameter sequence.

3. The method for evaluating the status of a heterogeneous uninterruptible power supply cluster according to claim 2, wherein: For each initial operating parameter sequence of each energy storage unit, the initial operating parameter sequence is filtered to obtain a filtered operating parameter sequence, including: Starting from the starting position of the initial operating parameter sequence, extracting multiple operating parameters based on a preset sliding window length, calculating the mean of all the extracted operating parameters, and obtaining a parameter mean; Moving the sliding window according to a preset step size, repeating the above calculation until all operating parameters in the initial operating parameter sequence participate in at least one mean calculation; Arrange all parameter means in order to obtain the filtered operating parameter sequence.

4. The evaluation method for heterogeneous uninterruptible power supply clusters according to claim 1, characterized in that: The mutation point detection algorithm is used to segment the corresponding operating parameter sequence to obtain multiple parameter subsequences, including: Based on the mutation point detection algorithm, mutation point detection is performed on each operating parameter in the operating parameter sequence to obtain all mutation parameters in the operating parameter sequence; The operating parameter sequence is divided according to all mutation parameters to obtain multiple parameter subsequences.

5. The method for evaluating the status of a heterogeneous uninterruptible power supply cluster according to claim 1, wherein: The method of calculating the similarity between each parameter subsequence based on the dynamic time warping algorithm and generating a similarity image includes: According to the type of operating parameters, all parameter subsequences are numbered in sequence; Calculate the similarity between each parameter subsequence based on the dynamic time warping algorithm, and use the numbers of the corresponding parameter subsequences as the row index and column index of the similarity respectively; Arrange all similarities in order according to the corresponding row index numbers and column index numbers to generate a similarity image.

6. The method for evaluating the status of a heterogeneous uninterruptible power supply cluster according to claim 1, wherein: The similarity between any two parameter subsequences is generated as follows: Calculate the parameter value difference between two parameter subsequences and construct the corresponding distance matrix; Based on the dynamic time warping algorithm, the alignment distance between two parameter subsequences is obtained according to the distance matrix; The alignment distance is normalized to obtain the similarity between the two parameter subsequences.

7. The method for evaluating the status of a heterogeneous uninterruptible power supply cluster according to claim 1, wherein: The evaluation results are multi-classified, the battery status evaluation model includes a feature extraction network and a plurality of expert networks with different network structures cascaded therewith, and the similarity image is input into the battery status evaluation model to generate the evaluation results of the heterogeneous uninterruptible power supply cluster, including: Inputting the similarity image into a feature extraction network of a battery state assessment model to extract shared image features; The shared image features are respectively input into multiple expert networks of the battery state assessment model. Each expert network extracts task features associated with its corresponding evaluation result type and generates an evaluation result of the corresponding type.

8. A device for evaluating the status of a heterogeneous uninterruptible power supply cluster, characterized in that: The device comprises: A parameter acquisition module is used to obtain multiple types of operating parameter sequences of each energy storage unit in a heterogeneous uninterruptible power supply cluster within a preset sampling time; A mutation point detection module is used to segment the corresponding operating parameter sequence of each type of operating parameter sequence of each energy storage unit based on the mutation point detection algorithm to obtain multiple parameter subsequences; An image generation module is used to calculate the similarity between any two parameter subsequences based on the dynamic time warping algorithm and generate a similarity image; An evaluation module is configured to input the similarity image into a battery status evaluation model to generate a status evaluation result of the heterogeneous uninterruptible power supply cluster; wherein the battery status evaluation model is a convolutional neural network.

9. An electronic device, characterized in that: The electronic device comprises: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the electronic device to implement the status assessment method for a heterogeneous uninterruptible power supply cluster as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor of a computer, the computer is caused to execute the method for assessing the status of a heterogeneous uninterruptible power supply cluster according to any one of claims 1 to 7.

Images