A partition control SOC balancing method and system in holographic data scenario

Abstract

The present invention relates to the field of battery management technology, and in particular to a partition control SOC balancing method and system in a holographic data scenario. The present invention inputs collected runtime data into a dynamic weight adjustment algorithm model based on MDSTFT to preprocess, extract features, and encode the runtime data to obtain local features, and obtain long-distance dependencies of the local features. The obtained long-distance dependencies are then processed to obtain corrected values ​​of the charge and discharge capacity weights and the SOC weights; according to the corrected values ​​of the charge and discharge capacity weights and the SOC weights, the charge and discharge capacity weights and the SOC weights are dynamically adjusted, which helps to flexibly adjust the weights between the charge and discharge capacity and the SOC, thereby improving the overall performance and reliability of the system.

Description

Technical Field

[0001] The present invention relates to the field of battery management technology, and in particular to a method and system for controlling SOC balancing by partitioning in a holographic data scenario. Background Art

[0002] Existing battery management systems often use fixed weights or manually set values ​​based on experience for charge and discharge capacity and SOC (state of charge) weight control strategies. However, with the development of existing battery equipment acquisition technology and the improvement of storage capacity, battery data mostly supports second-level or even millisecond-level acquisition, processing, and storage. Therefore, the current control strategy cannot flexibly adjust the weight between charge and discharge capacity and SOC when facing high-frequency actual changing operating conditions, resulting in a decrease in the overall system operating efficiency and battery life. Summary of the Invention

[0003] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a partition control SOC balancing method and system in a holographic data scenario, which helps to flexibly adjust the weight between charging and discharging capacity and SOC, thereby improving the overall performance and reliability of the system.

[0004] In order to implement the above technical solution, in a first aspect, the present invention provides a method for controlling SOC balancing by partitioning in a holographic data scenario, comprising:

[0005] Step 1: Collect the operating sequence data of the battery compartment;

[0006] Step 2: Input the collected runtime data into the dynamic weight adjustment algorithm model based on MDSTFT to output the charge and discharge capacity weight and SOC weight;

[0007] The second step includes:

[0008] Input the collected battery compartment operating sequence data into the dynamic weight adjustment algorithm model based on MDSTFT;

[0009] Preprocessing the input battery compartment operation sequence data;

[0010] Perform feature extraction and encoding on the preprocessed runtime data to capture local features in the runtime data;

[0011] Based on the acquired local features, the long-distance dependencies of the runtime series data are obtained;

[0012] The obtained long-distance dependency is processed to obtain a correction value of the charge and discharge capacity weight and a correction value of the SOC weight;

[0013] Dynamically adjusting the charge-discharge capacity weight and the SOC weight based on the obtained correction value of the charge-discharge capacity weight and the correction value of the SOC weight;

[0014] Output the dynamically adjusted charge and discharge capability weight and SOC weight.

[0015] Furthermore, the method further comprises:

[0016] Based on the output charge and discharge capacity weight and SOC weight, calculate the SOC energy loss function and charge and discharge loss function;

[0017] Based on the calculated SOC energy loss function and charge and discharge loss function, the model parameters of the dynamic weight adjustment algorithm based on MDSTFT are optimized.

[0018] Furthermore, the pre-processing of the inputted battery compartment operation sequence data includes:

[0019] Clean the input battery compartment runtime data to remove outliers and missing values;

[0020] Normalize real-time runtime data to remove outliers and missing values.

[0021] Furthermore, the method further comprises, before step 1:

[0022] Collect battery compartment operation data;

[0023] estimating the state of charge and state of health of the battery compartment based on the collected battery compartment operation data, and calculating a comprehensive health index of the battery compartment based on the estimated state of charge and state of health of the battery compartment;

[0024] Marking the health status of the battery according to the calculated comprehensive health index, where the health status includes normal and abnormal;

[0025] The battery compartment battery charging and discharging paths are dynamically adjusted and optimized based on the results of battery health diagnosis, and real-time dynamic adjustments are made to respond to changes in battery status.

[0026] In a second aspect, the present invention provides a partition control SOC balancing system in a holographic data scenario, comprising:

[0027] A data acquisition device, used for collecting operating sequence data of the battery compartment;

[0028] A dynamic weight adjustment algorithm model based on MDSTFT is used to process the collected battery compartment operating sequence data to obtain the charge and discharge capacity weight and SOC weight;

[0029] The dynamic weight adjustment algorithm model based on MDSTFT includes:

[0030] The input layer is used to input the collected battery compartment runtime data into the dynamic weight adjustment algorithm model based on MDSTFT;

[0031] The preprocessing layer is used to preprocess the input battery compartment runtime data;

[0032] One-dimensional convolutional layer, used to extract and encode features from preprocessed runtime data to obtain local features;

[0033] The multi-head self-attention layer is used to obtain long-range dependencies of runtime data based on local features;

[0034] A multilayer perceptron is used to process the obtained long-distance dependency to obtain a correction value of the charge and discharge capacity weight and a correction value of the SOC weight;

[0035] A dynamic weight adjustment layer dynamically adjusts the charge and discharge capacity weight and the SOC weight based on the obtained correction value of the charge and discharge capacity weight and the correction value of the SOC weight;

[0036] The output layer is used to output the dynamically adjusted charge and discharge capacity weights and SOC weights.

[0037] Furthermore, the system further comprises:

[0038] A loss function calculation module is used to calculate the SOC energy loss function and the charge and discharge loss function based on the output charge and discharge capacity weight and SOC weight;

[0039] The optimization module is used to optimize the parameters of the dynamic weight adjustment algorithm model based on MDSTFT according to the calculated SOC energy loss function and charge and discharge loss function.

[0040] Furthermore, the system further comprises:

[0041] A data collection module, used to collect battery compartment operation data;

[0042] A battery compartment comprehensive health index evaluation module is used to estimate the state of charge and health status of the battery compartment based on the collected battery compartment operation data, and calculate the comprehensive health index of the battery compartment based on the estimated state of charge and health status of the battery compartment;

[0043] A battery health diagnosis module is used to mark the health status of the battery based on the calculated comprehensive health index, where the health status includes normal and abnormal;

[0044] The dynamic adjustment module is used to dynamically adjust and optimize the battery compartment battery charging and discharging paths based on the results of battery health diagnosis, and make real-time dynamic adjustments to respond to changes in battery status.

[0045] In a third aspect, the present invention provides a computer-readable storage medium, which includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the partition control SOC balancing method in the holographic data scenario described above.

[0046] The beneficial effects of the present invention are:

[0047] (1) The present invention inputs the collected runtime data into a dynamic weight adjustment algorithm model based on MDSTFT, so that the dynamic weight adjustment algorithm model based on MDSTFT outputs the charge and discharge capacity weight and the SOC weight, which helps to flexibly adjust the weight between the charge and discharge capacity and the SOC, thereby improving the overall performance and reliability of the system.

[0048] (2) The runtime data collected usually captures and displays multi-dimensional information of the battery operating status more comprehensively, helping technicians better understand and analyze the health status and performance of the battery, and can reduce the impact on the battery compartment life when balancing the SOC of different battery cells. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0050] Figure 1 This is a flow chart of a partition control SOC balancing method in a holographic data scenario of the present invention.

[0051] Figure 2 This is a workflow diagram of the dynamic weight adjustment algorithm model based on MDSTFT of the present invention. DETAILED DESCRIPTION

[0052] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0053] It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the present invention. Unless otherwise specified, each technical and scientific term used in this embodiment has the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0054] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0055] In the present invention, terms such as "upper", "lower", "left", "right", "front", "back", "vertical", "horizontal", "side", "bottom", etc. indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are relational words determined only for the convenience of describing the structural relationships of the various parts or elements of the present invention, and do not specifically refer to any part or element in the present invention, and should not be understood as limiting the present invention.

[0056] In the present invention, terms such as "fixed connection," "connected," and "connection" should be interpreted broadly to mean a fixed connection, an integral connection, or a detachable connection; a direct connection or an indirect connection through an intermediary. Relevant researchers or technicians in this field may determine the specific meanings of these terms in the present invention based on specific circumstances, and they should not be construed as limitations of the present invention.

[0057] The dynamic weight adjustment algorithm model based on MDSTFT (Multi-Dimensional Sparse Time Series Fusion Transformer) used in this article is a deep learning model specifically designed to process high-frequency time series data. The core idea of ​​this model is to combine time series data analysis and sub-attention mechanism to fully explore the characteristics of data in both time and frequency dimensions, thereby achieving accurate modeling and prediction of complex time series data.

[0058] The model has three core mechanisms:

[0059] Multidimensional Sparse Time Series Analysis: This model first performs time-frequency analysis on the input high-frequency time series data, fusing time series signals such as the battery's charge and discharge status, SOC status, and power input and output from the time and frequency domains to capture the periodic and sudden changes in the time series data. Furthermore, to handle the massive amount of high-frequency time series data generated by energy storage power plants, a sparse representation method is introduced, retaining only the key frequency components that have the greatest impact on prediction, thereby reducing computational complexity and storage requirements.

[0060] Multi-head self-attention mechanism: In the improved MDSTFT, the self-attention mechanism is used to capture dependencies between different time steps in the time series. By calculating the attention weights between each time step and the others, the model can automatically learn and identify important patterns and features in the data. To enhance the model's expressiveness, a multi-head self-attention mechanism is employed, allowing the model to process attention in parallel across different subspaces, thereby capturing richer feature information and more efficiently processing the multi-dimensional features of the input data.

[0061] Space-time encoder: The space-time encoder in this algorithm combines the characteristics of both time and frequency dimensions, and generates a dynamic weight adjustment strategy that adapts to the current working conditions by encoding high-frequency data. Specifically, the model adjusts the charge and discharge capacity weights and SOC weights of each battery cell in real time based on the historical data and real-time data of the battery to ensure that the SOC is balanced at all times and to improve the overall charge and discharge efficiency of the system. In addition, the dynamic weight adjustment of this patent is optimized through a composite objective function. This objective function combines SOC balance and system efficiency. In actual operation, the model adjusts the charge and discharge capacity and SOC weights to achieve the optimal operating state of the battery system under high-frequency dynamic changes.

[0062] Example 1:

[0063] like Figure 1 As shown, this embodiment provides a partition control SOC balancing method in a holographic data scenario, including the following steps:

[0064] S1: Collect the operating sequence data of the battery compartment, where the operating sequence data includes: the working status of the energy storage converter (PCS), the SOC status, the active power of the battery compartment, the active power instruction of the battery compartment, the maximum charging capacity of the battery compartment, the maximum discharging capacity of the battery compartment and the comprehensive health index of the battery compartment.

[0065] Specifically, since the runtime data is sequence data, for the convenience of explanation and understanding, the real-time runtime data at the current time t is represented as :

[0066]

[0067] in, Time series data indicating the working status of the energy storage converter; Time series data indicating the SOC state; Time series data indicating the active power of the battery compartment; Indicates the timing data of the battery compartment active power command; Time series data indicating the maximum charging capacity of the battery compartment; Time series data indicating the maximum discharge capacity of the battery compartment; Time series data representing the comprehensive health index of the battery compartment.

[0068] S2: Input the collected battery compartment operation sequence data into the dynamic weight adjustment algorithm model based on MDSTFT, and process the collected battery compartment operation sequence data to obtain the charge and discharge capacity weight and SOC weight.

[0069] like Figure 2 As shown, the specific steps include:

[0070] S21: Preprocessing the inputted battery compartment operation sequence data.

[0071] The specific steps include:

[0072] S21-1: Clean the collected real-time runtime data of the battery compartment to remove outliers and missing values.

[0073] Specifically:

[0074] A) Use k-means clustering method to detect boundary points and non-cluster values, and identify outliers and missing values;

[0075] B) For the identified outliers, we delete and then interpolate them.

[0076] C) For missing values, Lagrange polynomial interpolation method is used to fill in the blanks.

[0077] S21-2: Normalize the cleaned real-time runtime data.

[0078] Specifically, the cleaned real-time operation data is normalized to be between 0 and 1.

[0079] Among them, the preprocessed runtime data is expressed as :

[0080]

[0081] in, Represents the timing data of the PCS working status after preprocessing, is the time series data of the SOC state after preprocessing, is the time series data of the active power of the battery compartment after preprocessing, Represents the time series data of the active power instruction of the battery compartment after preprocessing, is the time series data of the maximum charging capacity of the battery compartment after preprocessing, Represents the time series data of the maximum discharge capacity after preprocessing, Represents the time series data of the comprehensive health index of the battery compartment after preprocessing.

[0082] S22: Perform feature extraction and encoding on the preprocessed runtime series data to capture local features of the runtime series data.

[0083] Specifically, the following steps are performed through the convolutional layer (CNN) of the dynamic weight adjustment algorithm model based on the MDSTFT (Multi-Dimensional Sparse Temporal Fusion Transformer):

[0084] The preprocessed real-time runtime data is passed through a one-dimensional convolutional layer (1D CNN) for feature extraction to capture the local features of the real-time runtime data.

[0085] The feature extraction formula for runtime series data is as follows:

[0086] ;

[0087] in, is the extracted local feature matrix.

[0088] S23: Based on the acquired local features, long-distance dependencies in the runtime series data are obtained.

[0089] Specifically, each feature in the local feature matrix is ​​processed in parallel to obtain the long-distance dependency relationship between each feature in the local feature matrix.

[0090] The specific steps include:

[0091] A) Calculate the relationship between each local feature and other local features in the extracted local feature matrix (i.e., attention weight), and generate an enhanced representation of each local feature based on the calculation results.

[0092] Specifically, assuming that at the current The local feature matrix at time Include local features, and the dimension of each local feature is , then the local feature matrix It can be expressed as: ;

[0093] The local feature matrix Mapped to matrix, Matrix and Matrix, where ; ; .

[0094] in, , and Both represent trainable attention weight matrices, and , and All belong to ,in, is the dimension of the trainable attention weight matrix.

[0095] B) Normalize the trainable attention weight matrix through the Softmax function:

[0096]

[0097] where represents the scaling factor used to mitigate the The problem of too large dot product caused by the increase.

[0098] C) A multi-head self-attention mechanism is used to parallelly compute multiple different attention heads with normalized trainable attention weight matrices to focus on different information in different subspaces, thereby enhancing the expressiveness of the model.

[0099] For example, suppose the trainable attention weight matrix has Attention heads, each with its own independent Query, Key, Value matrix , then The Query, Key, and Value matrices of each head are 、 and ,and 、 and ∈ , then The output of an attention head is:

[0100]

[0101] Wherein, 1≤i≤h, h is a positive integer.

[0102] D) Concatenate the outputs of all attention heads through a linear transformation matrix , in order to obtain the long-distance dependencies of each local feature in the runtime series data.

[0103] Among them, the long-distance dependency relationship between local features is as follows:

[0104] ;

[0105] S24: Processing the obtained long-distance dependency to obtain a correction value of the charge-discharge capability weight and a correction value of the SOC weight.

[0106] Specifically, the multilayer perceptron obtains the long-distance dependencies between the features. Get the charge and discharge capacity weight Correction value and SOC weight Correction value .

[0107] S25: Dynamically adjust the charge-discharge capability weight and the SOC weight based on the obtained correction value of the charge-discharge capability weight and the correction value of the SOC weight.

[0108] The dynamic adjustment formula is:

[0109]

[0110] in, and Represents the coefficient.

[0111] S26: Output the dynamically adjusted charge and discharge capability weight and SOC weight.

[0112] S3: Based on the output charge and discharge capability weight and SOC weight, calculate the SOC energy loss function and the charge and discharge loss function.

[0113] S31: Calculate the SOC energy loss function. The specific formula is as follows:

[0114]

[0115] in, Indicates the degree of deviation of the battery compartment K from the group equilibrium, is the adjusted SOC weight.

[0116] S32: Calculate the charge and discharge energy loss function. The specific formula is as follows:

[0117]

[0118] in, Represents the charging efficiency of battery compartment K at time t. It represents the discharge instruction execution efficiency of battery compartment K at time t.

[0119] S4: Based on the calculated SOC energy loss function and charge and discharge loss function, the model parameters of the dynamic weight adjustment algorithm based on MDSTFT are optimized.

[0120] Example 2:

[0121] This embodiment provides a method for controlling SOC balancing by partition in a holographic data scenario. The method further includes collecting and calculating a comprehensive health index of a battery compartment before executing step S1. Specifically, the method includes the following steps:

[0122] Q1: Collect battery compartment operating data; it should be noted that the battery compartment operating data includes battery compartment current, voltage and temperature, etc., which provide basic data for subsequent estimation.

[0123] Specifically, first determine the type of data that needs to be collected, and set the frequency and time interval for data collection, and then perform data collection; it should be noted that data collection is completed by installing necessary sensors, such as voltage sensors, current sensors and temperature sensors, and the type of data that needs to be collected includes but is not limited to battery voltage, current, charge and discharge status, temperature, charge and discharge time, etc.

[0124] It should also be noted that setting the frequency and time interval of data collection can ensure the continuity and integrity of the data;

[0125] The collected data is then preprocessed. It should be noted that in this embodiment, data preprocessing includes data cleaning and data conversion, wherein data cleaning includes missing value processing: using the mean, median, or mode to fill missing values ​​in numerical data; for categorical data, the mode or specific mark can be used, and outliers can also be identified using box plots, Z-scores, or IQR (interquartile range) methods. For detected outliers, the detected outliers can be deleted, replaced, or further analyzed to determine their causes;

[0126] Data conversion includes standardization, which scales the collected data to a uniform range.

[0127] Q2: Based on the collected battery compartment operation data, estimate the charge state and health state of the battery compartment, and calculate the comprehensive health index of the battery compartment based on the estimated charge state and health state of the battery compartment.

[0128] Q21: Get a real-time estimate of the state of charge of the battery compartment battery .

[0129] A) Obtain the current integral of the battery compartment battery and the nominal capacity of the battery ;

[0130] It should be noted that, in this embodiment, the current integral of the battery compartment battery It is the amount of charge and discharge of the battery from the last real-time estimated state of charge measurement to the current time interval, and is obtained by multiplying the battery current by the time interval.

[0131] B) Based on the current integration of the battery compartment battery and the nominal capacity of the battery , calculate the real-time estimated value S of the battery state of charge, the specific calculation formula is as follows:

[0132] ;

[0133] Where n is the charge and discharge efficiency of the battery, This is the real-time estimated state of charge obtained from the last measurement. It should be noted that the discharge efficiency is usually 1, and the charge efficiency is slightly less than 1.

[0134] Q22: Obtain the estimated health status H of the battery in the battery compartment.

[0135] A) Obtain the capacity attenuation value of the battery in the battery compartment .

[0136] It should be noted that the available capacity of the battery decays over time. In this embodiment, the capacity decay value The method of obtaining the battery can be to divide the current available capacity of the battery by the initial capacity of the battery;

[0137] B) Obtain the internal resistance aging value of the battery compartment battery .

[0138] It should be noted that the internal resistance of the battery increases with aging. In this embodiment, the capacity attenuation value The method of obtaining the battery can be to divide the current internal resistance of the battery by the initial internal resistance of the battery;

[0139] C) Get the cycle value of the battery in the battery compartment .

[0140] It should be noted that the cycle life of the battery is related to its health status. In this embodiment, the capacity decay value The method of obtaining the cycle life of the battery can be divided by the initial cycle life of the battery.

[0141] D) Based on the obtained capacity decay value , internal resistance aging value and loop value , calculate the real-time estimated value of health status H, the specific calculation formula is as follows:

[0142]

[0143] Q23: Based on the real-time estimated state of health value H and the real-time estimated state of charge value S, the battery health index Q is calculated using the following formula.

[0144] ,in, and Represents known parameters.

[0145] Q24: Obtain the performance degradation index B of the battery in the battery compartment.

[0146] It should be noted that the battery performance degradation index B is calculated based on parameters such as battery capacity attenuation and internal resistance increase. In this embodiment, the battery performance degradation index is obtained by subtracting the real-time estimated value S of the battery's state of charge from 1 and dividing it by the health status value of the battery in a brand new state. The health status value of the battery in a brand new state is usually 1.

[0147] Q25: Based on the battery performance degradation index B and the battery health index Q, calculate the comprehensive health index of the battery compartment. The specific formula is as follows:

[0148] ,in, Comprehensive health index of the battery compartment .

[0149] In this embodiment, the method further includes: marking the health status of the battery according to the calculated comprehensive health index, wherein the health status includes normal and abnormal.

[0150] Specifically, a threshold range of the comprehensive health index of the battery compartment is set in advance, and the comprehensive health index of the battery compartment is The battery compartment's comprehensive health index is compared with the preset threshold range. If the battery compartment's comprehensive health index is If the battery compartment's comprehensive health index is greater than the threshold range, the battery's health status is marked as good. If the battery compartment's comprehensive health index is within the threshold range, the battery's health status is marked as normal. If the value is less than the threshold range of the comprehensive health index of the battery compartment, the health status of the battery is marked as abnormal.

[0151] In this embodiment, the method further includes:

[0152] The battery compartment battery charging and discharging paths are dynamically adjusted and optimized based on the results of battery health diagnosis, and real-time dynamic adjustments are made to respond to changes in battery status.

[0153] Specifically, the method includes the following steps:

[0154] A) If the battery is in good health, increase the charging current; this can further shorten the charging time.

[0155] B) If the battery's health status is normal, reduce the charging current; it should be noted that this can extend the battery life.

[0156] C) If the battery health status is abnormal, the charging current is further reduced. It should be noted that the above method can protect the battery.

[0157] It should also be noted that the specific way to obtain the above charging current is to multiply the standard charging current by the adjustment coefficient. The adjustment coefficient is obtained by adjusting the comprehensive health index of the battery compartment. Multiply by the second coefficient and add the first coefficient to obtain the comprehensive health index The difference between the median value in the threshold range of the comprehensive health index of the battery compartment, the first coefficient is a constant determined according to historical data, in the initial state, the first coefficient is 1, and in this embodiment is 0.986.

[0158] More specifically, the steps of adjusting the depth of discharge according to the current state and remaining state of charge of the battery to optimize the use of the battery and extend its life, and to prevent the battery from being over-discharged are as follows:

[0159] ;

[0160] in, is the depth of discharge, Is the lowest state of charge allowed for the battery. is the adjustment coefficient, D is the initial discharge depth, and S is the real-time estimated state of charge.

[0161] Among them, the adjustment coefficient is obtained according to the following formula:

[0162]

[0163] Among them, and are preset adjustment coefficients, corresponding to the health status of the battery. is the adjustment coefficient when the battery is in good condition, is the adjustment coefficient when the battery is in an abnormal state, k is the parameter of the control function steepness, and N is the middle value of the threshold range of the comprehensive health index of the battery compartment. It should be noted that in this embodiment, the parameter of the control function steepness depends on the type of battery, the health status of the battery, the charge and discharge history of the battery, and the temperature range of the battery. In this embodiment, the value of k is 1.023. The value of is 0.893, The value of is 0.293.

[0164] Embodiment 3: This embodiment provides a partition control SOC balancing system in a holographic data scenario, including:

[0165] (1) Data acquisition module, used to collect the operating sequence data of the battery compartment.

[0166] (2) A dynamic weight adjustment algorithm model based on MDSTFT is used to process the collected battery compartment operating sequence data to obtain the charge and discharge capacity weight and SOC weight.

[0167] The dynamic weight adjustment algorithm model based on MDSTFT includes:

[0168] (1) Input layer, used to input the collected battery compartment runtime data into the dynamic weight adjustment algorithm model based on MDSTFT.

[0169] (2) Preprocessing layer, used to preprocess the input battery compartment runtime data.

[0170] Among them, the pre-processing layer includes a data cleaning unit and a data processing unit;

[0171] The data cleaning unit is used to perform boundary point detection and non-cluster value detection on the input battery compartment operation sequence data using the k-means clustering method, identify abnormal values ​​and missing values, and process the identified abnormal values ​​and missing values.

[0172] The normalization unit is used to normalize the cleaned real-time runtime data.

[0173] (3) One-dimensional convolutional layer, used to extract and encode the preprocessed runtime data to obtain local features.

[0174] (4) Multi-head self-attention layer, which is used to obtain long-distance dependencies in runtime series data based on local features.

[0175] (5) A multilayer perceptron is used to process the obtained long-distance dependency to obtain a correction value of the charge and discharge capacity weight and a correction value of the SOC weight.

[0176] (6) A dynamic weight adjustment layer dynamically adjusts the charge-discharge capacity weight and the SOC weight based on the obtained correction value of the charge-discharge capacity weight and the correction value of the SOC weight.

[0177] (7) Output layer, used to output the dynamically adjusted charge and discharge capacity weights and SOC weights.

[0178] (3) A loss function calculation module, which is used to calculate the SOC energy loss function and the charge and discharge loss function based on the output charge and discharge capacity weight and SOC weight.

[0179] (4) An optimization module, used to optimize the parameters of the dynamic weight adjustment algorithm model based on MDSTFT according to the calculated SOC energy loss function and charge and discharge loss function.

[0180] (5) A data collection module, used to collect battery compartment operation data.

[0181] (6) a battery compartment comprehensive health index assessment module, used to estimate the state of charge and health status of the battery compartment based on the collected battery compartment operation data, and calculate the comprehensive health index of the battery compartment based on the estimated state of charge and health status of the battery compartment;

[0182] (7) A battery health diagnosis module, which marks the health status of the battery based on the calculated comprehensive health index, where the health status includes normal and abnormal.

[0183] (8) A dynamic adjustment module, which is used to dynamically adjust and optimize the battery compartment battery charging and discharging paths based on the results of battery health diagnosis, and to make real-time dynamic adjustments to respond to changes in battery status.

[0184] Example 3:

[0185] This embodiment provides a computer-readable storage medium, which includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the partition control SOC balancing method in a holographic data scenario described in Example 1 and Example 2.

[0186] In this specification, the same or similar parts between the various embodiments can be referred to each other. In particular, for the terminal embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant support can be referred to the description in the method embodiment.

[0187] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interface, indirect coupling or communication connection of the system or unit, which can be electrical, mechanical or other forms.

[0188] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0189] In addition, it should be noted that the flowcharts in the accompanying drawings show the methods of the embodiments of the present disclosure. In the descriptions corresponding to the flowcharts or block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in an order different from that disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps can actually be performed substantially in parallel, and they can sometimes be performed in the opposite order, which may depend on the functions involved. Each block in the block diagram and / or flow chart, and the combination of blocks in the block diagram and / or flow chart, can be implemented using a dedicated hardware-based system that performs the specified functions or actions, or can be implemented using a combination of dedicated hardware and computer instructions.

[0190] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for controlling SOC balancing by partitioning in a holographic data scenario, characterized in that: include: Step 1: Collect the operating sequence data of the battery compartment; Step 2: Input the collected runtime data into the dynamic weight adjustment algorithm model based on MDSTFT to output the charge and discharge capacity weight and SOC weight; The second step includes: Input the collected battery compartment operating sequence data into the dynamic weight adjustment algorithm model based on MDSTFT; Preprocessing the input battery compartment operation sequence data; Perform feature extraction and encoding on the preprocessed runtime data to capture local features in the runtime data; Based on the acquired local features, the long-distance dependencies of the runtime series data are obtained; The obtained long-distance dependency is processed to obtain a correction value of the charge and discharge capacity weight and a correction value of the SOC weight; Dynamically adjusting the charge-discharge capacity weight and the SOC weight based on the obtained correction value of the charge-discharge capacity weight and the correction value of the SOC weight; Output dynamically adjusted charge and discharge capacity weight and SOC weight; The method of obtaining long-distance dependencies in the runtime series data based on the acquired local features includes: Calculating the relationship between each local feature and other local features in the extracted local feature matrix, and generating an enhanced representation of each local feature based on the calculation result; Normalize the trainable attention weight matrix through the Softmax function; A multi-head self-attention mechanism is used to parallelly compute multiple different attention heads with normalized trainable attention weight matrices to focus on different information in different subspaces. The outputs of all attention heads are concatenated and transformed linearly to obtain the long-range dependencies of local features in the runtime data.

2. The method for controlling SOC balancing by partition in a holographic data scenario according to claim 1, characterized in that: The method further comprises: Based on the output charge and discharge capacity weight and SOC weight, calculate the SOC energy loss function and charge and discharge loss function; Based on the calculated SOC energy loss function and charge and discharge loss function, the model parameters of the dynamic weight adjustment algorithm based on MDSTFT are optimized.

3. The method for controlling SOC balancing by partition in a holographic data scenario according to claim 1, characterized in that: The preprocessing of the inputted battery compartment operation sequence data includes: Clean the input battery compartment runtime data to remove outliers and missing values; Normalize real-time runtime data to remove outliers and missing values.

4. The method for controlling SOC balancing by partition in a holographic data scenario according to claim 2, characterized in that: The method further comprises, before step 1: Collect battery compartment operation data; estimating the state of charge and state of health of the battery compartment based on the collected battery compartment operation data, and calculating a comprehensive health index of the battery compartment based on the estimated state of charge and state of health of the battery compartment; Marking the battery's health status based on the calculated comprehensive health index, where the health status includes normal and abnormal; The battery compartment battery charging and discharging paths are dynamically adjusted and optimized based on the results of battery health diagnosis, and real-time dynamic adjustments are made to respond to changes in battery status.

5. A zone-controlled SOC balancing system in a holographic data scenario, characterized in that: include: Data acquisition equipment, used to collect operating sequence data of the battery compartment; A dynamic weight adjustment algorithm model based on MDSTFT is used to process the collected battery compartment operating sequence data to obtain the charge and discharge capacity weight and SOC weight; The dynamic weight adjustment algorithm model based on MDSTFT includes: The input layer is used to input the collected battery compartment runtime data into the dynamic weight adjustment algorithm model based on MDSTFT; The preprocessing layer is used to preprocess the input battery compartment runtime data; One-dimensional convolutional layer, used to extract and encode features from preprocessed runtime data to obtain local features; The multi-head self-attention layer is used to obtain long-range dependencies of runtime data based on local features; A multilayer perceptron is used to process the obtained long-distance dependency to obtain a correction value of the charge and discharge capacity weight and a correction value of the SOC weight; A dynamic weight adjustment layer dynamically adjusts the charge and discharge capacity weight and the SOC weight based on the obtained correction value of the charge and discharge capacity weight and the correction value of the SOC weight; The output layer is used to output the dynamically adjusted charge and discharge capacity weights and SOC weights; The method of obtaining long-distance dependencies in the runtime series data based on the acquired local features includes: Calculating the relationship between each local feature and other local features in the extracted local feature matrix, and generating an enhanced representation of each local feature based on the calculation result; Normalize the trainable attention weight matrix through the Softmax function; A multi-head self-attention mechanism is used to parallelly compute multiple different attention heads with normalized trainable attention weight matrices to focus on different information in different subspaces. The outputs of all attention heads are concatenated and transformed linearly to obtain the long-range dependencies of local features in the runtime data.

6. The zone-controlled SOC balancing system in the holographic data scenario according to claim 5 is characterized in that: The system further comprises: A loss function calculation module is used to calculate the SOC energy loss function and the charge and discharge loss function based on the output charge and discharge capacity weight and SOC weight; The optimization module is used to optimize the parameters of the dynamic weight adjustment algorithm model based on MDSTFT according to the calculated SOC energy loss function and charge and discharge loss function.

7. The zone-controlled SOC balancing system in the holographic data scenario according to claim 6 is characterized in that: The system further comprises: A data collection module, used to collect battery compartment operation data; A battery compartment comprehensive health index evaluation module is used to estimate the state of charge and health status of the battery compartment based on the collected battery compartment operation data, and calculate the comprehensive health index of the battery compartment based on the estimated state of charge and health status of the battery compartment; A battery health diagnosis module is used to mark the health status of the battery according to the calculated comprehensive health index, where the health status includes normal and abnormal; The dynamic adjustment module is used to dynamically adjust and optimize the battery compartment battery charging and discharging paths based on the results of battery health diagnosis, and make real-time dynamic adjustments to respond to changes in battery status.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the partition control SOC balancing method in the holographic data scenario according to any one of claims 1 to 4.

Images